fshy / Compute / Crawling / Databases and KV / Egress and tunnels / IP and proxies / Storage
google.com · Compute, Crawling, Databases and KV, Egress and tunnels, IP and proxies, Storage · 63 records · JSON
- Cloud Firestore Databases and KV★ qualifiesemailno cardDrequest-rate
- Google Workspace for Education Fundamentals (pooled tenant) Storage★ qualifiesinstitutionno cardRper-object-cap
- Firebase Hosting Computecaveatemailno cardDvolume-quota
- Google Compute Engine (Linux VMs) ComputecaveatcardcardDrequest-ratedisagrees
- Google Compute Engine GPU VMs Computecaveatemail?card?Dwrite-gated
- Google Compute Engine (Windows Server) Computecaveatemail?card?Drequest-ratedisagrees
- GKE Agent Sandbox Computecaveatemail?card?U
- Google Agent Runtime / Agent Engine Computecaveatemail?card?Drequest-rate
- Google Cloud Build Computecaveatemail?card?Drequest-rate
- Google Cloud Free Tier (e2-micro) ComputecaveatcardcardDper-ip-shaping
- Google Cloud Run ComputecaveatcardcardDvolume-quota
- Google Cloud Shell Computecaveatemailno cardDvolume-quota
- Google Cloud Spot ComputecaveatcardcardD
- Google Cloud Text-to-Speech ComputecaveatcardcardDvolume-quota
- Google Colab Computecaveatemailno cardDvolume-quotadisagrees
- Google Open Source Computecaveatemailno cardDwrite-gated
- Google Summer Computecaveatemail?card?Dwrite-gated
- Google Agent Platform Sandbox (Code Execution / Shell / Computer Use) Computecaveatemail?card?Rwrite-gated
- Gemini Notebook (formerly NotebookLM) — source corpus Storagecaveatemailno cardRper-object-cap
- Google Chat spaces as a quota-exempt attachment store (Chat API media.upload/media.download) Storagecaveatinstitutionno cardRper-object-cap
- Google Drive shared drives written by service accounts (the SA chain, corrected) Storagecaveatinstitutionno cardRper-ip-shaping
- Google for Startups Cloud Program (Start / Scale tiers) Storagecaveatinstitutionno cardRper-object-cap
- Google Play Books Partner Center Storagecaveatemailno cardRper-object-cap
- Public Datasets hosting programme (Google Cloud) Storagecaveatcard?cardRwrite-gated
- Google Cloud Free Trial (USD 300) ComputetrialcardcardphoneDvolume-quota
- Google Cloud external passthrough Network Load Balancer Egress and tunnelspaidcardcardDwrite-gated
- Google Computefails: identityinstitutionno cardDwrite-gated
- Google Cloud research Computefails: identityinstitutionno cardDwrite-gated
- Google Cloud Free Tier (e2-micro) Egress and tunnelsfails: egresscardcardDvolume-quota
- Google Cloud Run Egress and tunnelsfails: egresscardcardIvolume-quota
- Firebase Realtime Database Storagefails: egressemail?card?Rper-ip-shaping
- Firebase Storage (Cloud Storage for Firebase) Storagefails: egressemail?card?Rper-ip-shaping
- Google Artifact Registry (incl. the gcr.io domain) Storagefails: egresscountry-phoneno cardphoneRdisagrees
- google artifact registry Storagefails: egressemail?card?Rdisagrees
- google cloud source repositories Storagefails: egressemail?card?R
- Google Colab (paid tiers) Computefails: quotacardcardDrequest-ratedisagrees
- Google Cloud Free Trial Egress and tunnelsfails: quotacardcardDvolume-quota
- Cloud Firestore Storagefails: quotaemail?card?Rrequest-rate
- cloud identity free (as a source of free identities / unlimited tenants) Storagefails: quotaemail?card?R
- Cloud Storage for Firebase (Spark/Blaze) Storagefails: quotacardcardRper-ip-shaping
- GCP research credits programme (Cloud for Higher Education) Storagefails: quotainstitutionno cardRwrite-gated
- Gmail (free @gmail.com) — 15 GB shared, and the honest ceiling of 'Gmail as infinite storage' Storagefails: quotaemailno cardRper-ip-shaping
- gmail as bulk storage Storagefails: quotaemail?card?R
- google artifact registry (as a blob store) Storagefails: quotaemail?card?Rdisagrees
- Google Cloud Free Tier — Cloud Storage Storagefails: quotacardcardRper-ip-shaping
- google cloud free trial ($300 / 90 days) Storagefails: quotaemail?card?R
- google cloud public datasets (as an intake) Storagefails: quotaemail?card?R
- Google Cloud Shell Storagefails: quotaemailno cardRretention-clamped
- Google Cloud Storage Always Free tier Storagefails: quotacardcardRper-ip-shaping
- Google Colab Storagefails: quotainstitutionno cardR
- google colab as a serving box Storagefails: quotaemail?card?R
- google colab free-tier runtime disk (as scratch storage) Storagefails: quotaemail?card?R
- google container registry (gcr.io) as a separate free tier Storagefails: quotaemail?card?R
- google docs / sheets as unlimited storage (uds, base64 embedding) Storagefails: quotaemail?card?R
- google groups as a file store Storagefails: quotaemail?card?R
- google photos Storagefails: quotaemail?card?R
- Google Photos Library API (append-only upload) Storagefails: quotainstitutionno cardR
- Google Play Books (personal upload library) Storagefails: quotainstitutionno cardRper-object-cap
- google sites (new sites) as free hosting/storage Storagefails: quotaemail?card?R
- Google Vertex AI Model Garden — custom model import Storagefails: quotacardcardR
- Firebase Studio Computedeademailno cardphoneD
- Google Programmable Search / Custom Search JSON API CrawlingdeadcardcardDrequest-rate
- LunaProxy, PIA S5 Proxy and 911 S5 (911.re) IP and proxiesdeadanonymousno cardD
Cloud Firestore Databases and KV
★ qualifiesemailno cardDrequest-rate
| Free tier | 1 GiB stored, 50,000 document reads per day, 20,000 writes per day, 20,000 deletes per day, exactly one free database per project unit GiB; metered on request-count |
|---|---|
| Price | none stated |
| Per-object cap | 1 GiB stored; a 1 MiB document size limit applies to the engine |
| Retention | data persists; there is no inactivity deletion on the free quota |
| Egress | 10 GiB a month of outbound data transfer is included in the free quota; beyond it the transfer is billed at the standard Cloud Firestore network rate, and reads are counted as operations against the daily quota free_allowance: 10 GiB per month |
| Identity | email a Google account |
| Limits |
a project may hold up to 100 databases, but only one is free, so multiple logical databases is a paid feature |
| Binding | request-rate |
| Interface | rest, graphql a REST and gRPC API with realtime listeners; SDKs for web, mobile and server; a GraphQL-ish query surface is not offered, the REST and gRPC APIs are |
| Tooling | the Firebase Admin SDK; the gcloud and firebase CLIs; REST |
| Composes with | 1 GiB of document state with client SDKs and realtime sync, which no other free tier here gives |
| Reliability | no reliability event recorded |
| Notes | The one-database rule is the constraint that a multi-tenant design meets first, not the 1 GiB. |
| Evidence rung | documented the quotas page was read today and states 'exactly one free database per project'. |
| Checked | 2026-10-23 (live); stale on 2026-11-22 |
| Parsed numbers |
|
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | firebase.google.com/docs/firestore/quotas | Free quota: 1 GiB stored data, 50,000 document reads/day, 20,000 writes/day, 20,000 deletes/day; 'Cloud Firestore allows exactly one free database per project' |
Unknown or withheld values (1)
$.reliability_notenone-observed
Google Workspace for Education Fundamentals (pooled tenant) Storage
★ qualifiesinstitutionno cardRper-object-cap
| Free tier | 100 TB of pooled cloud storage per institution, at 'No cost'. Pooled across Drive + Gmail + Photos and all shared drives. Confirmed twice on Google's own pages: 'Schools and universities with Google Workspace for Education subscriptions have a baseline of 100 TB of pooled storage shared across all users' and 'Google Workspace for Education Fundamentals / Standard — 100 TB total for all End Users'. unit TB; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | 5 TB per single upload (5000000000000 bytes) |
| Retention | Over-quota: at +25% or 14 days, uploads and Docs creation stop but download keeps working. (14 days) |
| Egress | Not metered for Drive reads. Google's own storage page says over-quota tenants 'can still sign in to ... view and download their files' — i.e. reads are never blocked by quota. Direct-download URLs are served from Google's edge; put behind Cloudflare if you want a cache. |
| Identity | institution derived from the record's prose, not from an identity field, and no signup was executed: 100 TB of pooled cloud storage per institution, at 'No cost'. Pooled across Drive + Gmail + Photos and all shared drives. Confirmed twice on Google's own pages: 'Scho |
| Limits |
per-object-cap 5 TB per single upload, and 750 GB per user per 24 h (upload+copy) — this is the real throttle, and it is refreshing not cumulative. request-rate: standard Drive API quota (per-project, raiseable). retention: none while the tenant is active; data survives account suspension but suspended users still consume the pool. Item cap 500 million items per user. Over-quota: at +25% or 14 days, uploads and Docs creation stop but download keeps working. |
| Binding | per-object-cap the per-object cap sets the shard size and therefore the object count; it bounds the total more tightly than any headline quota |
| Interface | rest, rclone rest |
| Tooling | rclone backend 'drive' (Tier 1, server-side copy, shared-drive support); Google Drive API v3; GAM (github.com/GAM-team/GAM) for admin-side bulk operations; gclone for service-account fan-out. |
| Composes with | One qualifying institution = one tenant = 100 TB. No user should be part of multiple tenants') and an institution with ineligible business 'must establish a separate domain for the school' — so the multiplier is *domains*, not users. Inside a tenant the multiplier is shared drives (unlimited count) and user count: up to 2,500 licenses per new domain, all drawing on the same 100 TB pool. Adding Teaching & Learning add-on licences adds 100 GB each 'with no cap' (paid). |
| Reliability | Stable for a decade, but Google already abolished the predecessor's unlimited tier (2021 archive: 'Existing customers ... will continue to have unlimited storage until early 2022'). Quota is now hard-capped and admin-auditable, so a tenant can be spotted by whoever administers it. |
| Notes | 100 TB of pooled cloud storage per institution, at 'No cost'. Pooled across Drive + Gmail + Photos and all shared drives. Confirmed twice on Google's own pages: 'Schools and universities with Google Workspace for Education subscriptions have a baseline of 100 TB of pooled storage shared across all users' and 'Google Workspace for Education Fundamentals / Standard — 100 TB total for all End Users'. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
| Parsed numbers |
|
Evidence (5)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | knowledge.workspace.google.com/admin/getting-started/edition | 'baseline of 100 TB of pooled storage shared across all users' |
| 2026-10-21 | support.google.com/a/answer/172541 | table: 'Google Workspace for Education Fundamentals / Standard | 100 TB total for all End Users'; 'Each user can upload and copy 750 GB to Drive within 24 hours ... up to 5 TB in size' |
| 2026-10-21 | knowledge.workspace.google.com/admin/getting-started/edition | K-12 and degree-granting institutions only; 'must establish a separate domain for the school to qualify' |
| 2026-10-21 | withheld | probe: GET knowledge.workspace.google.com/admin/getting-started/editions/overview-of-google-workspace-for-education-storage -> 200 (quota quoted live, 2026-10-21) |
| 2026-10-21 | withheld | probe: GET www.googleapis.com/drive/v3/about?fields=storageQuota -> 401 (anonymous; confirms API is the read path, auth required) |
Unknown or withheld values (5)
$.evidence[3].urlprivate-url$.evidence[4].urlprivate-url$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.retention.notesnot-extracted
Firebase Hosting Compute
caveatemailno cardDvolume-quota
| Free tier | On the no-cost Spark plan, Hosting storage is free up to 10 GB and CDN data transfer is free up to 10 GB a month, with a 2 GB ceiling on any single file. Over either limit without the Blaze plan, deployment stops or the sites are disabled until the month turns. unit GiB; metered on wall-clock; perpetual-allowance |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | 'Data transfer from the CDN to your end users is at no cost up to 10 GB/month'; past it there is a short grace period and then sites are disabled until the start of the next month, quoted |
| Identity | email not published |
| Limits |
10 GB a month of CDN transfer is the smaller of the two: at 10 GB stored and 10 GB served, the allowance is spent by serving the whole site once a month. |
| Binding | volume-quota 10 GB a month of CDN transfer is the smaller of the two: at 10 GB stored and 10 GB served, the allowance is spent by serving the whole site once a month. |
| Interface | unknown inferred from the tooling the provider documents; not probed |
| Tooling | not stated |
| Composes with | 10 GB stored and 10 GB a month served is a documentation site or a demo, not a data host; it composes as the front door for a project whose real artefacts sit in a bucket with a larger allowance. |
| Reliability | no reliability event recorded |
| Notes | Candidates: (a) Firebase Hosting has no free tier since the Spark plan was retired; refuted, the quota page still describes the no-cost level and the Blaze upgrade path. (b) 10 GB of transfer per day rather than per month; refuted, the page states per month and says sites stay disabled until the start of the next month. The dossier holds firebase-studio.json, which is a different product. |
| Evidence rung | documented |
| Checked | 2026-10-22 (live); stale on 2026-11-21 |
| Parsed numbers |
|
Evidence (3)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | firebase.google.com/docs/hosting/usage-quotas-pricing | 'Storage for your Hosting content is at no cost up to 10 GB'; on the non-Blaze plan reaching 10 GB blocks new deployments, quoted |
| 2026-10-22 | firebase.google.com/docs/hosting/usage-quotas-pricing | 'Data transfer from the CDN to your end users is at no cost up to 10 GB/month'; past it there is a short grace period and then sites are disabled until the start of the next month, quoted |
| 2026-10-22 | firebase.google.com/docs/hosting/usage-quotas-pricing | Hosting has a maximum size limit of 2 GB for individual files; paid overage is USD 0.026/GB stored and USD 0.15/GB transferred on Blaze, quoted |
More fields (1)
{
"wave2": {
"class": "perpetual-allowance",
"free_tier_kind": "perpetual-allowance",
"shape": {
"vcpu": null,
"ram_gib": null,
"gpu_class": null,
"gpu_count": null
}
}
}Unknown or withheld values (8)
$.identity_cost.detailprocess$.interfaceinterface-unknown$.per_object_cap.notesprocess$.per_object_cap.statednot-applicable$.reliability_notenone-observed$.retention.notesprocess$.retention.statednot-applicable$.toolingnot-stated
Google Compute Engine (Linux VMs) Compute
caveatcardcardDrequest-ratedisagrees
Sibling records disagree:
- identity_kind:
compute/gcp-computesays card, compute/gcp-windows says email
| Free tier | 1 non-preemptible e2-micro instance for the whole of each month in us-west1, us-central1 or us-east1, with 30 GB-months of standard persistent disk and 1 GB a month of outbound transfer from North America; the allowance is counted in hours, not instances unit hours; metered on wall-clock |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | not stated |
| Identity | card 'You won't be charged until you activate your full paid account' and the $300 welcome credit require a payment method to activate the paid account; the free tier itself is available to an eligible customer, quoted from cloud.google.com/free read 2026-10-22 |
| Limits |
not stated |
| Binding | request-rate |
| Interface | unknown |
| Tooling | not stated |
| Composes with | not stated |
| Reliability | no reliability event recorded |
| Evidence rung | documented the Google Cloud free-tier list and the free-features document were both read on 2026-10-22. The free tier is a monthly usage allowance that does not expire, which is a different product from the USD 300 credit the record carried; the credit is a 90-day trial that sits beside it |
| Checked | 2026-10-23 (live); stale on 2026-11-22 |
| Parsed numbers |
|
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-02 | cloud.google.com/products/compute/pricing/general-purpose | not published |
| 2026-10-22 | cloud.google.com/products/compute/pricing/general-purpose | '1 non-preemptible e2-micro VM instance per month in one of the following US regions: Oregon, Iowa, South Carolina. 30 GB-months standard persistent disk. 1 GB of outbound data transfer from North America', and 'Your Free Tier e2-micro instance limit is by time, not by instance ... free until you have used a number of hours equal to the total hours in the current month', and 'GPUs and TPUs are not included in the Free Tier offer', quoted |
More fields (16)
{
"talaria_m10h_usd": 20.1,
"talaria_m24h_usd": 48.25,
"nemo_m10h_usd": 20.93,
"keep_rate": 1.0,
"keep_rate_basis": null,
"billing_model": "unclassified",
"shape": {
"vcpu": 2,
"ram": 1,
"disk": 30,
"gpu_class": null,
"gpu_count": null,
"note": "e2-micro: 2 vCPU on a shared core, 1 GiB, 30 GB standard persistent disk"
},
"session_cap": {
"wall_clock_min": null,
"idle_reap_s": null,
"restart_required": false,
"note": "the allowance is the total hours in the month, so a single always-on e2-micro consumes all of it; the clock resets monthly, not on activity"
},
"persistent_storage": null,
"network_out": null,
"price_note": null,
"free_tier_kind": "allowance",
"granularity": {
"min_billed_s": 60,
"rounding_s": null
},
"burst_x": 1.0,
"ps": {
"gib": 30,
"price": "30 GB-months included, then standard PD rates",
"durability": "persistent disk",
"on_stop": true,
"on_destroy": false,
"retention": "while the disk exists"
},
"net": {
"out": true,
"in": true,
"ports": "any",
"static_ip": false,
"ipv6": false,
"allowance": "1 GB a month outbound from North America"
}
}Unknown or withheld values (15)
$.chainingnot-stated$.egress.meteredegress-unknown$.egress.termsnot-stated$.evidence[0].provedprocess$.extras.keep_rate_basisprocess$.extras.price_noteprocess$.interfaceinterface-unknown$.interface_detailwithheld$.limits_notenot-stated$.per_object_cap.notesprocess$.per_object_cap.statednot-stated$.reliability_notenone-observed$.retention.notesnot-extracted$.retention.statednot-stated$.toolingnot-stated
Corrections
- 2026-10-22 Google Cloud's free tier is a recurring usage allowance, and the record priced it as a credit
Google Compute Engine GPU VMs Compute
caveatemail?card?Dwrite-gated
| Free tier | not published unit none; metered on wall-clock |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | not stated |
| Identity | email (placeholder) not published |
| Limits |
excluded as gpu-only |
| Binding | write-gated |
| Interface | unknown |
| Tooling | not stated |
| Composes with | not stated |
| Reliability | no reliability event recorded |
| Evidence rung | documented both the accelerator-optimized rate card and the free-features document were read on 2026-10-22. The page proves the per-GPU rate and the free-features document proves there is no free GPU allowance |
| Checked | 2026-10-23 (live); stale on 2026-11-22 |
| Parsed numbers |
|
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-02 | cloud.google.com/products/compute/pricing/accelerator-optimi | not published |
| 2026-10-22 | cloud.google.com/products/compute/pricing/accelerator-optimi | 'GPUs and TPUs are not included in the Free Tier offer. You are always charged for GPUs and TPUs that you add to VM instances' (free-features document), against the accelerator-optimized rate card read the same day: a2-ultragpu-1g at $5.06879789 an hour on demand |
More fields (13)
{
"talaria_m10h_usd": null,
"talaria_m24h_usd": null,
"nemo_m10h_usd": null,
"keep_rate": null,
"keep_rate_basis": "no field in the source",
"billing_model": "unclassified",
"shape": {
"vcpu": null,
"ram": null,
"disk": null,
"gpu_class": "A100 80GB",
"gpu_count": 1,
"note": "each accelerator-optimized type fixes its own ratio; a2-ultragpu-1g is 1 GPU, 12 vCPU, 170 GB, 275 GiB SSD"
},
"session_cap": {
"wall_clock_min": null,
"idle_reap_s": null,
"restart_required": false,
"note": "no free allowance and no session cap; a GPU VM bills until it is stopped or deleted"
},
"persistent_storage": null,
"network_out": null,
"price_note": null,
"free_tier_kind": "none",
"net": null
}Unknown or withheld values (17)
$.chainingnot-stated$.egress.meteredegress-unknown$.egress.termsnot-stated$.evidence[0].provedprocess$.extras.price_noteprocess$.free_tier.noteswithheld$.free_tier.statedprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.per_object_cap.notesprocess$.per_object_cap.statednot-stated$.reliability_notenone-observed$.retention.notesnot-extracted$.retention.statednot-stated$.toolingnot-stated
Google Compute Engine (Windows Server) Compute
caveatemail?card?Drequest-ratedisagrees
Sibling records disagree:
- identity_kind:
compute/gcp-computesays card, compute/gcp-windows says email
| Free tier | not published unit mixed; metered on wall-clock |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | not stated |
| Identity | email (placeholder) not published |
| Limits |
not stated |
| Binding | request-rate |
| Interface | unknown |
| Tooling | not stated |
| Composes with | not stated |
| Reliability | no reliability event recorded |
| Evidence rung | documented the disks and images pricing page was read on 2026-10-22 and states the Windows Server premium-image hourly rate in two vCPU bands, plus the 90-day archive-snapshot minimum |
| Checked | 2026-10-23 (live); stale on 2026-11-22 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-02 | cloud.google.com/compute/disks-image-pricing#windows_server_ | not published |
| 2026-10-22 | cloud.google.com/compute/disks-image-pricing#windows_server_ | '$0.06 USD/hour for instances with 4 or fewer vCPUs' and '$0.13 USD/hour for instances with more than 4 vCPUs' on the Windows Server premium image, and 'Archive snapshots are billed for a minimum of 90 days of storage', quoted |
More fields (15)
{
"talaria_m10h_usd": 10.05,
"talaria_m24h_usd": 24.12,
"nemo_m10h_usd": null,
"keep_rate": 1.0,
"keep_rate_basis": null,
"billing_model": "unclassified",
"shape": null,
"session_cap": {
"wall_clock_min": null,
"idle_reap_s": null,
"restart_required": false,
"note": "the licence bills with the instance hour; the free e2-micro allowance does not carry a Windows image"
},
"persistent_storage": null,
"network_out": null,
"price_note": null,
"free_tier_kind": "allowance",
"granularity": {
"min_billed_s": 60,
"rounding_s": null
},
"ps": {
"gib": null,
"price": "archive snapshots metered at a 90-day floor",
"durability": "snapshot",
"on_stop": true,
"on_destroy": true,
"retention": "90 days of billing even if deleted sooner"
},
"net": null
}Unknown or withheld values (19)
$.chainingnot-stated$.egress.meteredegress-unknown$.egress.termsnot-stated$.evidence[0].provedprocess$.extras.keep_rate_basisprocess$.extras.price_noteprocess$.free_tier.noteswithheld$.free_tier.statedwithheld$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_notenot-stated$.per_object_cap.notesprocess$.per_object_cap.statednot-stated$.reliability_notenone-observed$.retention.notesnot-extracted$.retention.statednot-stated$.toolingnot-stated
GKE Agent Sandbox Compute
caveatemail?card?U
| Free tier | not published unit mixed; metered on wall-clock |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | not stated |
| Identity | email (placeholder) not published |
| Binding | |
| Interface | unknown |
| Tooling | not stated |
| Composes with | not stated |
| Reliability | no reliability event recorded |
| Notes | The GKE free tier is documented at cloud.google.com/free/docs/free-cloud-features: one Autopilot or zonal Standard cluster per month, and the credit applies to the cluster charge only. What is not documented is an agent-sandbox product: kubernetes-engine/docs/concepts/agent-sandbox returns 404 as of 2026-10-22. Do not read the cluster allowance as an agent sandbox until a page states one. |
| Evidence rung | unverified not verified on 2026-10-22: the page that named a GKE agent sandbox returns 404 on every route in the challenge protocol. The GKE free tier itself is documented elsewhere and says one Autopilot or zonal Standard cluster per month; the agent-sandbox product is not. |
| Checked | 2026-10-23 (inherited); stale on 2026-11-06 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-02 | cloud.google.com/kubernetes-engine/docs/concepts/agent-sandb | not published |
| 2026-10-22 | cloud.google.com/free/docs/free-cloud-features | unverified: the documented page returns 404 on 2026-10-22. Tried: broker read (404), ?__provider=firecrawl (404), raw fetch (404), archive closest month 2026-10. The GKE free-tier row is on https://cloud.google.com/free/docs/free-cloud-features and states one free Autopilot or zonal Standard cluster per month applying to the cluster charge only |
More fields (14)
{
"talaria_m10h_usd": 105.61,
"talaria_m24h_usd": 151.26,
"nemo_m10h_usd": 73.0,
"keep_rate": 1.0,
"keep_rate_basis": null,
"billing_model": "unclassified",
"shape": null,
"session_cap": null,
"persistent_storage": null,
"network_out": null,
"price_note": null,
"free_tier_kind": "one-time-credit",
"granularity": {
"min_billed_s": 0,
"rounding_s": null
},
"burst_x": 1.0
}Unknown or withheld values (19)
$.chainingnot-stated$.egress.meteredegress-unknown$.egress.termsnot-stated$.evidence[0].provedprocess$.extras.keep_rate_basisprocess$.extras.price_noteprocess$.free_tier.noteswithheld$.free_tier.statedwithheld$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_notenot-stated$.per_object_cap.notesprocess$.per_object_cap.statednot-stated$.reliability_notenone-observed$.retention.notesnot-extracted$.retention.statednot-stated$.toolingnot-stated
Google Agent Runtime / Agent Engine Compute
caveatemail?card?Drequest-rate
| Free tier | 180,000 vCPU-seconds and 360,000 GiB-seconds a month on Agent Engine, which is 50 vCPU-hours and 100 GiB-hours, recurring unit seconds; metered on wall-clock |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | not stated |
| Identity | email (placeholder) not published |
| Limits |
not stated |
| Binding | request-rate |
| Interface | unknown |
| Tooling | not stated |
| Composes with | not stated |
| Reliability | no reliability event recorded |
| Evidence rung | documented the free-features document and the Gemini Enterprise Agent Platform pricing page were both read on 2026-10-22. The allowance is stated in vCPU-seconds, which is the unit the expansion plan asks for and which a dollar figure hides |
| Checked | 2026-10-23 (live); stale on 2026-11-22 |
| Parsed numbers |
|
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-02 | cloud.google.com/products/gemini-enterprise-agent-platform/p | not published |
| 2026-10-22 | cloud.google.com/products/gemini-enterprise-agent-platform/p | 'Agent Engine: First 180,000 vCPU-seconds (50 hours) per month, First 360,000 GiB-seconds (100 hours) per month', quoted from the free-features document, against the per-node-hour rates on the pricing page read the same day |
More fields (15)
{
"talaria_m10h_usd": 61.8,
"talaria_m24h_usd": 148.32,
"nemo_m10h_usd": 61.8,
"keep_rate": 1.0,
"keep_rate_basis": "billed while running; can be stopped by the caller",
"billing_model": "unclassified",
"shape": null,
"session_cap": {
"wall_clock_min": null,
"idle_reap_s": null,
"restart_required": false,
"note": "the allowance is vCPU-seconds a month; there is no session clock, and a deployment that stays up exhausts the month early"
},
"persistent_storage": null,
"network_out": null,
"price_note": null,
"free_tier_kind": "allowance",
"granularity": {
"min_billed_s": 0,
"rounding_s": null
},
"burst_x": 1.0,
"net": null
}Unknown or withheld values (16)
$.chainingnot-stated$.egress.meteredegress-unknown$.egress.termsnot-stated$.evidence[0].provedprocess$.extras.price_noteprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_notenot-stated$.per_object_cap.notesprocess$.per_object_cap.statednot-stated$.reliability_notenone-observed$.retention.notesnot-extracted$.retention.statednot-stated$.toolingnot-stated
Corrections
- 2026-10-22 Google Cloud's free tier is a recurring usage allowance, and the record priced it as a credit
Google Cloud Build Compute
caveatemail?card?Drequest-rate
| Free tier | 2,500 build-minutes a month on e2-standard-2 in the default pool, plus the first 100 GB of worker SSD; both are promotional and the minute limit is stated as subject to change unit minutes; metered on wall-clock |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | not stated |
| Identity | email (placeholder) not published |
| Limits |
not stated |
| Binding | request-rate |
| Interface | unknown |
| Tooling | not stated |
| Composes with | not stated |
| Reliability | no reliability event recorded |
| Evidence rung | documented the Cloud Build pricing page was read on 2026-10-22. The allowance is a recurring monthly pool of build-minutes, and the page calls it promotional, which is a weaker commitment than a tier |
| Checked | 2026-10-23 (live); stale on 2026-11-22 |
| Parsed numbers |
|
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-02 | cloud.google.com/build/pricing | not published |
| 2026-10-22 | cloud.google.com/build/pricing | 'Each billing account comes with 2,500 free build-minutes per month' and 'This promotional free tier is for e2-standard-2 machine types in the default pool, and the limit is subject to change' and 'The first 100GB is free' of worker SSD, quoted |
More fields (15)
{
"talaria_m10h_usd": 108.0,
"talaria_m24h_usd": 259.2,
"nemo_m10h_usd": 93.0,
"keep_rate": 1.0,
"keep_rate_basis": null,
"billing_model": "unclassified",
"shape": {
"vcpu": 2,
"ram": 8,
"disk": 100,
"gpu_class": null,
"gpu_count": null,
"note": "e2-standard-2 with up to 100 GB of free worker SSD"
},
"session_cap": {
"wall_clock_min": null,
"idle_reap_s": null,
"restart_required": false,
"note": "a build runs to completion; the allowance is minutes a month, and partial minutes are billed at the actual number of seconds"
},
"persistent_storage": null,
"network_out": null,
"price_note": null,
"free_tier_kind": "allowance",
"granularity": {
"min_billed_s": 0,
"rounding_s": null
},
"burst_x": 1.0,
"net": null
}Unknown or withheld values (17)
$.chainingnot-stated$.egress.meteredegress-unknown$.egress.termsnot-stated$.evidence[0].provedprocess$.extras.keep_rate_basisprocess$.extras.price_noteprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_notenot-stated$.per_object_cap.notesprocess$.per_object_cap.statednot-stated$.reliability_notenone-observed$.retention.notesnot-extracted$.retention.statednot-stated$.toolingnot-stated
Corrections
- 2026-10-22 Google Cloud's free tier is a recurring usage allowance, and the record priced it as a credit
Google Cloud Free Tier (e2-micro) Compute
caveatcardcardDper-ip-shaping
| Free tier | 1 non-preemptible e2-micro VM instance a month in us-west1, us-central1 or us-east1; 30 GB-months of standard persistent disk; 1 GB of outbound data transfer from North America unit hours; metered on wall-clock; the limit is by time, not by instance: a month's worth of hours in total; the Free Tier is a separate product from the USD 300 / 90-day trial that sits on the same billing account |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | the Free Tier has no end date; Google reserves the right to change or eliminate limits with 30 days notice |
| Egress | 1 GB a month of outbound transfer from North America, excluding China and Australia; then standard rates free_allowance: 1 GB a month |
| Identity | card card or other payment method at signup; the published authorisation is between USD 0.00 and USD 1.00, a pending request rather than a charge, released within a few days to a month; debit accepted with a Visa or Mastercard logo, but a debit card that requires two-factor authentication is refused, as are prepaid and virtual cards; name, address and payment method are collected, and a tax ID only when the law requires it; no phone number is named on any signup page read, and no identity document |
| Limits |
e2-micro is a shared-core shape: 2 vCPU bursting at about 12.5% each, so 0.25 vCPU sustained, and a burst lasts 30 seconds |
| Binding | per-ip-shaping 1 GB of outbound a month is 0.09 per cent of one TiB: egress binds before the instance-hour allowance does |
| Interface | rest, cli Compute Engine API, gcloud CLI, SSH keys in instance or project metadata |
| Tooling | gcloud, the Compute Engine API, Terraform; a real Linux VM |
| Composes with | one always-on micro VM for a relay or a scheduler, with an egress allowance too small to serve from |
| Reliability | no reliability event recorded |
| Notes | SIGNUP PATH, read today. There is no Google Cloud signup page: cloud.google.com/signup returns a 404 and console.cloud.google.com/freetrial redirects to a Google account signup at accounts.cloud.google.com, whose first field is an email address. The Cloud form comes after the Google account exists and asks for name, address and payment method, and in some countries a bank account instead. A card is not optional: 'you must provide a credit card or other payment method'. THE HOLD IS 0.00 TO 1.00 USD, published on the free-features page, and it is the widest published range of the three providers: a bank may convert it and hold it for 'a few days to a month'. THE DEBIT RULE has a second half that matters more than the first: debit is accepted with a Visa or Mastercard logo, but 'Debit cards that require two-factor authentication aren't accepted for payment', which excludes most European and many Asian debit cards that push a one-time password by SMS. Prepaid and virtual are refused. THE GRANT IS TWO PRODUCTS. The always-free e2-micro is a permanent monthly allowance. The USD 300 is a 90-day trial on the same billing account, and when it closes Google stops 'all of its associated projects and resources' - the documentation does not carve out the always-free instance the way Oracle's FAQ does for Always Free. A user who wants the e2-micro to survive month four has to upgrade to a paid billing account, and then the always-free allowance still covers the instance and the bill is zero, but the account is now billable and the 1 GB egress overage is chargeable. That is the caveat, and it is the difference between this row and Oracle's. WHAT IT IS USABLE FOR: 1 GB a month of egress out of North America is 0.09 per cent of one TiB (1e9 bytes against 1.0995e12), so it is a relay, a scheduler, a monitoring node or an SSH jump box, and not a host. Inbound listen works and it has a public ephemeral IP, not a static one; a static IP is billed separately. The shape is 0.25 sustained vCPU with a 30-second burst, so anything CPU-bound is slow. What I could not establish: whether a phone number is required at signup, because no page read names one and the console form is behind a Google login I did not pass; the phrase 'Google accounts with only a phone number are not supported' is about the account type, not about a verification step. |
| Evidence rung | documented The grant, the egress term, the shape, the card rules and the hold each come from a first-party page read and quoted on 2026-10-24. The signup field list is documented at the FAQ level; the console form itself is behind a Google login that I did not pass, so the phone question is left unverified rather than guessed. |
| Checked | 2026-10-25 (live); stale on 2026-11-24 |
| Parsed numbers |
|
Evidence (27)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | cloud.google.com/free/docs/free-cloud-features | 1 non-preemptible e2-micro a month in us-west1, us-central1 or us-east1, 30 GB-months standard PD, 1 GB outbound |
| 2026-10-22 | cloud.google.com/free/docs/free-cloud-features | the e2-micro limit is by time, not by instance: eligible use is free until the hours equal the hours in the month |
| 2026-10-22 | cloud.google.com/free/docs/free-cloud-features | the Free Tier has no end date, but Google may change or eliminate limits with 30 days notice |
| 2026-10-24 | cloud.google.com/free/docs/free-cloud-features | quoted: the authorisation hold is USD 0.00 to 1.00, and the always-free e2-micro is 1 GB of outbound data transfer per month from North America |
| 2026-10-24 | cloud.google.com/signup | quoted: 'During the sign up, you must provide a credit card or other payment method' |
| 2026-10-24 | cloud.google.com/free/docs/free-cloud-features | quoted: '1 non-preemptible e2-micro VM instance per month' in us-west1, us-central1 or us-east1 with '30 GB-months standard persistent disk' and '1 GB of outbound data transfer', and 'A Google Cloud billing account is required to access the Google Cloud Free Tier' |
| 2026-10-24 | docs.cloud.google.com/billing/docs/how-to/payment-methods | quoted: 'Debit cards with the Visa or MasterCard logo' are accepted; 'Debit cards that require two-factor authentication aren't accepted for payment.'; 'Prepaid cards aren't accepted for payment.' |
| 2026-10-24 | docs.cloud.google.com/billing/docs/how-to/manage-billing-acc | quoted, the hold: 'When you create a new, self-serve Cloud Billing account, you might notice a $0.00 transaction from Google on your bank or credit card statement. These are authorization requests only, not actual charges.' |
| 2026-10-24 | cloud.google.com/signup-faqs | quoted: 'During signup, we ask for your name, address, and payment method to verify your identity.' and 'Authorizations may stay in your account for 1-14 business days depending on your bank.' |
| 2026-10-24 | console.cloud.google.com/freetrial | the signup URL, opened today, redirects to https://accounts.cloud.google.com/signup/email?continueUrl=https://console.cloud.google.com/freetrial and the only readable field is 'Email address' with 'Your work, personal, or any other email address you own' - so the account is created at a Google account signup, and cloud.google.com/signup returns a 404 |
| 2026-10-24 | cloud.google.com/signup-faqs | quoted, the form: 'During signup, we ask for your name, address, and payment method to verify your identity.' and 'Authorizations may stay in your account for 1-14 business days depending on your bank.' |
| 2026-10-24 | docs.cloud.google.com/free/docs/free-cloud-features | quoted, the card gate: 'During the sign up, you must provide a credit card or other payment method that is valid for the period of the Free Trial. Depending on your country, you might also need to verify your bank account.' |
| 2026-10-24 | docs.cloud.google.com/free/docs/free-cloud-features | quoted, the hold with its amount: 'Temporary hold: The hold often appears as a pending transaction from Google, with a value between $0.00 and $1.00 USD. Your bank might convert the amount to a local currency.' and 'The hold is typically released within a few days to a month.' |
| 2026-10-24 | docs.cloud.google.com/billing/docs/how-to/payment-methods | quoted, the accepted cards: 'Debit cards with the Visa or MasterCard logo' are accepted; 'Debit cards that require two-factor authentication aren't accepted for payment.'; 'Prepaid cards aren't accepted for payment.' |
| 2026-10-24 | docs.cloud.google.com/free/docs/free-cloud-features | quoted, the grant: '1 non-preemptible e2-micro VM instance per month in one of the following US regions: Oregon: us-west1. Iowa: us-central1. South Carolina: us-east1. 30 GB-months standard persistent disk. 1 GB of outbound data transfer from North America to all region destinations (excluding China and Australia) per month.' |
| 2026-10-24 | docs.cloud.google.com/free/docs/free-cloud-features | quoted, the time-denominated limit: 'Your Free Tier e2-micro instance limit is by time, not by instance. Each month, eligible use of all of your e2-micro instances is free until you have used a number of hours equal to the total hours in the current month.' |
| 2026-10-24 | docs.cloud.google.com/free/docs/free-cloud-features | quoted, the two products separated: 'The Free Trial is a 90-day program' with 'a Free Trial billing account that is preloaded with $300 in free Welcome credit which is valid for 90 days', and 'The Free Tier has no end date, but Google reserves the right to change the offering, including changing or eliminating usage limits, with 30 days' advance notice.' |
| 2026-10-24 | docs.cloud.google.com/free/docs/free-cloud-features | quoted, what the trial end does: 'If you don't upgrade to a Paid billing account before 90 days pass or if you spend the $300 in free credit, then your Free Trial billing account will be closed and all of its associated projects and resources will be stopped.' and 'Resources are stopped: All resources you created during the trial are stopped.' |
| 2026-10-24 | docs.cloud.google.com/free/docs/free-cloud-features | quoted, the billing account requirement for the free product: 'A Google Cloud billing account is required to access the Google Cloud Free Tier.' |
| 2026-10-24 | docs.cloud.google.com/billing/docs/how-to/close-or-reopen-bi | quoted, that a non-billable billing account stops even free services: 'if your projects are linked to a Cloud Billing account that is not in a billable state (such as an account suspension or billing account closure), you can't use the products and services enabled in your projects. This is true even if your projects only use Google Cloud or other billing-eligible services that are free.' |
| 2026-10-24 | cloud.google.com/terms/free-trial | quoted, the trial clock and its end: 'The Free Trial starts when Customer creates a billing account in the Google Cloud console ... and ends on the earlier of: (i) the date that Customer's fees for usage exceed the $300 credit provided under the Free Trial, (ii) 3 months from the Free Trial Start Date (on this date any unused credit will expire), or (iii) the date that Customer upgrades to a paid Google Cloud Platform account.' |
| 2026-10-24 | cloud.google.com/terms/free-trial | quoted, the core limit inside the trial: 'Customer's use of Google Compute Engine is restricted to only eight cores of concurrent usage per Free Trial' and 'Google may terminate or modify the Free Trial at any time with or without notice to Customer.' |
| 2026-10-24 | docs.cloud.google.com/compute/docs/general-purpose-machines | quoted, the shape: 'e2-micro sustains 2 vCPUs, each for 12.5% of CPU time totaling 25% CPU time.' and the table row 'e2-micro | 2 | 0.25 | 1' for vCPUs, fractional vCPUs and memory in GB; and 'e2-micro: 30 seconds' of burst at 100% |
| 2026-10-24 | docs.cloud.google.com/support/docs/troubleshoot-signup | quoted, the signup failure modes: 'Billing address: Verify that the billing address, name, and contact details that you entered during the signup match the exact information on file with your payment card issuer or bank.' and 'Disable VPNs and proxies' and 'Make sure that your card supports international transactions and automatic recurring payments.' |
| 2026-10-24 | cloud.google.com/signup-faqs | quoted, the account prerequisite: 'Your Google account must have an email associated with it. Google accounts with only a phone number are not supported.' and the named error codes 'OR-CCSEH-26, OR-CCSEH-11, OR-CCSEH-25, OR-BSBBF-55, or OR-CCSEH-05' |
| 2026-10-24 | cloud.google.com/signup-faqs | quoted, per-person eligibility: 'Each user is eligible for the $300 free credit ... if you have never been a paying customer of Google Cloud, Google Maps Platform, or Firebase and you haven't previously signed up for the Google Cloud Free Trial.' |
| 2026-10-24 | docs.cloud.google.com/billing/docs/how-to/verify-bank | quoted, the alternative payment route and its gate: 'To use a bank account as a primary payment method for self-serve (online) billing accounts, your account must have an established billing history with Google Cloud, including a minimum account age and a record of on-time payments.' - so a new account cannot start with a bank account |
More fields (14)
{
"shape": {
"vcpu": 2,
"ram_gib": 1,
"disk_gib": 30,
"gpu_class": null,
"gpu_count": null
},
"billing_model": "allocated",
"granularity": {
"min_billed_s": 60,
"rounding_s": 60
},
"session_cap": {
"wall_clock_min": null,
"idle_reap_s": null,
"restart_required": false,
"note": "no idle reaping; suspend and resume are user-initiated only; the 90-day trial clock is the clock that ends the account"
},
"keep_rate": {
"value": 1.0,
"basis": "an instance bills its running hours against the monthly allowance",
"source": "https://docs.cloud.google.com/free/docs/free-cloud-features"
},
"persistent_storage": {
"included_gib": 30,
"price_gib_month": "included to 30 GB-months",
"durability": "persistent disk",
"survives_stop": true,
"survives_destroy": true,
"max_retention": "no end date while a billing account is active"
},
"network_out": {
"allow_outbound": true,
"allow_inbound_listen": true,
"ports": "a public ephemeral external IP; all ports reachable subject to firewall rules",
"static_ip": false,
"ipv6": null,
"per_session_bytes": "1 GB a month"
},
"free_tier_kind": "allowance",
"m10h_usd": 0.0,
"m24h_usd": 0.0,
"price_arithmetic": "one month of hours at 0.25 sustained vCPU; the whole allowance is one instance",
"census_rows": [
"gcp-e2-micro",
"google-cloud-free-e2-micro"
],
"payment_methods": {
"card_at_signup": "yes",
"debit": "accepted-with-condition",
"prepaid": "refused",
"virtual": "refused",
"hold_amount": "USD 0.00 to USD 1.00",
"hold_detail": "an authorisation request, not a charge; released in a few days to a month, or 1 to 14 business days per the FAQ",
"identity": "name, address and payment method; no identity document",
"source_url": "https://docs.cloud.google.com/free/docs/free-cloud-features",
"read_on": "2026-10-24"
},
"signup": {
"url": "https://console.cloud.google.com/freetrial",
"redirects_to": "https://accounts.cloud.google.com/signup/email?continueUrl=https://console.cloud.google.com/freetrial",
"form_fields": [
"Google account email (created first, at accounts.cloud.google.com)",
"name",
"address",
"payment method: card number, expiry, billing address",
"tax ID, when the law requires it",
"bank account instead of a card, in some countries, only with billing history"
],
"company_required": false,
"id_document_required": false,
"phone_required": "not named on any signup page read; unverified",
"source_url": "https://cloud.google.com/signup-faqs",
"read_on": "2026-10-24"
}
}Unknown or withheld values (3)
$.per_object_cap.noteswithheld$.per_object_cap.statednot-applicable$.reliability_notenone-observed
Google Cloud Run Compute
caveatcardcardDvolume-quota
| Free tier | per billing account, monthly: 2 million requests, 180,000 vCPU-seconds and 360,000 GiB-seconds under request-based billing; 240,000 vCPU-seconds and 450,000 GiB-seconds under instance-based billing; plus 1 GiB of outbound transfer from North America unit seconds; metered on wall-clock |
|---|---|
| Price | none stated |
| Per-object cap | a request may run up to 60 minutes; memory and CPU per instance are configurable |
| Retention | no persistent disk; a container is stateless unless it mounts a volume or a bucket |
| Egress | 1 GiB of data transfer within North America free per month, then Google Cloud network pricing at the Premium tier free_allowance: 1 GiB |
| Identity | card a card at signup; the free tier is a spending-based discount applied after the fact, not a prepaid allowance |
| Limits |
it was not resolved either |
| Binding | volume-quota an always-on instance spends the monthly grant in about ten days: 450,000 GiB-seconds at 512 MiB is 900,000 seconds, 10.4 days. |
| Interface | unknown an OCI image behind an HTTPS endpoint; the service listens on a port |
| Tooling | gcloud, the Cloud Run API, Docker; any language with an HTTP server |
| Composes with | the only free serverless tier here with 50 vCPU-hours and a listening socket |
| Reliability | a service scales to zero by default and pays a cold start on the next request; a minimum instance removes the cold start and starts the idle meter |
| Notes | Arithmetic from the published free tiers, not a measurement: at 0.25 vCPU and 512 MiB, an always-on instance needs 648,000 vCPU-seconds and 1,296,000 GiB-seconds a month. The memory grant runs out first, at 900,000 seconds, so the free tier holds an always-on 0.25 vCPU service for about 10.4 days a month. What is not verified is whether the 'Idle time (Min instance)' SKU is drawn from the same free tier; the page presents it as a separate rate. |
| Evidence rung | documented both free-tier tables and the per-second rates were read on the pricing page today |
| Checked | 2026-10-25 (live); stale on 2026-11-24 |
| Parsed numbers |
|
Evidence (5)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | cloud.google.com/free/docs/free-cloud-features | Cloud Run free tier: 2 million requests, 360,000 GB-seconds of memory, 180,000 vCPU-seconds, 1 GB of outbound transfer a month |
| 2026-10-24 | cloud.google.com/run/pricing | quoted: 'Free tier (based on us-central1 pricing): CPU - First 240,000 vCPU-seconds free per month; RAM - First 450,000 GiB-seconds free per month' under instance-based billing |
| 2026-10-24 | cloud.google.com/run/pricing | quoted: 'CPU - First 180,000 vCPU-seconds free per month; RAM - First 360,000 GiB-seconds free per month; Requests - 2 million requests free per month' under request-based billing |
| 2026-10-24 | cloud.google.com/run/pricing | quoted: 'The free tier usage is aggregated across projects by billing account and resets every month' |
| 2026-10-24 | cloud.google.com/run/pricing | quoted rate table: 'CPU (per vCPU-second) ... $0.000018', 'Memory (per GiB-second) ... $0.000002', 'Idle time (Min instance1) $0.0000025' |
More fields (20)
{
"talaria_m10h_usd": 3.7,
"talaria_m24h_usd": 16.19,
"nemo_m10h_usd": 4.36,
"keep_rate": {
"value": 0.0,
"basis": "a scaled-to-zero service does not bill",
"source": "https://cloud.google.com/free/docs/free-cloud-features"
},
"keep_rate_basis": null,
"billing_model": "per-100ms",
"shape": {
"vcpu": 0.25,
"ram_gib": 0.5,
"disk_gib": null,
"gpu_class": null,
"gpu_count": null
},
"session_cap": {
"wall_clock_min": 60,
"idle_reap_s": null,
"restart_required": false,
"note": "a request may run 60 minutes; the service scales to zero when idle unless a minimum instance is set"
},
"persistent_storage": {
"included_gib": 0,
"price_gib_month": "Cloud Storage is separate",
"durability": "none in the runtime",
"survives_stop": false,
"survives_destroy": false,
"max_retention": "none"
},
"network_out": {
"allow_outbound": true,
"allow_inbound_listen": true,
"ports": "the container port on a generated HTTPS hostname",
"static_ip": false,
"ipv6": false,
"per_session_bytes": null
},
"price_note": "the two seed columns differ by 15 per cent because each applied a grant and neither established which meter it attaches to, the Services meter or the Instances meter. This one is genuinely unresolved: no number here is wrong and none is established, so the record carries the gap rather than choosing",
"granularity": {
"min_billed_s": 0.1,
"rounding_s": 0.1
},
"free_tier_kind": "monthly-spending-discount",
"open_question": "whether the grant applies to the Services meter or the Instances meter",
"gap": "the grant's meter (Services or Instances) is not resolved; the claim page's Instances section and footnotes would close it",
"burst_x": 1.0,
"term": "monthly",
"m10h_usd": 0.0,
"m24h_usd": 0.0,
"price_arithmetic": "free within the monthly grant; 512 MiB at 0.25 vCPU consumes the 450,000 GiB-second grant in 900,000 s = 10.4 days of 24/7"
}Unknown or withheld values (2)
$.extras.keep_rate_basisprocess$.interfaceinterface-unknown
Google Cloud Shell Compute
caveatemailno cardDvolume-quota
| Free tier | 50 hours a week of Cloud Shell, a 12-hour session cap, non-interactive sessions ended after 40 minutes, and 5 GB of persistent disk mounted as $HOME unit hours; metered on wall-clock; The 50 hours is weekly, not monthly, which makes it the largest published hour budget. |
|---|---|
| Price | none stated |
| Per-object cap | not applicable: the unit is a session, not an object |
| Retention | 5 GB of $HOME persists between sessions and counts towards the 5 GB limit; anything outside $HOME is lost when the instance terminates |
| Egress | no separate egress allowance published for Cloud Shell itself |
| Identity | email a Google account; no card, and no Cloud project is needed to open Cloud Shell itself |
| Limits |
the 12-hour cap and the 40-minute non-interactive timeout are two different clocks and both bind |
| Binding | volume-quota the 40-minute non-interactive cut is what kills a daemon; the 50-hour week is generous but unrunnable unattended |
| Interface | ssh the browser terminal, or the gcloud CLI over a tunnel; no inbound port |
| Tooling | a full shell with gcloud, git, docker and language runtimes |
| Composes with | a 5 GB home directory that outlives the VM, and nothing else |
| Reliability | the backing VM is not dedicated and terminates after 40 minutes of inactivity, so a background process does not survive the tab |
| Notes | RAM and CPU are not on the quotas page read today, so the shape is unverified. The 40-minute non-interactive rule is the decisive sentence for a daemon, and it is quoted verbatim. |
| Evidence rung | documented quota, session caps and disk are quoted from Google's quotas-and-limits page today. The page I first tried, /shell/docs/quotas, is a 404; the live path is /shell/docs/quotas-limits. |
| Checked | 2026-10-25 (live); stale on 2026-11-24 |
| Parsed numbers |
|
Evidence (11)
| date | source | what it proved |
|---|---|---|
| 2026-10-06 | cloud.google.com/shell/docs/limitations | 'Non-interactive sessions are ended automatically after 40 minutes. Cloud Shell sessions are capped at 12 hours' |
| 2026-10-22 | cloud.google.com/shell/docs/limitations | read live: 12-hour session cap and 40-minute non-interactive timeout |
| 2026-10-24 | cloud.google.com/shell/docs/quotas-limits | quoted: 'The default weekly Cloud Shell quota is 50 hours.' |
| 2026-10-24 | cloud.google.com/shell/docs/quotas-limits | quoted: 'Cloud Shell provisions 5 GB of free persistent disk storage mounted as your $HOME directory' |
| 2026-10-24 | cloud.google.com/shell/docs/quotas-limits | quoted: 'terminates if your session is inactive for 40 minutes. After the instance is terminated, any modifications that you made outside your $HOME directory are lost.' |
| 2026-10-24 | cloud.google.com/shell/docs/quotas-limits | quoted: 'Cloud Shell sessions are capped at 12 hours, after which sessions automatically terminate.' |
| 2026-10-24 | cloud.google.com/shell/docs/quotas-limits | quoted: 'To prevent your $HOME directory from being deleted due to inactivity for 120 days, start a session.' |
| 2026-10-24 | cloud.google.com/shell/docs/quotas-limits | quoted: 'Cloud Shell is intended for interactive use only. Non-interactive sessions are ended automatically after 40 minutes. Cloud Shell sessions are capped at 12 hours, after which sessions automatically terminate.' |
| 2026-10-24 | cloud.google.com/shell/docs/quotas-limits | quoted, with a link and code span in the original: 'Cloud Shell provisions 5 GB of free [persistent disk storage](https://docs.cloud.google.com/compute/docs/disks/persistent-disks) mounted as your $HOME directory on the virtual machine instance.' |
| 2026-10-24 | cloud.google.com/shell/docs/quotas-limits | quoted, with the code span preserved: 'terminates if your session is inactive for 40 minutes. After the instance is terminated, any modifications that you made outside your $HOME directory are lost.' |
| 2026-10-24 | cloud.google.com/shell/docs/quotas-limits | quoted: 'If you reach your quota, you'll need to wait until the date and time specified before you can use Cloud Shell again' |
More fields (16)
{
"shape": {
"vcpu": null,
"ram_gib": null,
"disk_gib": 5,
"gpu_class": null,
"gpu_count": null,
"note": "the quotas page does not state vCPU or RAM; the ephemeral VM shape is unverified"
},
"billing_model": "free",
"session_cap": {
"wall_clock_min": 720,
"idle_reap_s": 2400,
"restart_required": true,
"note": "12 h maximum; 40 min non-interactive cut; 50 h/week"
},
"keep_rate": {
"value": 0.0,
"basis": "the VM is ephemeral and free",
"source": "https://cloud.google.com/shell/docs/limitations"
},
"persistent_storage": {
"included_gib": 5,
"price_gib_month": "free",
"durability": "$HOME only",
"survives_stop": true,
"survives_destroy": false,
"max_retention": "while the account exists"
},
"network_out": {
"allow_outbound": true,
"allow_inbound_listen": false,
"ports": "outbound; inbound only through the web preview proxy",
"static_ip": false,
"ipv6": false,
"per_session_bytes": null
},
"free_tier_kind": "free-forever-with-quota",
"census_rows": [
"google-cloud-shell"
],
"shape_note": "Google does not publish the vCPU or RAM of the Cloud Shell VM on the quotas page; only the 5 GB $HOME is stated. Both are null rather than guessed.",
"term": "none",
"m10h_usd": 0.0,
"m24h_usd": 0.0,
"price_arithmetic": "USD 0.00 for the weekly allowance",
"unattended": false,
"unattended_mechanism": "none: 'Non-interactive sessions are ended automatically after 40 minutes'",
"units": "RAM and disk are quoted in the vendor's own decimal GB, not GiB. The ram_gib and disk_gib keys carry the vendor's number unconverted, so where a vendor writes 30 GB the key holds 30 and the true value is about 27.9 GiB. No conversion was performed."
}Google Cloud Spot Compute
caveatcardcardD
| Free tier | Google publishes two numbers that answer the survival question directly and neither is an interruption rate. Estimated uptime is 'the minimum amount of time that you can expect most of your Spot VMs to run before Compute Engine preempts them', and it is bucketed into exactly three values: '60 minutes (3,600 seconds)', '10 minutes (600 seconds)' and '1 minute (60 seconds)'. Obtainability is a separate score from 0.0 to 1.0 for whether the creation request succeeds at all. The preemption rate from advice.capacityHistory is defined as 'Total preempted Spot VMs in a day' divided by 'Total Spot VMs that stopped running in a day'. unit hours; metered on wall-clock; none |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | not stated |
| Identity | card card: 'the Free Trial is USD 300 over 90 days and requires a credit card or other payment method at signup'. No authorisation or hold amount is stated in the evidence, so the hold is unknown, and whether debit or prepaid is accepted is not stated: unknown. Spot VMs sit outside the Free Tier - 'The Google Cloud Free Tier credits for Compute Engine don't apply to Spot VMs' - and a non-billable Free trial account cannot add GPUs or request a quota increase. |
| Binding | none Estimated uptime is bucketed to three values and is a floor for the majority of VMs, so it cannot be converted into a per-hour probability without assuming a distribution. |
| Interface | unknown inferred from the tooling the provider documents; not probed |
| Tooling | not stated |
| Composes with | Estimated uptime is the one vendor-published figure a scheduler can gate on before launching: if it returns 10 minutes and the checkpoint interval is 10 minutes, is not worth the setup cost and the scheduler should move zone. |
| Reliability | no reliability event recorded |
| Notes | The comparability trap is the denominator. A GCP preemption rate of 0.50 does not mean a VM has a 50 per cent chance of preemption in an hour; it means half of the VMs that stopped that day were stopped by preemption rather than by their owner. Published third-party tables that put 'GCP ~2.3 per cent per hour' next to an Azure per-hour eviction probability are mixing two statistics. Also of note: preemption history is unavailable for N1 machine types with GPUs attached, which is exactly the cheap GPU lane, and the whole feature is marked Preview. fails-identity via a05-spot. The one recorded is the better-evidenced. |
| Evidence rung | documented |
| Checked | 2026-10-22 (live); stale on 2026-11-21 |
Evidence (10)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | cloud.google.com/compute/docs/instances/view-vm-availability | estimated uptime is 'the minimum amount of time that you can expect most of your Spot VMs to run' and takes only the values 60 minutes, 10 minutes or 1 minute, quoted |
| 2026-10-22 | cloud.google.com/compute/docs/instances/view-spot-preemption | the preemption rate formula is total preempted Spot VMs in a day divided by total Spot VMs that stopped running in a day, and the rate is expressed as a fraction of VMs rather than of time, quoted |
| 2026-10-22 | cloud.google.com/compute/docs/instances/view-spot-preemption | preemption history covers 30 days per machine type and zone, and the feature cannot report N1 machine types with GPUs attached, custom machine types or TPUs |
| 2026-10-22 | cloud.google.com/compute/docs/instances/spot | the preemption notice duration can be set to 120 seconds or 0 seconds, 120 is recommended for workloads needing more than 30 seconds, and the shutdown period is best effort up to 30 seconds, quoted |
| 2026-10-22 | cloud.google.com/compute/docs/instances/create-use-spot | the gcloud flag is --preemption-notice-duration and the value must be either 120s or 0s, quoted |
| 2026-10-22 | cloud.google.com/compute/docs/instances/spot | 'The Google Cloud Free Tier credits for Compute Engine don't apply to Spot VMs', quoted in the Spot VMs limitations list |
| 2026-10-22 | cloud.google.com/free/docs/free-cloud-features | on a non-billable Free trial account you cannot 'Add GPUs to your VM instances' and cannot 'Request a quota increase', quoted |
| 2026-10-22 | cloud.google.com/free/docs/free-cloud-features | the Free Trial is USD 300 over 90 days and requires a credit card or other payment method at signup, quoted |
| 2026-10-22 | cloud.google.com/spot-vms/pricing | the accelerator table lists 'Current Spot pricing (USD)' per machine type, including a2-highgpu-1g at $2.20401, a3-highgpu-1g at $6.273932575 and g2-standard-4 at $0.423016, all for 1 hour |
| 2026-10-22 | cloud.google.com/compute/docs/instances/spot | Spot VMs offer 'up to 91% discounts for many machine types, GPUs, TPUs, and Local SSDs', and 'The Google Cloud Free Tier credits for Compute Engine don't apply to Spot VMs', quoted |
More fields (1)
{
"wave2": {
"class": "spot-arbitrage",
"free_tier_kind": "none",
"shape": null
}
}Unknown or withheld values (9)
$.egress.meteredegress-unknown$.egress.termsnot-stated$.interfaceinterface-unknown$.per_object_cap.notesprocess$.per_object_cap.statednot-applicable$.reliability_notenone-observed$.retention.notesprocess$.retention.statednot-applicable$.toolingnot-stated
Google Cloud Text-to-Speech Compute
caveatcardcardDvolume-quota
| Free tier | A recurring monthly free allowance per voice class, not a one-time credit: 'WaveNet voices 0 to 4 million characters', 'Standard voices 0 to 4 million characters', 'Chirp 3: HD voices 0 to 1 million characters', 'Studio voices 0 to 1 million characters', 'Neural2 voices 0 to 1 million characters'. Over the allowance the price is 'US$4 per 1 million characters' for WaveNet and Standard, 'US$30 per 1 million' for Chirp 3 HD and Studio at 'US$160 per 1 million'. At the common 1,000 characters a minute, 4 million characters is about 66 hours of speech a month. unit hours; metered on wall-clock; recurring-monthly-allowance |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | not stated |
| Identity | card not published |
| Limits |
the allowance resets monthly but the account must be a billing account: 'You must enable billing to use Text-to-Speech, and will be automatically charged if your usage exceeds the number of free characters allowed per month', so a card is required before the first character is synthesised and an overrun bills silently. |
| Binding | volume-quota the allowance resets monthly but the account must be a billing account: 'You must enable billing to use Text-to-Speech, and will be automatically charged if your usage exceeds the number of free characters allowed per month', so a card is required before the first character is synthesised and an overrun bills silently. |
| Interface | unknown inferred from the tooling the provider documents; not probed |
| Tooling | not stated |
| Composes with | What breaks: the Gemini-family TTS models have 'Not available' in the free column, so the newest voices are excluded, and an accidental loop over a long document bills at USD 4 per million characters. |
| Reliability | no reliability event recorded |
| Notes | These are inference credits, not general compute. This is the largest recurring voice allowance found in the wave and it sits with a hyperscaler rather than a startup, which is why the free tiers of Google and Amazon are worth more than the grants of the voice startups. The 66-hour figure is our division at 1,000 characters a minute; Google publishes characters, not hours. Not probed: no project was created; the card requirement is documented on the pricing page. |
| Evidence rung | documented |
| Checked | 2026-10-22 (live); stale on 2026-11-21 |
Evidence (4)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | cloud.google.com/text-to-speech/pricing | the free usage table prints 'WaveNet voices 0 to 4 million characters', 'Standard voices 0 to 4 million characters', 'Chirp 3: HD voices 0 to 1 million characters', 'Studio voices 0 to 1 million characters', 'Neural2 voices 0 to 1 million characters', quoted |
| 2026-10-22 | cloud.google.com/text-to-speech/pricing | 'You must enable billing to use Text-to-Speech, and will be automatically charged if your usage exceeds the number of free characters allowed per month', quoted |
| 2026-10-22 | cloud.google.com/text-to-speech/pricing | 'Chirp 3: HD voices ... US$0.00003 per character (US$30 per 1 million characters)' and 'WaveNet voices ... US$0.000004 per character (US$4 per 1 million characters)', quoted |
| 2026-10-22 | cloud.google.com/text-to-speech/pricing | the Gemini TTS rows print 'Free usage limit: Not available', so the newest models are outside the allowance, quoted |
More fields (1)
{
"wave2": {
"class": "perpetual-free-allowance",
"free_tier_kind": "recurring-monthly-allowance",
"shape": {
"vcpu": null,
"ram_gib": null,
"gpu_class": null,
"gpu_count": 0
}
}
}Unknown or withheld values (10)
$.egress.meteredegress-unknown$.egress.termsnot-stated$.identity_cost.detailprocess$.interfaceinterface-unknown$.per_object_cap.notesprocess$.per_object_cap.statednot-applicable$.reliability_notenone-observed$.retention.notesprocess$.retention.statednot-applicable$.toolingnot-stated
Google Colab Compute
caveatemailno cardDvolume-quotadisagrees
Sibling records disagree:
- verdict:
compute/google-colabsays qualifies-with-caveat, compute/google-colab-paid says fails-quota - identity_kind:
compute/google-colabsays email, compute/google-colab-paid says card
| Free tier | free of charge access to computing resources including GPUs and TPUs; a notebook runs for at most 12 hours; usage limits, idle timeout, maximum VM lifetime and GPU types are not published and vary over time unit hours; metered on unstated; There is no published GPU-hour number. Anything quoted for Colab's free GPU quota is somebody's recollection. |
|---|---|
| Price | none stated |
| Per-object cap | not applicable: the unit is a session, not an object |
| Retention | the VM disk is deleted when the runtime is recycled; only Google Drive survives |
| Egress | no published egress allowance or price for the notebook runtime |
| Identity | email a Google account; no card for the free tier |
| Limits |
a documented absence: 'Colab does not publish these limits' is a discretionary throttle, not a quota |
| Binding | volume-quota the quota is real but unquantified, so it cannot be planned against; the 12-hour wall is the only number the vendor will commit to |
| Interface | rest, cli a notebook in the browser; no SSH and no inbound port |
| Tooling | Jupyter; the Colab client; no shell beyond the notebook's |
| Composes with | a notebook is not a machine: nothing survives the runtime, so it cannot hold state or serve |
| Reliability | Colab states outright that it prioritises interactive users and that runtimes time out when idle, so a long job is cut for reasons the client cannot see |
| Notes | RAM, GPU class and the idle timeout are all unpublished, so no record can carry them. What would settle them is running a probe notebook that reports /proc/meminfo and nvidia-smi, which needs an account. |
| Evidence rung | documented the 12-hour cap and the refusal to publish the rest are quoted from the vendor FAQ. RAM and GPU class are unverified because the page states a 'standard memory profile' without a number. |
| Checked | 2026-10-25 (live); stale on 2026-11-24 |
| Parsed numbers |
|
Evidence (10)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | research.google.com/colaboratory/faq.html | in the free version notebooks can run for at most 12 hours, depending on availability and usage patterns |
| 2026-10-22 | research.google.com/colaboratory/faq.html | 'Colab does not publish these limits, in part because they can vary over time' |
| 2026-10-22 | research.google.com/colaboratory/faq.html | virtual machines are deleted when idle for a while and have a maximum lifetime enforced by the service |
| 2026-10-22 | research.google.com/colaboratory/faq.html | 'In the version of Colab that is free of charge notebooks can run for at most 12 hours, depending on availability and your usage patterns', with 'Colab does not publish these limits, in part because they can vary over time' and 'access to expensive resources like GPUs is heavily restricted', all quoted |
| 2026-10-24 | research.google.com/colaboratory/faq.html | quoted: 'In the version of Colab that is free of charge notebooks can run for at most 12 hours, depending on availability and your usage patterns.' |
| 2026-10-24 | research.google.com/colaboratory/faq.html | quoted: 'Colab does not publish these limits, in part because they can vary over time.' |
| 2026-10-24 | research.google.com/colaboratory/faq.html | quoted: 'Virtual machines are deleted when idle for a while, and have a maximum lifetime enforced by the Colab service.' |
| 2026-10-24 | research.google.com/colaboratory/faq.html | quoted: 'Colab prioritizes interactive compute. Runtimes will time out if you are idle.' |
| 2026-10-24 | research.google.com/colaboratory/faq.html | quoted: 'In the version of Colab that is free of charge you are able to access VMs with a standard system memory profile.' — no number is given |
| 2026-10-24 | research.google.com/colaboratory/faq.html | quoted: 'on higher plans, additional capabilities such as background execution' — background execution is a paid-plan feature, so the free tier cannot run unattended |
More fields (16)
{
"shape": {
"vcpu": null,
"ram_gib": null,
"disk_gib": null,
"gpu_class": null,
"gpu_count": null,
"note": "the vendor publishes no numbers; GPU type 'varies over time'"
},
"billing_model": "free",
"granularity": {
"min_billed_s": null,
"rounding_s": null
},
"session_cap": {
"wall_clock_min": 720,
"idle_reap_s": null,
"restart_required": true,
"note": "12 h cap; idle timeout and VM lifetime unpublished"
},
"keep_rate": {
"value": 0.0,
"basis": "a deleted runtime costs nothing",
"source": "https://research.google.com/colaboratory/faq.html"
},
"persistent_storage": {
"included_gib": null,
"price_gib_month": null,
"durability": "none: the VM is deleted",
"survives_stop": false,
"survives_destroy": false,
"max_retention": "the runtime"
},
"network_out": {
"allow_outbound": true,
"allow_inbound_listen": false,
"ports": "outbound only",
"static_ip": false,
"ipv6": false,
"per_session_bytes": null
},
"free_tier_kind": "free-forever-with-unpublished-quota",
"m10h_usd": 0.0,
"m24h_usd": 0.0,
"wave2": {
"verdict_change": "trial-only -> qualifies-with-caveat on 2026-10-22"
},
"term": "none",
"price_arithmetic": "USD 0.00 for the free tier",
"unattended": false,
"unattended_mechanism": "none on the free tier; background execution is listed as a paid-plan capability",
"units": "RAM and disk are quoted in the vendor's own decimal GB, not GiB. The ram_gib and disk_gib keys carry the vendor's number unconverted, so where a vendor writes 30 GB the key holds 30 and the true value is about 27.9 GiB. No conversion was performed."
}Corrections
- 2026-10-22 Google Colab's free tier is a standing service, not a trial
Google Open Source Compute
caveatemailno cardDwrite-gated
| Free tier | Quoted: 'you can claim rewards for proactive improvements you've made to security in open source projects. Any patch (typically a merged GitHub pull request) that you can demonstrate to have improved the security of an in-scope project will be considered for a reward.' Published tiers: USD 500 for a one-liner, USD 2,000 for modest complexity, USD 7,500 for moderately complex, USD 15,000 for complicated high-impact work, with a '2x multiplier' and a '3x multiplier' for secure-by-design memory-safety work until the end of 2026. unit none; metered on wall-clock; cash-bounty |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | not stated |
| Identity | email not published |
| Limits |
A merged pull request in an in-scope project comes before any payment, and the multiplier that makes the ceiling interesting expires 2026-12-31, so the top of the range is a two-month window. |
| Binding | write-gated A merged pull request in an in-scope project comes before any payment, and the multiplier that makes the ceiling interesting expires 2026-12-31, so the top of the range is a two-month window. |
| Interface | rest inferred from the tooling the provider documents; not probed |
| Tooling | not stated |
| Composes with | A 3x multiplier on a tier-1 core data parser pays USD 45,000 for one API-compatible Rust rewrite, which is 4.5 times the USD 10,000 GitHub grant and 6.4 times the USD 7,000 Outreachy stipend, and it needs no company, no card and no interview. |
| Reliability | no reliability event recorded |
| Notes | This is the highest cash-per-action figure and the only one with no application, no deadline and no cohort. The gate is real skill plus scope: tier 1 is a fixed project list and tier 2 is the OSS-Fuzz set, so the work has to land in code somebody else depends on. Recorded after finding that the more famous Secure Open Source Rewards pilot is gone: sos.dev states it was decommissioned and names this programme as the replacement, which is why the dead programme belongs in the notes of the live one rather than in the corpus as a dead record. |
| Evidence rung | documented |
| Checked | 2026-10-22 (live); stale on 2026-11-21 |
Evidence (3)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | bughunters.google.com/open-source-security/patch-rewards | quoted the four reward tiers, the 2x and 3x memory-safety multipliers 'until the end of 2026', and the two worked examples: USD 10,000 for a Struts sandbox, USD 5,000 for a patched Go cors denial-of-service |
| 2026-10-22 | bughunters.google.com/about/rules/open-source/patch-rewards- | the in-scope definition: tier 1 is an explicit project list at github.com/google/bughunters/blob/main/patch-rewards-program/scope.md; any project that receives a Google vulnerability report is also in scope for the patch that fixes it; tier 2 is every project integrated into OSS-Fuzz |
| 2026-10-22 | sos.dev/ | SOS Rewards was decommissioned after paying USD 353,000 for 189 improvements in 24 months, and the page redirects individuals to this Patch Rewards programme instead |
More fields (1)
{
"wave2": {
"class": "bounty-individual",
"free_tier_kind": "cash-bounty",
"shape": null
}
}Unknown or withheld values (9)
$.egress.meteredegress-unknown$.egress.termsnot-stated$.identity_cost.detailprocess$.per_object_cap.notesprocess$.per_object_cap.statednot-applicable$.reliability_notenone-observed$.retention.notesprocess$.retention.statednot-applicable$.toolingnot-stated
Google Summer Compute
caveatemail?card?Dwrite-gated
| Free tier | Quoted: 'large projects range from $3000 to $6600 USD, depending on location', medium 'from $1500 to $3300', small 'from $750 to $1650', paid 45 percent after the midterm evaluation and 55 percent at the end, to a 12+ week project. unit none; metered on wall-clock; cash-stipend |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | not stated |
| Identity | email (placeholder) unverified: no page stating what a contributor account requires was read; the record's evidence covers the stipend bands, the 45/55 payment split and payment through Payoneer across 190+ countries, but not the application form. The email kind is inferred from an account-based application. A Payoneer account is needed to be paid, and whether Payoneer demands a bank account or an identity document before paying is not stated. No card, hold, debit or prepaid term is stated. |
| Limits |
One application window a year, a mentor organization must select the proposal, and payment is split across two evaluations, so nothing arrives before roughly week six. |
| Binding | write-gated One application window a year, a mentor organization must select the proposal, and payment is split across two evaluations, so nothing arrives before roughly week six. |
| Interface | unknown inferred from the tooling the provider documents; not probed |
| Tooling | not stated |
| Composes with | This is the only lane here that pays a person in a country the vendor has never heard of at a published per-country rate with no company and no card: USD 3,000 to 6,600 for twelve weeks, paid in two instalments, and the stipend is the only number either side publishes. |
| Reliability | no reliability event recorded |
| Notes | Two real limits. The intake is annual and closed for 2026; the next window is inferred to be March 2027 from the 2026 dates, not announced. The stipend is PPP-adjusted downward, so the published table has to be read for the country of residence. The role is titled 'contributor' rather than student on the pages read. |
| Evidence rung | documented |
| Checked | 2026-10-22 (live); stale on 2026-11-21 |
Evidence (4)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | developers.google.com/open-source/gsoc/help/student-stipends | quoted the 2026 stipend bands, the PPP adjustment, the 45/55 payment split, and payment through Payoneer across 190+ countries |
| 2026-10-22 | developers.google.com/open-source/gsoc/timeline | quoted the 2026 dates: contributor applications 2026-03-18 to 2026-04-02, accepted projects revealed 2026-05-01, and 'The timeline for Google Summer of Code can shift a bit each year' |
| 2026-10-22 | summerofcode.withgoogle.com/ | the live front page reads 'GSoC 2026 Contributors Announced!', so the 2026 round is complete |
| 2026-10-22 | summerofcode.withgoogle.com/how-it-works | the phases: organization applications, contributor application period, proposal review, community bonding, coding, two evaluations |
More fields (1)
{
"wave2": {
"class": "stipend",
"free_tier_kind": "cash-stipend",
"shape": null
}
}Unknown or withheld values (10)
$.egress.meteredegress-unknown$.egress.termsnot-stated$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.per_object_cap.notesprocess$.per_object_cap.statednot-applicable$.reliability_notenone-observed$.retention.notesprocess$.retention.statednot-applicable$.toolingnot-stated
Google Agent Platform Sandbox (Code Execution / Shell / Computer Use) Compute
caveatemail?card?Rwrite-gated
| Free tier | not published unit mixed; metered on wall-clock |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | not stated |
| Identity | email (placeholder) not published |
| Limits |
not stated |
| Binding | write-gated |
| Interface | unknown |
| Tooling | not stated |
| Composes with | not stated |
| Reliability | no reliability event recorded |
| Evidence rung | reported the pricing page and the free-features document were read on 2026-10-22. |
| Checked | 2026-10-23 (live); stale on 2026-11-06 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-02 | cloud.google.com/products/gemini-enterprise-agent-platform/p | not published |
| 2026-10-22 | cloud.google.com/products/gemini-enterprise-agent-platform/p | the Gemini Enterprise Agent Platform pricing page was re-read on 2026-10-22 and prices training, prediction and batch prediction per node-hour; the free-features document lists Agent Engine's vCPU-second allowance and states no computer-use allowance |
More fields (13)
{
"talaria_m10h_usd": 61.8,
"talaria_m24h_usd": 148.32,
"nemo_m10h_usd": null,
"keep_rate": 1.0,
"keep_rate_basis": null,
"billing_model": "unclassified",
"shape": null,
"session_cap": {
"wall_clock_min": null,
"idle_reap_s": null,
"restart_required": false,
"note": "no allowance or session clock published for computer use"
},
"persistent_storage": null,
"network_out": null,
"price_note": null,
"free_tier_kind": "one-time-credit",
"net": null
}Unknown or withheld values (20)
$.chainingnot-stated$.confidence_notepartly-held$.egress.meteredegress-unknown$.egress.termsnot-stated$.evidence[0].provedprocess$.extras.keep_rate_basisprocess$.extras.price_noteprocess$.free_tier.noteswithheld$.free_tier.statedwithheld$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_notenot-stated$.per_object_cap.notesprocess$.per_object_cap.statednot-stated$.reliability_notenone-observed$.retention.notesnot-extracted$.retention.statednot-stated$.toolingnot-stated
Gemini Notebook (formerly NotebookLM) — source corpus Storage
caveatemailno cardRper-object-cap
| Free tier | 100 notebooks/user x 50 sources/notebook x 200 MB per source = 1,000,000 MB ≈ 1 TB free. Also 500,000 words/source cap and no page limit. unit TB; metered on unstated |
|---|---|
| Price | none stated |
| Per-object cap | per-object-cap 200 MB (200000000 bytes) |
| Retention | not stated |
| Egress | No egress metering or fee anywhere in the product. Reads happen through Google's own UI/API, not a CDN URL. |
| Identity | email email |
| Limits |
per-object-cap 200 MB or 500,000 words. Notebook cap is per-user; source cap is per-notebook; 'sharing a notebook does not change the source limit for any collaborator'. Copy-protected PDFs are rejected outright. Consumer API is undocumented and Google has changed *enforced* limits without updating the published tables. |
| Binding | per-object-cap the per-object cap sets the shard size and therefore the object count; it bounds the total more tightly than any headline quota |
| Interface | rest, cli No official public API for consumer accounts. Working path is the reverse-engineered internal batchexecute API: notebooklm-py (19,838 stars, pushed 2026-10-21) exposes Python, CLI, MCP server, and a local REST server; notebooklm-mcp (202) and notebooklm-rest-api (91) wrap it. |
| Tooling | pip install notebooklm-py; notebooklm source add, notebooklm ask, notebooklm generate, notebooklm download. Google Cloud has a *separate* documented Enterprise API (notebooks.sources.uploadFile) that requires a Workspace/Cloud entitlement. |
| Composes with | 100 notebooks x 50 x 200 MB = 1 TB per Google account. Google accounts are free and email-only, so N accounts linearise. Upskilling to Plus/Pro only raises the per-notebook source cap (100/300), so the notebook count stays at 200/500 — the free tier already gives the widest notebook multiplier per dollar. |
| Reliability | READ-BACK IS EXTRACTED TEXT, NOT ORIGINAL BYTES. There is no 'download source' button; the platform keeps a static extracted copy and the original file is not retrievable. Third-party extensions (Notebook Clipper, nlm tools, notebooklm-tui) export sources as Markdown/TXT/PDF of the *extracted text*; notebooklm-tui saves PDFs as per-page PNGs. So this is the brief's rule 4 case — a host that gives storage but not your bytes back. It is a genuine 1 TB of free ingest, and a genuinely usable grounded-RAG corpus, but it is not a byte-exact archive. Rate limits apply and heavy use is throttled; the shared CI account-pool discussion in the wrapper's own docs is the tell. |
| Notes | 100 notebooks/user x 50 sources/notebook x 200 MB per source = 1,000,000 MB ≈ 1 TB free. Also 500,000 words/source cap and no page limit. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
| Parsed numbers |
|
Evidence (6)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | support.google.com/gemininotebook/answer/16269187?hl=en | 'The current limit is 500,000 words per source or up to 200MB for local uploads. There's no page limit.' |
| 2026-10-21 | support.google.com/gemininotebook/answer/16213268?hl=en | 'Notebooks | 100/user'; 'Sources | 50/notebook' (Standard column); 'Sharing a notebook does not change the source limit' |
| 2026-10-21 | github.com/teng-lin/notebooklm-py | 19,838 stars, description 'Full programmatic access to Gemini Notebook's features—including capabilities the web UI doesn't expose'; pushed_at 2026-10-21T19:38:11Z |
| 2026-10-21 | notebookclipper.com/docs/notebook/export-sources | export is 'the text Gemini Notebook extracted from each source, as Markdown, plain text or PDF' — not the original file |
| 2026-10-21 | ainotechou.com/notelm-source/ | '元のPDFには戻らない!HTML形式への変換' (does not return to the original PDF) |
| 2026-10-21 | docs.cloud.google.com/gemini/enterprise/notebooklm-enterpris | separate Enterprise API exists but is entitlement-gated |
Unknown or withheld values (4)
$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
Google Chat spaces as a quota-exempt attachment store (Chat API media.upload/media.download) Storage
caveatinstitutionno cardRper-object-cap
| Free tier | Per-record unlimited: 200 MB per attachment, and Chat messages AND attachments explicitly do NOT count toward Google storage quota. 600 attachment uploads/min/project x 200 MB = ~120 GB/min theoretical ingest. unit MB; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | 200 MB per file (200000000 bytes) |
| Retention | not stated |
| Egress | Not metered — downloaded through Chat API media.download (3,000 attachment reads/min/project), supports Range requests. No per-GB charge in the docs. |
| Identity | institution derived from the record's prose, not from an identity field, and no signup was executed: nt. 20 attachments per interactive message on web, but the API path is per-media-upload, so the binding constraint is the 600 uploads/min/project rate, and projects are multiplicativ |
| Limits |
200 MB per file hard cap; some file types blocked; 600 media.upload/min/project and 3,000 media.download/min/project; requires a Google Workspace account + Cloud project + Chat API enabled (personal @gmail.com accounts are not a supported developer surface; Chat API auth doc requires Workspace). Chat API quotas are per-minute only — the docs state 'no limit to the number of requests you can make per day' if you stay under them. |
| Binding | per-object-cap the per-object cap sets the shard size and therefore the object count; it bounds the total more tightly than any headline quota |
| Interface | rest, rclone rest |
| Tooling | Official Chat API with service-account (app) auth; media.upload for write, GET /v1/media/{name}?alt=media for read with Range. No dedicated rclone backend exists — this is a build-it-yourself REST store; google-api-* client libs cover both directions. |
| Composes with | Unlimited messages x 200 MB per attachment. 20 attachments per interactive message on web, but the API path is per-media-upload, so the binding constraint is the 600 uploads/min/project rate, and projects are multiplicative (30-project default quota per user). Storage is free because Chat attachments are quota-exempt — this is the strongest multiplier in the Google stack. |
| Reliability | Messages/attachments live in a space — deleting or archiving the space, or removing the app from it, can orphan the data; no documented retention guarantee. Blocked MIME types must be avoided. |
| Notes | Per-record unlimited: 200 MB per attachment, and Chat messages AND attachments explicitly do NOT count toward Google storage quota. 600 attachment uploads/min/project x 200 MB = ~120 GB/min theoretical ingest. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
| Parsed numbers |
|
Evidence (4)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | knowledge.workspace.google.com/admin/storage/google-workspac | under 'The following content doesn't count toward storage:' lists 'Google Chat messages and attachments' |
| 2026-10-21 | developers.google.com/workspace/chat/upload-media-attachment | 'You can upload file sizes up to 200 MB' + 'Some file types aren't supported' |
| 2026-10-21 | developers.google.com/workspace/chat/limits | 'Attachment writes per minute | media.upload | 600' and 'Attachment reads per minute | spaces.messages.attachments.get media.download | 3000' + 'there's no limit to the number of requests you can make per day' |
| 2026-10-21 | developers.google.com/workspace/chat/api/reference/rest/v1/m | download supports Range partial downloads via /v1/media/{name}?alt=media |
Unknown or withheld values (4)
$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
Google for Startups Cloud Program (Start / Scale tiers) Storage
caveatinstitutionno cardRper-object-cap
| Free tier | Scale and Scale AI tiers: '$200K - $350k in credits' for startups with verifiable Pre-Seed to Series A funding. Credits apply to 'most Google Cloud platform products ... used in a pay-as-you-go (PayG) or consumption billed format' — which includes Cloud Storage buckets. unit mixed; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | per-object-cap: GCS single-object max 5 TiB (5497558138880 bytes) |
| Retention | not stated |
| Egress | GCP egress is metered and comes out of the same credit pool; free egress only to Cloudflare-class partners or via the always-free 200 GB/mo Cloud Storage egress allowance. free_allowance: 200 GB per month |
| Identity | institution derived from the record's prose, not from an identity field, and no signup was executed: gram qualifications change annually'). Excludes educational institutions, government, nonprofits, personal blogs/content, dev shops and consultancies — a startup wrapper is required. iaas |
| Limits |
per-object-cap: GCS single-object max 5 TiB — not binding. request-rate: per-project GCS quotas (raiseable). retention: credits are time-boxed (program is annual, 'Program qualifications change annually'). Excludes educational institutions, government, nonprofits, personal blogs/content, dev shops and consultancies — a startup wrapper is required. |
| Binding | per-object-cap the per-object cap sets the shard size and therefore the object count; it bounds the total more tightly than any headline quota |
| Interface | s3-compatible, rest, rclone s3 |
| Tooling | rclone backend 'gcs' (Tier 1), gsutil, GCS XML/JSON API, GCS interop keys give an S3-compatible endpoint. |
| Composes with | Credits are per billing account and explicitly non-transferable ('Credits may only be used by the original applicant and cannot be shared, split, or resold'). The multiplier is *re-application in later years* — Google allows re-applying if credits went unused ('You may still be eligible for credits if you did not use your credits in a previous enrollment'). $200K at coldline-class rates is petabyte-scale for a year; the binding constraint becomes the 12-month expiry, not bytes. |
| Reliability | Requires a funded company, a matching domain email and a billing account owned by that domain; credits land in a specific billing account ID and cannot be moved afterwards. Year-over-year re-enrolment is not guaranteed. |
| Notes | Scale and Scale AI tiers: '$200K - $350k in credits' for startups with verifiable Pre-Seed to Series A funding. Credits apply to 'most Google Cloud platform products ... used in a pay-as-you-go (PayG) or consumption billed format' — which includes Cloud Storage buckets. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
| Parsed numbers |
|
Evidence (3)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | cloud.google.com/startup | 'Our Funded program tiers (Scale, and Scale AI) range from $200K - $350k in credits'; exclusion list; 'Credits may only be used by the original applicant' |
| 2026-10-21 | cloud.google.com/startup | 'Can I re-apply if I participated in this program a few years ago but did not use my credits? You may still be eligible ... if you did not use your credits' |
| 2026-10-21 | withheld | probe: GET cloud.google.com/startup -> 200 (numbers quoted live 2026-10-21) |
Unknown or withheld values (5)
$.evidence[2].urlprivate-url$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
Google Play Books Partner Center Storage
caveatemailno cardRper-object-cap
| Free tier | 2 GB per content file ('Each file must be less than 2 GB. This limit includes cover image files'); audiobook audio files under 100 MB each, but multiple parts allowed plus ZIP. unit GB; metered on unstated |
|---|---|
| Price | none stated |
| Per-object cap | per-object-cap 2 GB (2000000000 bytes) |
| Retention | not stated |
| Egress | Google-served to purchasers; no public hot-link. |
| Identity | email email |
| Limits | per-object-cap 2 GB; PDF/EPUB only for ebooks; PDF unencrypted required; EpubCheck validation required. |
| Binding | per-object-cap the per-object cap sets the shard size and therefore the object count; it bounds the total more tightly than any headline quota |
| Interface | unknown Partner Center web upload; also automated content fetching (ONIX/feed) per answer/2763162. |
| Tooling | Web + feed-based automation. |
| Composes with | 2 GB/file with multi-part naming (ISBN_1of3.pdf) means a large work can be split across parts with no stated part count — the sharding is built into the naming convention. |
| Reliability | Payout requires tax/payment setup; publishing is write-gated by Google's review. Ingest is heavily normalised — this is not a byte-preserving store. |
| Notes | 2 GB per content file ('Each file must be less than 2 GB. This limit includes cover image files'); audiobook audio files under 100 MB each, but multiple parts allowed plus ZIP. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
| Parsed numbers |
|
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | support.google.com/books/partner/answer/3424254?hl=en | 'Each file must be less than 2 GB. This limit includes cover image files'; audiobook 'Under 100 MB in size' |
| 2026-10-21 | help.vellum.pub/uploading/ | independent comparison table: Google Play 2048 MB, Apple Books 2048 MB, Amazon 650 MB, Kobo 100 MB, B&N 20 MB |
Unknown or withheld values (5)
$.free_tier.notesprocess$.interfaceinterface-unknown$.per_object_cap.notesnot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
Public Datasets hosting programme (Google Cloud) Storage
caveatcard?cardRwrite-gated
| Free tier | No published quota - Google 'pays for the hosting of these datasets'. Existing hosted sets (Sentinel-2, Landsat, ARCO-ERA5, NEXRAD) run to petabytes. unit unlimited; metered on unstated |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | Free public access to hosted datasets; anonymous API links work without auth. |
| Identity | card (placeholder) inferred from the category 'iaas'; the source record states no identity cost, so this is a class assumption, not evidence |
| Limits |
There is no application form. The only stated route is 'If you have any questions about listing your public dataset on Cloud Storage, please contact us at [address removed]'. Effectively invitation-only, and the catalogue has grown slowly relative to AWS's 300+ PB. |
| Binding | write-gated the write is a queue with a person or a rule in it, so the upload rate is set by somebody else |
| Interface | s3, rclone s3 |
| Tooling | GCS API / gcloud storage / rclone 'gcs'. |
| Composes with | Not applicable - this is Google's own catalogue, not an open intake. |
| Reliability | No published process means no predictable outcome; do not plan around it. |
| Notes | No published quota - Google 'pays for the hosting of these datasets'. Existing hosted sets (Sentinel-2, Landsat, ARCO-ERA5, NEXRAD) run to petabytes. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | docs.cloud.google.com/storage/docs/public-datasets | 'Google pays for the hosting of these datasets'; the entire intake instruction is an email address |
| 2026-10-21 | cloud.google.com/datasets | catalogue is Google-owned, commercial and government datasets only |
Unknown or withheld values (6)
$.free_tier.notesprocess$.identity_cost.kindidentity-defaulted$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
Google Cloud Free Trial (USD 300) Compute
trialcardcardphoneDvolume-quota
| Free tier | Signing up for the Free Trial creates a Free Trial billing account that is preloaded with $300 in free Welcome credit which is valid for 90 days. unit mixed; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not applicable: the offer is a spending balance, not a stored object |
| Retention | the balance is deleted at day 90 or when it is spent, whichever comes first; the trial cannot be paused or extended (90 days) |
| Egress | egress bills against the credit at list price; the separate always-free allowance is 1 GB a month from North America |
| Identity | card a credit card or other payment method is required at signup, with a temporary authorisation hold of USD 0.00 to USD 1.00; a consumer Google account is accepted and the page does not ask for a work address; bank or phone verification applies in some countries |
| Limits |
the credit cannot pay for the Gemini API in AI Studio, for a generative AI partner model sold as a managed API, for Windows Server images, or for Marketplace fees |
| Binding | volume-quota the 90-day clock ends the balance before any usage ceiling is reached |
| Interface | rest, cli the Google Cloud console, gcloud CLI and every product API |
| Tooling | gcloud, the client libraries, Terraform, and every tool that speaks a Google Cloud API |
| Composes with | this is the general-purpose credit behind a Google Cloud deployment; it pairs with the separate always-free e2-micro record, which has no end date |
| Reliability | the credit expires at day 90 even after the account is upgraded to a paid billing account, and the trial cannot be paused |
| Notes | This is a different record from google-cloud-free-tier, which carries the e2-micro allowance that has no end date. The trial is the largest single balance Google publishes to an individual, and the exclusions decide what it is worth: no GPU on a trial billing account, and no Gemini API spend. A workload that plans to train on a GPU cannot use this credit for it. |
| Evidence rung | documented the free-features page was read on 2026-10-23 and states the amount, the 90-day window, the card requirement and the product exclusions. No account was created, so the identity gate is the vendor's sentence and not a completed signup |
| Checked | 2026-10-24 (live); stale on 2026-11-23 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-23 | cloud.google.com/free/docs/free-cloud-features | 'Signing up for the Free Trial creates a Free Trial billing account that is preloaded with $300 in free Welcome credit which is valid for 90 days', quoted; the same page states the card requirement, the 90-day expiry after upgrade, and the Gemini API and GPU exclusions |
More fields (8)
{
"free_tier_kind": "one-time-credit",
"credit_usd": 300,
"credit_days": 90,
"recurring": false,
"reachable_by": "consumer email plus a card",
"excluded_products": "Gemini API in AI Studio, partner managed models, GPUs on a trial account, Windows Server, Marketplace",
"hourly_usd": {
"e2_standard_4": 0.13402284,
"e2_standard_2_derived": 0.06701142
},
"rate_basis": "e2-standard-4 read from the Google Cloud VM instance pricing page on 2026-10-23; e2-standard-2 is half of it and is derived, not read"
}Google Cloud external passthrough Network Load Balancer Egress and tunnels
paidcardcardDwrite-gated
| Free tier | No free tier is stated for the load balancer. Backends must be instance groups or NEGs: These backends must be in the same region and project. unit none; metered on none |
|---|---|
| Price | unverified |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | No byte allowance: a forwarding rule is billed hourly and the traffic is billed at Cloud egress rates. The price page could not be read - its extraction returned an unrelated domain-price table.; ingress: tcp+udp free_allowance: none published |
| Identity | card A Google Cloud project with billing enabled; the free trial credit is a credit, not a free tier for this product. |
| Limits |
|
| Binding | write-gated At 1 TiB/month the first thing that stops the client is not a byte count but the backend requirement: the service must already run in the same GCP project and region, which means the origin already has a public address of its own. |
| Interface | rest, cli a forwarding rule selects address, protocol and ports and points at one regional backend service |
| Tooling | gcloud, the API or the console; a forwarding rule plus a backend service and health check |
| Composes with | only with a GCP backend, so it chains with a free GCP e2-micro but that VM already has a public IP |
| Reliability | none recorded |
| Notes | UDP is documented and the address can be Premium Tier, which is anycast to the nearest region. It still adds no reachability a GCP VM's own external IP does not already provide, and it is regional rather than global for L4. Hybrid NEGs can point at on-premises hosts through Cloud VPN or Interconnect, which is a route but not a cheap one. |
| Evidence rung | documented The protocol list and the backend constraint are quoted from the page read today. The price is unverified because the pricing page extraction failed, so the record cannot support a cost claim. |
| Checked | 2026-10-23 (live); stale on 2026-11-22 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-23 | cloud.google.com/load-balancing/docs/network | Backend service-based regional external passthrough Network Load Balancers can load-balance TCP, UDP, ESP, GRE, ICMP, and ICMPv6 traffic. |
| 2026-10-23 | cloud.google.com/load-balancing/docs/network | These backends must be in the same region and project but can be in different VPC networks. |
More fields (7)
{
"wg_path": "inside-udp",
"wg_detail": "The backend service forwards TCP, UDP, ESP, GRE, ICMP and ICMPv6, so a WireGuard datagram reaches a GCE backend on the forwarding rule's port.",
"ingress": "tcp+udp",
"volume_tib_month": null,
"volume_stated": "No byte allowance: a forwarding rule is billed hourly and the traffic is billed at Cloud egress rates. The price page could not be read - its extraction returned an unrelated domain-price table.",
"bytes_transit": "yes",
"transit_note": "Google's network carries the bytes to the backend"
}Unknown or withheld values (2)
$.per_object_cap.statednot-applicable$.retention.statednot-applicable
Google Compute
fails: identityinstitutionno cardDwrite-gated
| Free tier | Start tier is USD 2k for startups founded in the past 2 years with a working MVP, a clear business model and plans to seek venture funding soon. The Funded program tiers, Scale and Scale AI, range from USD 200K to USD 350k in credits and require verifiable Pre-Seed to Series A funding. unit none; metered on wall-clock; one-shot-credit |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | not stated |
| Identity | institution The application requires a business email matching the startup public website domain, a billing account administered by a user on that domain, and an active website describing company, team and product. Scale tier funding must come from a known investor; non-venture funding such as grants, loans or friends and family does not qualify. |
| Limits |
The domain-matched business email plus a verifiable funding round fail first, and Google states it screens for AI-generated site content, so a thin one-person site is rejected at review. |
| Binding | write-gated The domain-matched business email plus a verifiable funding round fail first, and Google states it screens for AI-generated site content, so a thin one-person site is rejected at review. |
| Interface | unknown inferred from the tooling the provider documents; not probed |
| Tooling | not stated |
| Composes with | The Start tier at USD 2,000 is the only slice a pre-funding founder could plausibly reach, and it needs a live MVP site. The USD 350,000 figure is locked behind an investor relationship that also constitutes the funding. |
| Reliability | no reliability event recorded |
| Notes | dossier/storage/google-for-startups-cloud-program.json records the USD 200K to 350k tiers as qualifies-with-caveat and derives identity_cost from prose. The live page shows the door is the funding round and the domain email, so the verdict here is fails-identity. This finding supersedes that record's verdict, not its numbers. |
| Evidence rung | documented |
| Checked | 2026-10-22 (live); stale on 2026-11-21 |
Evidence (3)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | cloud.google.com/startup | Start tier is USD 2k and the Funded tiers range from USD 200K to USD 350k, with Pre-Seed to Series A verifiable funding, quoted |
| 2026-10-22 | cloud.google.com/startup | the site is screened and applications may be rejected for a high volume of AI generated text, image or other content, quoted |
| 2026-10-22 | cloud.google.com/startup | non-venture funding such as private equity, bank loans, government innovation programs, grants, crowdfunding or friends and family do not qualify for the Scale tier, quoted |
More fields (1)
{
"wave2": {
"class": "startup-credit",
"free_tier_kind": "one-shot-credit",
"shape": null
}
}Unknown or withheld values (9)
$.egress.meteredegress-unknown$.egress.termsnot-stated$.interfaceinterface-unknown$.per_object_cap.notesprocess$.per_object_cap.statednot-applicable$.reliability_notenone-observed$.retention.notesprocess$.retention.statednot-applicable$.toolingnot-stated
Google Cloud research Compute
fails: identityinstitutionno cardDwrite-gated
| Free tier | 'PhD students conducting research may apply for $1000 in GCP Credit per year' and 'Faculty and Postdoctoral researchers are eligible for one award of up to $5,000 USD'. The application form requires 'Organization Name *' chosen by 'Begin typing and select institution from the provided list', a 'Link to your profile in your institution's faculty or researcher directory', a 'Google Cloud Billing Account ID', a 250-word proposal, and a pricing-calculator estimate. Review is '6-8 weeks'. unit none; metered on wall-clock; recurring-credit |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | not stated |
| Identity | institution not published |
| Limits |
the organisation field is a closed picker over an accredited-institution list plus a faculty-directory URL, so an unaffiliated applicant cannot fill the form at all. |
| Binding | write-gated the organisation field is a closed picker over an accredited-institution list plus a faculty-directory URL, so an unaffiliated applicant cannot fill the form at all. |
| Interface | rest inferred from the tooling the provider documents; not probed |
| Tooling | not stated |
| Composes with | Only for a student: the PhD line is USD 1,000 a year and renews, which buys roughly 800 TPU v5e chip-hours at USD 1.20, so it funds TRC's host VM and GCS bucket rather than the accelerator. The faculty line is one award per person, ever, so it does not renew. |
| Reliability | no reliability event recorded |
| Notes | Distinct from the TRC record and worth keeping separate, because the two Google programs have opposite doors: TRC takes a free-text affiliation, this one takes a picker over accredited institutions. The referring loophole is the only path an outsider might use, and it requires two university applicants who name you, so it is a finding about the door rather than a route. the compute-side reading is fails-identity, because the gate is the form field rather than the amount. |
| Evidence rung | documented |
| Checked | 2026-10-22 (live); stale on 2026-11-21 |
Evidence (4)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | web.archive.org/web/20260317102442id_/https://support.google | Wayback capture 2026-03-17 of the vendor help page: 'PhD students conducting research may apply for $1000 in GCP Credit per year', 'Faculty and Postdoctoral researchers are eligible for one award of up to $5,000 USD', 'Students on a Masters degree program are not eligible', 'we review and make decisions on applications within 6-8 weeks', 'Only one person per research proposal may apply', quoted |
| 2026-10-22 | edu.google.com/programs/credits/research/ | the live form's required fields: 'Organization Name *' with 'Begin typing and select institution from the provided list', 'Link to your profile in your institution's faculty or researcher directory', 'Google Cloud Billing Account ID *', 'The program is currently only available in the countries listed'; page 200 |
| 2026-10-22 | support.google.com/google-cloud-higher-ed/answer/10724468?hl | the live help page returned a bot challenge (429 through six providers), so the quoted terms come from the 2026-03-17 capture rather than the live page |
| 2026-10-22 | www.reddit.com/r/googlecloud/comments/1rrllgu/new_free_trial | quoted OP: 'you are no longer allowed to use free trial credits on Google Studio API anymore'; a reply suggests the change targets accounts created for free AI access; the OP's own workaround is to use Vertex AI instead [re-read 2026-10-22] |
More fields (1)
{
"wave2": {
"class": "research-credit",
"free_tier_kind": "recurring-credit",
"shape": {
"vcpu": null,
"ram_gib": null,
"gpu_class": null,
"gpu_count": null
}
}
}Unknown or withheld values (9)
$.egress.meteredegress-unknown$.egress.termsnot-stated$.identity_cost.detailprocess$.per_object_cap.notesprocess$.per_object_cap.statednot-applicable$.reliability_notenone-observed$.retention.notesprocess$.retention.statednot-applicable$.toolingnot-stated
Google Cloud Free Tier (e2-micro) Egress and tunnels
fails: egresscardcardDvolume-quota
| Free tier | 1 non-preemptible e2-micro VM instance per month in one of the following US regions: Oregon: us-west1. Iowa: us-central1. South Carolina: us-east1. 30 GB-months standard persistent disk. 1 GB of outbound data transfer from North America to all region destinations (excluding China and Australia) per month. unit GB; metered on downloaded-bytes |
|---|---|
| Price | USD 0 for the instance; about USD 3.65 a month for the external IPv4 at USD 0.005 an hour |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | 1 GB of outbound data transfer from North America to all region destinations (excluding China and Australia) per month; ingress: udp free_allowance: 1 GB of outbound data transfer from North America to all region destinations (excluding China and Australia) per month |
| Identity | card 'A Google Cloud billing account is required to access the Google Cloud Free Tier'; the account is created through the Free Trial, which preloads USD 300 of credit for 90 days and takes a payment method |
| Limits |
|
| Binding | volume-quota 1 GB a month is 33.3 MB a day, 0.09% of 1 TiB a month, so a run at a 1 TiB/month pace exhausts the allowance in about 39 minutes, which makes the byte allowance the first wall by three orders of magnitude; the second wall is the address: 'Static and ephemeral IP addresses in use on standard VM instances - USD 0.005 / 1 hour', which the Free Tier usage table does not list among the free amounts, so the free instance carries a public IPv4 bill of about USD 3.65 a month |
| Interface | ssh SSH to the instance's external IP, with a firewall rule for any further port |
| Tooling | wg-quick, nothing vendor-specific |
| Composes with | 1 GB a month cannot carry bulk traffic; it composes as a control plane or a stable WireGuard endpoint that hands off to Oracle, not as an origin |
| Reliability | no idle reclamation was found; the Free Tier has no end date, so the ceiling is the 1 GB allowance rather than the offer expiring |
| Notes | The 1 GB figure is stated on the Free Tier page, not inferred, and it is the Compute Engine line rather than the Cloud Storage or Cloud Run line, which are also 1 GB and 100 GB respectively. No reclamation rule for an idle e2-micro appears on the page: unlike Oracle's Always Free, the instance is not taken back for being idle. The address charge is an inference from two documented facts, the USD 0.005 an hour line and the absence of an address line in the Free Tier table. |
| Evidence rung | documented both figures are on Google's own pages read today |
| Checked | 2026-10-23 (live); stale on 2026-11-22 |
| Parsed numbers |
|
Evidence (4)
| date | source | what it proved |
|---|---|---|
| 2026-10-23 | cloud.google.com/free/docs/free-cloud-features | Compute Engine Free Tier reads '1 GB of outbound data transfer from North America to all region destinations (excluding China and Australia) per month' and names the three US regions for the e2-micro |
| 2026-10-23 | cloud.google.com/vpc/pricing | 'Static and ephemeral IP addresses in use on standard VM instances - USD 0.005 / 1 hour', and the same page's internet egress table, so the public address is billed separately from the free instance |
| 2026-10-23 | cloud.google.com/firewall/docs/firewalls | a firewall rule carries an IP protocol and a UDP destination port, and the default rules use udp:0-65535, so UDP ingress including 51820 is allowed by rule and not blocked by the platform |
| 2026-10-23 | cloud.google.com/free/docs/free-cloud-features | 'The Free Tier has no end date, but Google reserves the right to change the offering, including changing or eliminating usage limits, with 30 days advance notice.' |
More fields (7)
{
"wg_path": "native-conf",
"wg_detail": "an e2-micro is an ordinary Linux VM; VPC firewall rules accept a UDP protocol and a destination port ('IP Protocol' plus 'Destination port ... For TCP or UDP, you can specify all destination ports, or optionally specify a single destination port number'), and the default-allow-internal row in the same document shows udp:0-65535, so udp:51820 is one rule away and wg-quick runs natively",
"ingress": "udp",
"volume_tib_month": 0.00091,
"volume_stated": "1 GB of outbound data transfer from North America to all region destinations (excluding China and Australia) per month",
"bytes_transit": "yes",
"transit_note": "bytes leave through the VM's external address on Google's network and count against the 1 GB monthly allowance"
}Unknown or withheld values (2)
$.per_object_cap.statednot-applicable$.retention.statednot-applicable
Google Cloud Run Egress and tunnels
fails: egresscardcardIvolume-quota
| Free tier | CPU - First 180,000 vCPU-seconds free per month ... RAM - First 360,000 GiB-seconds free per month ... Requests - 2 million requests free per month. unit mixed; metered on request-count |
|---|---|
| Price | USD 0 |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | Outbound internet data transfer is charged at Google Cloud networking rates "with a free tier of 1GiB free data transfer within North America per month".; ingress: http-only free_allowance: Outbound internet data transfer is charged at Google Cloud networking rates "with a free tier of 1GiB free data transfer within North America per month". |
| Identity | card A Google Cloud billing account with a payment method; the free tier is a billing discount, not an account without a card. |
| Limits | |
| Binding | volume-quota 1 GiB/month free transfer is 0.001 TiB; egress past it is billed at Premium tier rates. The transport fails before that. |
| Interface | rest, cli gcloud CLI, REST API, and console. |
| Tooling | gcloud, Dockerfile, or source deploy. |
| Composes with | Cloud Run can host an HTTP API; a WireGuard endpoint needs a VM, not a service. |
| Reliability | Request timeout caps a stream; WebSockets must reconnect. Scale to zero adds cold starts. |
| Notes | Cloud Run also requires a card, so it fails identity for a no-card client even before the transport does. Included here because it is the most common answer to "where do I run this". |
| Evidence rung | inferred Cloud Run publishes no page that says "UDP is not supported"; the conclusion follows from the container contract requiring an HTTP listener on one port, and from Google's own developer forum answer that it will not work with UDP. A first-party negative statement would raise this to documented. |
| Checked | 2026-10-23 (live); stale on 2026-11-22 |
| Parsed numbers |
|
Evidence (4)
| date | source | what it proved |
|---|---|---|
| 2026-10-23 | cloud.google.com/run/pricing | Free tier: 180,000 vCPU-s, 360,000 GiB-s, 2M requests; 1 GiB free data transfer within North America |
| 2026-10-23 | docs.cloud.google.com/run/docs/container-contract | "The ingress container within an instance must listen for requests on 0.0.0.0 on the port to which requests are sent." |
| 2026-10-23 | docs.cloud.google.com/run/docs/triggering/websockets | "WebSockets streams are HTTP requests" subject to the request timeout |
| 2026-10-23 | stackoverflow.com/questions/56525366/how-to-deploy-an-udp-se | third party: the container contract requires an HTTP server, so UDP will not work |
More fields (7)
{
"wg_path": "none",
"wg_detail": "No raw port. \"The ingress container within an instance must listen for requests on 0.0.0.0 on the port to which requests are sent\"; requests are HTTP, and WebSocket streams \"are HTTP requests\".",
"ingress": "http-only",
"volume_tib_month": 0.000977,
"volume_stated": "Outbound internet data transfer is charged at Google Cloud networking rates \"with a free tier of 1GiB free data transfer within North America per month\".",
"bytes_transit": "no",
"transit_note": "Google's front end terminates HTTP and forwards requests."
}Unknown or withheld values (2)
$.per_object_cap.statednot-applicable$.retention.statednot-applicable
Firebase Realtime Database Storage
fails: egressemail?card?Rper-ip-shaping
| Free tier | 1 GB stored on the Spark (no-cost) plan; 10 GB/month downloaded (~360 MB/day) unit GB; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | Metered on Blaze at $1/GB, with 'No-cost up to 360 MB/day (equivalent to ~10 GB/month)' on Spark. 360 MB/day vs 500 GB/day = 1,400x short. free_allowance: 10 GB per month |
| Identity | email (placeholder) not published |
| Limits |
1 GB stored; 10 MB max string (so ~7.5 MB of binary as base64); 256 MB max response per read; 32 levels max depth; 768 byte key length; 1,000 writes/sec per DB; 100 simultaneous connections on Spark; 10 GB/month download |
| Binding | per-ip-shaping one stream is shaped, and a user's throughput is the shaped rate times the number of streams |
| Interface | rest, cli rest |
| Tooling | Firebase SDKs, REST API (PUT/PATCH/GET .json), Admin SDK, firebase-tools CLI; data is a JSON tree |
| Composes with | 1 GB per database instance; multiple instances are possible via sharding, but the Spark plan limits you to a small number of instances and each shares the project's 10 GB/month download. 500 GB would need 500 projects. |
| Reliability | Durable and simple, but base64 in JSON costs +33% and the 10 MB per-string cap means chunking into ~7.5 MB pieces. The 100-connection Spark limit and 360 MB/day download cap both bind hard. |
| Notes | 1 GB stored on the Spark (no-cost) plan; 10 GB/month downloaded (~360 MB/day) |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | firebase.google.com/pricing | 'Realtime Database | GB stored 1 GB (no-cost)'; 'GB downloaded 10 GB/month (no-cost)'; 'No-cost up to 360 MB/day (equivalent to ~10 GB/month) Then $1/GB' |
| 2026-10-21 | firebase.google.com/docs/database/usage/limits | 'Maximum size of a string | 10 MB'; 'Size of a single response served by the database | 256 MB'; 'Maximum depth of child nodes | 32'; 'Write rate | 1,000 writes/second'; 'The Spark plan limit on simultaneous connections is 100' |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "1 GB stored on the Spark (no-cost) plan; 10 GB/month downloaded (~360 MB/day)",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (8)
$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
Firebase Storage (Cloud Storage for Firebase) Storage
fails: egressemail?card?Rper-ip-shaping
| Free tier | 5 GB stored on Spark (no-cost) plan; 1 GB downloaded per day; 50k download ops/day unit GB; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | 1 GB/day download (~30 GB/month) free, then $0.12/GB; outbound networking 5 GB/month. The 1 GB/DAY download cap is the binding constraint, not the 5 GB stored. free_allowance: 30 GB per month |
| Identity | email (placeholder) not published |
| Limits |
1 GB/day egress (~30 GB/mo) is the practical ceiling — a 5 GB store is refetchable only 6x/month. Download-op cap 50k/day. Spark also caps Cloud Storage at 5 GB-months and 100 GB/month downloaded for the *default* bucket variant. No billing = hard stop, not overage. |
| Binding | per-ip-shaping one stream is shaped, and a user's throughput is the shaped rate times the number of streams |
| Interface | rest, rclone rest |
| Tooling | Firebase SDKs (web/Android/iOS), Admin SDK, gsutil/gcloud storage against the underlying GCS bucket, firebase storage:upload; rclone can hit the bucket with GCS (not rclone's 'firebase' — that is the Realtime DB). Public download URLs via getDownloadURL / Firebase download tokens. |
| Composes with | 5 GB per PROJECT. Projects per Google account on Spark are limited to 'a small number of projects (usually around 5-10)' — so ~25-50 GB/account, not a big multiplier. Multiple Google accounts is the only axis. Deleting a project takes 30 days and still counts against quota, which badly hurts churn-and-burn cycling. |
| Reliability | Hard daily reset at midnight; exceeding it fails requests until reset. Project deletion is 30 days and blocked from quota reuse, so account cycling is slow. |
| Notes | 5 GB stored on Spark (no-cost) plan; 1 GB downloaded per day; 50k download ops/day |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
| Parsed numbers |
|
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | firebase.google.com/pricing | 'GB stored | No-cost up to 5 GB | Then $0.026/GB', 'GB downloaded | No-cost up to 1 GB/day', 'Download operations | No-cost up to 50K/day' |
| 2026-10-21 | firebase.google.com/docs/projects/learn-more | 'Spark pricing plan: Project-creation quota is limited to a small number of projects (usually around 5-10)' |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "5 GB stored on Spark (no-cost) plan; 1 GB downloaded per day; 50k download ops/day",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (8)
$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
Google Artifact Registry (incl. the gcr.io domain) Storage
fails: egresscountry-phoneno cardphoneRdisagrees
Sibling records disagree:
- identity_kind:
storage/google-artifact-registrysays country-phone, storage/google-artifact-registry-2 says email - verdict:
storage/google-artifact-registrysays fails-egress, storage/google-artifact-registry-3 says fails-quota - identity_kind:
storage/google-artifact-registrysays country-phone, storage/google-artifact-registry-3 says email
| Free tier | 0.5 GiB-month, once per billing account, shared across every project on it: 'Artifact Registry Storage | 0 gibibyte month to 0.5 gibibyte month | $0.00 (Free) / 1 gibibyte hour, per 1 month / account'; above that $0.000136986/GiB-hour (~$0.10/GiB-month). The page explicitly closes the multi-project loophole: 'Pricing applies to billing accounts, not individual Google Cloud projects ... the limit for the storage free tier is for total usage across all the attached projects.' unit GiB; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | NOT free — this is the clearest egress failure among the three hyperscalers: 'Pricing for outbound Internet data transfer is based on Premium tier pricing. This is network traffic leaving a repository to a client that is not a Google product, such as using a local server that is downloading artifacts.' Intra-Google traffic is $0.00, and same-continent repository-to-service moves are $0.00, but anything to the open internet is metered at Google's premium internet egress rates. There is no anonymous or public-repo egress allowance. |
| Identity | country-phone derived from the record's prose, not from an identity field, and no signup was executed: gibibyte hour, per 1 month / account'; above that $0.000136986/GiB-hour (~$0.10/GiB-month). The page explicitly closes the multi-project loophole: 'Pricing applies to billing account |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | oci oci |
| Tooling | docker/ORAS/gcloud artifacts; gcr.io domain repositories are billed as Artifact Registry. |
| Composes with | 0.5 GiB per billing account is not worth chaining; a second billing account requires a second payment instrument. |
| Reliability | Stable, but the free tier is effectively decorative and the egress meter is the hard wall. |
| Notes | 0.5 GiB-month, once per billing account, shared across every project on it: 'Artifact Registry Storage | 0 gibibyte month to 0.5 gibibyte month | $0.00 (Free) / 1 gibibyte hour, per 1 month / account'; above that $0.000136986/GiB-hour (~$0.10/GiB-month). The page explicitly closes the multi-project loophole: 'Pricing applies to billing accounts, not individual Google Cloud projects ... the limit for the storage free tier is for total usage across all the attached projects.' |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (3)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | cloud.google.com/artifact-registry/pricing | '0 gibibyte month to 0.5 gibibyte month $0.00 (Free)', 'Pricing applies to billing accounts, not individual Google Cloud projects' |
| 2026-10-21 | cloud.google.com/artifact-registry/pricing | 'Pricing for outbound Internet data transfer is based on Premium tier pricing' |
| 2026-10-21 | cloud.google.com/network-tiers/pricing | 'Premium Tier egress is priced at internet egress rates' |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "0.5 GiB-month, once per billing account, shared across every project on it: 'Artifact Registry Storage | 0 gibibyte month to 0.5 gibibyte month | $0.00 (Free) / 1 gibibyte hour, per 1 month / account'; above that $0.000136986/GiB-hour (~$0.10/GiB-month). The page explicitly closes the multi-project loophole: 'Pricing applies to billing accounts, not individual Google Cloud projects ... the limit for the storage free tier is for total usage across all the attached projects.'",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (6)
$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
google artifact registry Storage
fails: egressemail?card?Rdisagrees
Sibling records disagree:
- identity_kind:
storage/google-artifact-registrysays country-phone, storage/google-artifact-registry-2 says email - verdict:
storage/google-artifact-registry-2says fails-egress, storage/google-artifact-registry-3 says fails-quota
| Free tier | Free tier is 1 GiB-month of storage per billing account, shared across all projects, and egress between regions is $0.01-0.05/GB. Fails both the quota and egress filters. unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | Free tier is 1 GiB-month of storage per billing account, shared across all projects, and egress between regions is $0.01-0.05/GB. Fails both the quota and egress filters. |
| Notes | Free tier is 1 GiB-month of storage per billing account, shared across all projects, and egress between regions is $0.01-0.05/GB. Fails both the quota and egress filters. |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | cloud.google.com/artifact-registry/pricing | Free tier is 1 GiB-month of storage per billing account, shared across all projects, and egress between regions is $0.01-0.05/GB. Fails both the quota and egress filters. |
More fields (1)
{
"dead_end": {
"failure_class": "egress",
"why": "Free tier is 1 GiB-month of storage per billing account, shared across all projects, and egress between regions is $0.01-0.05/GB. Fails both the quota and egress filters.",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
google cloud source repositories Storage
fails: egressemail?card?R
| Free tier | 50 GB/month storage and 50 GB/month egress on the free tier, but the product is end-of-sale and being replaced by Secure Source Manager. Also a git-only interface — poor fit for 10 TiB of ra unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | 50 GB/month storage and 50 GB/month egress on the free tier, but the product is end-of-sale and being replaced by Secure Source Manager. Also a git-only interface — poor fit for 10 TiB of ra |
| Notes | 50 GB/month storage and 50 GB/month egress on the free tier, but the product is end-of-sale and being replaced by Secure Source Manager. Also a git-only interface — poor fit for 10 TiB of ra |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | cloud.google.com/free/docs/free-cloud-features | 50 GB/month storage and 50 GB/month egress on the free tier, but the product is end-of-sale and being replaced by Secure Source Manager. Also a git-only interface — poor fit for 10 TiB of ra |
| 2026-10-21 | docs.cloud.google.com/source-repositories/docs | cited as the basis of the rejection: 50 GB/month storage and 50 GB/month egress on the free tier, but the product is end-of-sale and being replaced by Secure Source Manager. Also a git-only interface — poor fit for 10 TiB of ra |
More fields (1)
{
"dead_end": {
"failure_class": "egress",
"why": "50 GB/month storage and 50 GB/month egress on the free tier, but the product is end-of-sale and being replaced by Secure Source Manager. Also a git-only interface — poor fit for 10 TiB of ra",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
Google Colab (paid tiers) Compute
fails: quotacardcardDrequest-ratedisagrees
Sibling records disagree:
- verdict:
compute/google-colabsays qualifies-with-caveat, compute/google-colab-paid says fails-quota - identity_kind:
compute/google-colabsays email, compute/google-colab-paid says card
| Free tier | no published quota on any tier: Colab states its limits vary and are not published, and paid tiers fall back to the free policies when compute units are exhausted unit none; metered on wall-clock; a paid tier whose limits are unpublished cannot be budgeted |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | a VM is deleted when idle and has a maximum lifetime enforced by the service |
| Egress | not published |
| Identity | card a paid tier requires a payment |
| Limits |
a documented absence of limits is a discretionary throttle, not a quota, and it does not become one when paid for |
| Binding | request-rate the compute-unit allowance binds and the policies behind it are unpublished, so there is no ceiling to schedule against |
| Interface | unknown a notebook in the browser |
| Tooling | Jupyter; compute units on the paid tiers |
| Composes with | none for a scheduled workload: the ceiling is discretionary on every tier |
| Reliability | no reliability event recorded |
| Notes | The paid tiers do not publish a quota either. A workload that needs a published ceiling cannot be scheduled here at any price, which is a stronger statement than 'the free tier is small'. |
| Evidence rung | documented |
| Checked | 2026-10-22 (live); stale on 2026-11-21 |
| Parsed numbers |
|
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | research.google.com/colaboratory/faq.html | Colab does not publish its limits because they vary over time; paid users revert to free-tier policies when units run out |
More fields (9)
{
"shape": null,
"billing_model": "wall-clock",
"session_cap": null,
"keep_rate": null,
"persistent_storage": null,
"network_out": null,
"free_tier_kind": "none",
"m10h_usd": 19.91,
"m24h_usd": 19.91
}Unknown or withheld values (5)
$.egress.meteredegress-unknown$.interfaceinterface-unknown$.per_object_cap.notesprocess$.per_object_cap.statednot-applicable$.reliability_notenone-observed
Google Cloud Free Trial Egress and tunnels
fails: quotacardcardDvolume-quota
| Free tier | Signing up for the Free Trial creates a Free Trial billing account that is preloaded with $300 in free Welcome credit which is valid for 90 days. unit mixed; metered on downloaded-bytes |
|---|---|
| Price | USD 300 of credit |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | $300 of credit at $0.12 per GiB for the first 1,024 GiB per month, so a 90-day trial spends $100 a month and buys 826 GiB a month, or 2.57 TiB in total; ingress: udp free_allowance: $300 of credit at $0.12 per GiB for the first 1,024 GiB per month, so a 90-day trial spends $100 a month and buys 826 GiB a month, or 2.57 TiB in total |
| Identity | card During the sign up, you must provide a credit card or other payment method that is valid for the period of the Free Trial. Depending on your country, you might also need to verify your bank account. |
| Limits |
|
| Binding | volume-quota at 1 TiB a month the money runs out first: $100 a month of credit buys 826 GiB, so the trial carries 0.86 TiB a month and never reaches the 90-day expiry with credit left. The bandwidth is not the limit; the 2 Gbps egress ceiling on an e2-micro is far above 1 TiB a month. |
| Interface | cli, rest gcloud compute and the Compute Engine REST API |
| Tooling | wg-quick on the VM; Google Cloud CLI for the instance and firewall rule |
| Composes with | the origin half of a tunnel: run wg-quick here and the client connects straight to the public IPv4, with the 2.57 TiB credit as the entire budget |
| Reliability | a trial account that is not upgraded has its projects and resources stopped when the 90 days lapse or the credit is spent |
| Notes | Arithmetic: $300 over 90 days is a $100 monthly budget. The first 1 GiB of each month is free, then $0.12 per GiB: $100 buys 1 + 825 GiB = 826 GiB = 0.81 TiB a month. Spent in one month instead, $300 buys 1 free GiB, 1,023 GiB at $0.12 = $122.76, then $177.24 at $0.11 = 1,611 GiB, so 2,635 GiB = 2.57 TiB total. Either way the credit is short of three months at 1 TiB a month, so the credit binds and the 90-day clock does not. Candidates considered: the Free Tier e2-micro line, which gives '1 GB of outbound data transfer from North America' a month and is three orders of magnitude short. The free-trial page also bars GPUs, quota increases and Windows images on a trial account, none of which touch a WireGuard ingress. |
| Evidence rung | documented a probe would settle whether a fresh trial account can open a UDP port on a public IPv4 from a non-card-verified country |
| Checked | 2026-10-23 (live); stale on 2026-11-22 |
| Parsed numbers |
|
Evidence (3)
| date | source | what it proved |
|---|---|---|
| 2026-10-23 | cloud.google.com/free/docs/free-cloud-features | '$300 in free Welcome credit which is valid for 90 days' and 'you must provide a credit card or other payment method' |
| 2026-10-23 | cloud.google.com/vpc/network-pricing | 'Inbound data transfer No charge for inbound data transfer'; 'Network (data transfer out) TO North America (per GiB in USD) 0 gibibyte to 1 gibibyte $0.00 (Free)'; '1 gibibyte to 1,024 gibibyte $0.12'; '1,024 gibibyte to 10,240 gibibyte $0.11' |
| 2026-10-23 | cloud.google.com/vpc/network-pricing | 'During the Free Trial period, these prices are charged against the Free Trial credit amount', so egress is spendable against the $300 rather than excluded from it |
More fields (7)
{
"wg_path": "native-conf",
"wg_detail": "a full Compute Engine VM with a public IPv4 address; wg-quick installs and binds a UDP port with no vendor component in the path",
"ingress": "udp",
"volume_tib_month": 0.86,
"volume_stated": "$300 of credit at $0.12 per GiB for the first 1,024 GiB per month, so a 90-day trial spends $100 a month and buys 826 GiB a month, or 2.57 TiB in total",
"bytes_transit": "yes",
"transit_note": "the VM's outbound packets cross Google's network and are metered as internet data transfer out"
}Unknown or withheld values (2)
$.per_object_cap.statednot-applicable$.retention.statednot-applicable
Cloud Firestore Storage
fails: quotaemail?card?Rrequest-rate
| Free tier | 1 GiB stored data on the free tier; 50,000 document reads/day; 20,000 writes/day; 10 GiB/month outbound unit GiB; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | Metered — 10 GiB/month outbound free, then GCP rates. 10 GiB/month = ~340 MB/day, 1,500x short of 500 GB/day. free_allowance: 10 GiB per month |
| Identity | email (placeholder) not published |
| Limits |
1 GiB total; 1 MiB per document; 1 MiB - 89 bytes per field value; 50K reads/day; 20K writes/day; 7.5 KiB max index entry; only ONE free database per project |
| Binding | request-rate calls are bounded rather than bytes, which is fatal to a nightly job on a monthly counter |
| Interface | rest rest |
| Tooling | Firebase Admin SDK, client SDKs, REST/gRPC API, gcloud firestore export to a bucket; documents are JSON-ish with a native Bytes type (no base64 bloat) |
| Composes with | 1 GiB per database; 'Cloud Firestore allows exactly one free database per project'. 500 GB needs 500 projects, each with its own billing-less free quota and its own 10 GiB/month egress. |
| Reliability | Reliable; has a native Bytes type so no base64 overhead. The 20K writes/day cap means 1 GiB at 1 MiB/doc = 1,024 documents, which is trivially fine, but re-writing or migrating burns the quota. Single-field indexes add storage you do not control. |
| Notes | 1 GiB stored data on the free tier; 50,000 document reads/day; 20,000 writes/day; 10 GiB/month outbound |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (3)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | firebase.google.com/docs/firestore/quotas | 'Stored data | 1 GiB'; 'Outbound data transfer | 10 GiB per month'; 'Maximum size for a document | 1 MiB (1,048,576 bytes)'; 'Maximum size of a field value | 1 MiB - 89 bytes' |
| 2026-10-21 | firebase.google.com/docs/firestore/pricing | 'Document reads 50,000 per day'; 'Document writes 20,000 per day'; 'Cloud Firestore allows exactly one free database per project' |
| 2026-10-21 | withheld | arithmetic: 50,000 reads/day x 1 MiB max document = 52.4 GB/day of theoretical read — 10x short of 500 GB/day, and the 10 GiB/month egress cap cuts it to 340 MB/day |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "1 GiB stored data on the free tier; 50,000 document reads/day; 20,000 writes/day; 10 GiB/month outbound",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (9)
$.evidence[2].urlprivate-url$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
cloud identity free (as a source of free identities / unlimited tenants) Storage
fails: quotaemail?card?R
| Free tier | Attractive in theory (free identities, Drive for desktop supported), but it grants no Workspace storage entitlement and no shared drives, so it produces identities with zero bytes — the exac unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | Attractive in theory (free identities, Drive for desktop supported), but it grants no Workspace storage entitlement and no shared drives, so it produces identities with zero bytes — the exac |
| Notes | Attractive in theory (free identities, Drive for desktop supported), but it grants no Workspace storage entitlement and no shared drives, so it produces identities with zero bytes — the exac |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | cloud.google.com/identity/docs/editions | Attractive in theory (free identities, Drive for desktop supported), but it grants no Workspace storage entitlement and no shared drives, so it produces identities with zero bytes — the exac |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "Attractive in theory (free identities, Drive for desktop supported), but it grants no Workspace storage entitlement and no shared drives, so it produces identities with zero bytes — the exac",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
Cloud Storage for Firebase (Spark/Blaze) Storage
fails: quotacardcardRper-ip-shaping
| Free tier | 5 GB stored free, but only on the Blaze pay-as-you-go plan; Spark project buckets return 402/403. unit GB; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | 1 GB/day free for legacy *.appspot.com buckets, then $0.12/GB; 100 GB/month for *.firebasestorage.app buckets. Metered beyond that. free_allowance: 100 GB per month |
| Identity | card derived from the record's prose, not from an identity field, and no signup was executed: ark project buckets return 402/403. Requires attaching a credit card (Blaze). No free path at all since the Sept 2024 change — this is the clearest 'free that is not free' in the Google st |
| Limits |
5 GB stored; 20K upload ops/day (appspot) / 5K/month (firebasestorage.app); 1 GB/day or 100 GB/month egress; bucket creation post-Sept-2024 uses PROJECT_ID.firebasestorage.app naming. |
| Binding | per-ip-shaping one stream is shaped, and a user's throughput is the shaped rate times the number of streams |
| Interface | rest, rclone rest |
| Tooling | Firebase Storage SDK / GCS JSON+XML API (rclone 'gs'/'gcs'); buckets are ordinary GCS buckets so rclone works. |
| Composes with | 5 GB x N Firebase projects — each project needs a billing account on Blaze, which defeats the purpose since Blaze is a metered surface. |
| Reliability | Requires attaching a credit card (Blaze). No free path at all since the Sept 2024 change — this is the clearest 'free that is not free' in the Google stack. |
| Notes | 5 GB stored free, but only on the Blaze pay-as-you-go plan; Spark project buckets return 402/403. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
| Parsed numbers |
|
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | firebase.google.com/pricing | 'Storage | * Not applicable * | No-cost up to 5 GB Then $0.10/GB'; 'GB downloaded | No-cost up to 1 GB/day Then $0.12/GB' |
| 2026-10-21 | firebase.google.com/docs/storage/faqs-storage-changes-announ | 'If your Firebase project is on the Spark pricing plan, you won't have access to any Cloud Storage buckets … your API calls to buckets will return 402 or 403 errors' |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "5 GB stored free, but only on the Blaze pay-as-you-go plan; Spark project buckets return 402/403.",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (6)
$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
GCP research credits programme (Cloud for Higher Education) Storage
fails: quotainstitutionno cardRwrite-gated
| Free tier | $5,000 one-time for faculty/postdoc; $1,000 per year for PhD students. Credit-denominated, buys GCS at list price (~$20/TB-month Standard). unit TB; metered on unstated |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | Metered at standard GCS rates unless fronted by Cloudflare; credit covers it. |
| Identity | institution derived from the record's prose, not from an identity field, and no signup was executed: $5,000 one-time for faculty/postdoc; $1,000 per year for PhD students. Credit-denominated, buys GCS at list price (~$20/TB-month Standard). Cred |
| Limits |
Only available in the countries on Google's dropdown; must be a faculty member, postdoc, or PhD student at a listed institution; masters students ineligible. Faculty may apply only once, ever ('you are only eligible to apply one time'). 6-8 week review. One applicant per research proposal. |
| Binding | write-gated the write is a queue with a person or a rule in it, so the upload rate is set by somebody else |
| Interface | s3, rclone s3 |
| Tooling | GCS JSON/XML API, gsutil, rclone 'gcs' backend. |
| Composes with | PhD students can re-apply annually - the only genuinely recurring line here. $1,000/yr is roughly 4 TB-months of Standard storage, so it funds a working set, not an archive. |
| Reliability | Credits expire; GCS bills continue after the credit is exhausted. No data loss, just a bill. |
| Notes | $5,000 one-time for faculty/postdoc; $1,000 per year for PhD students. Credit-denominated, buys GCS at list price (~$20/TB-month Standard). |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (3)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | support.google.com/google-cloud-higher-ed/answer/10724468?hl | 'PhD students conducting research may apply for $1000 in GCP Credit per year'; 'Faculty and Postdoctoral researchers are eligible for one award of up to $5,000 USD' |
| 2026-10-21 | support.google.com/google-cloud-higher-ed/answer/10724468?hl | 'if you are a faculty member or postdoctoral researcher you are only eligible to apply one time' |
| 2026-10-21 | edu.google.com/programs/credits/research/ | 'The program is currently only available in the countries listed'; form requires organisation name and faculty-directory link |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "$5,000 one-time for faculty/postdoc; $1,000 per year for PhD students. Credit-denominated, buys GCS at list price (~$20/TB-month Standard).",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (6)
$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
gmail as bulk storage Storage
fails: quotaemail?card?R
| Free tier | Gmail shares the same 15 GB (or pooled Workspace) quota as Drive, and attachments are capped at 50 MB inbound / 25 MB outbound — so it multiplies nothing while consuming the same bytes. Stri unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | Gmail shares the same 15 GB (or pooled Workspace) quota as Drive, and attachments are capped at 50 MB inbound / 25 MB outbound — so it multiplies nothing while consuming the same bytes. Stri |
| Notes | Gmail shares the same 15 GB (or pooled Workspace) quota as Drive, and attachments are capped at 50 MB inbound / 25 MB outbound — so it multiplies nothing while consuming the same bytes. Stri |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | support.google.com/accounts/answer/9312312 | Gmail shares the same 15 GB (or pooled Workspace) quota as Drive, and attachments are capped at 50 MB inbound / 25 MB outbound — so it multiplies nothing while consuming the same bytes. Stri |
| 2026-10-21 | workspaceupdates.googleblog.com/2026/02/ending-larger-attach | cited as the basis of the rejection: Gmail shares the same 15 GB (or pooled Workspace) quota as Drive, and attachments are capped at 50 MB inbound / 25 MB outbound — so it multiplies nothing while consuming the same bytes. Stri |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "Gmail shares the same 15 GB (or pooled Workspace) quota as Drive, and attachments are capped at 50 MB inbound / 25 MB outbound — so it multiplies nothing while consuming the same bytes. Stri",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
google artifact registry (as a blob store) Storage
fails: quotaemail?card?Rdisagrees
Sibling records disagree:
- verdict:
storage/google-artifact-registrysays fails-egress, storage/google-artifact-registry-3 says fails-quota - identity_kind:
storage/google-artifact-registrysays country-phone, storage/google-artifact-registry-3 says email - verdict:
storage/google-artifact-registry-2says fails-egress, storage/google-artifact-registry-3 says fails-quota
| Free tier | 0.5 GB/month free — three orders of magnitude short. Only GC's own container/image storage is bigger, and that is already covered by the known GHCR slot. unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | 0.5 GB/month free — three orders of magnitude short. Only GC's own container/image storage is bigger, and that is already covered by the known GHCR slot. |
| Notes | 0.5 GB/month free — three orders of magnitude short. Only GC's own container/image storage is bigger, and that is already covered by the known GHCR slot. |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | cloud.google.com/free/docs/free-cloud-features | 0.5 GB/month free — three orders of magnitude short. Only GC's own container/image storage is bigger, and that is already covered by the known GHCR slot. |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "0.5 GB/month free — three orders of magnitude short. Only GC's own container/image storage is bigger, and that is already covered by the known GHCR slot.",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
Google Cloud Free Tier — Cloud Storage Storage
fails: quotacardcardRper-ip-shaping
| Free tier | 5 GB-months of regional storage per month (us-east1/us-west1/us-central1 only), 5,000 Class A ops, 50,000 Class B ops, 100 GB outbound/month from North America. unit GB; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | '100 GB of outbound data transfer from North America to all region destinations (excluding China and Australia) per month' — free, recurring monthly, applies to all GCP customers, not a trial. |
| Identity | card derived from the record's prose, not from an identity field, and no signup was executed: ottling surprises; the limit is purely quota size. 5 GB per billing account. Billing accounts are tied to a payment instrument, so account multiplication is not free/trivial. The real GCP lever i |
| Limits |
5 GB is ~1% of the 500 GB floor. 100 GB/month egress is 1/100th of what a 500 GB/day serving workload needs. Storage only free in 3 US regions. |
| Binding | per-ip-shaping one stream is shaped, and a user's throughput is the shaped rate times the number of streams |
| Interface | s3-compatible, rclone s3 |
| Tooling | GCS JSON/XML API, S3-compatible XML endpoint (rclone 'gcs' and 's3' with GCS provider); gsutil/gcloud storage; signed URLs for public reads. |
| Composes with | 5 GB per billing account. Billing accounts are tied to a payment instrument, so account multiplication is not free/trivial. The real GCP lever is not this tier — it is the $300/90-day trial and the startup credits (below). |
| Reliability | Rock solid, no data loss or throttling surprises; the limit is purely quota size. |
| Notes | 5 GB-months of regional storage per month (us-east1/us-west1/us-central1 only), 5,000 Class A ops, 50,000 Class B ops, 100 GB outbound/month from North America. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
| Parsed numbers |
|
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | cloud.google.com/free/docs/free-cloud-features | Cloud Storage row: '5 GB-months of regional storage (US regions only) per month ... 100 GB of outbound data transfer from North America ... per month' |
| 2026-10-21 | cloud.google.com/storage/pricing | Standard regional $0.000030137/GiB-hour ≈ $0.022/GiB-month (useful to convert credits to GB-months) |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "5 GB-months of regional storage per month (us-east1/us-west1/us-central1 only), 5,000 Class A ops, 50,000 Class B ops, 100 GB outbound/month from North America.",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (6)
$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
google cloud free trial ($300 / 90 days) Storage
fails: quotaemail?card?R
| Free tier | Explicitly time-boxed: 'The Free Trial is a 90-day program'; at day 90 or credit exhaustion 'your Free Trial billing account will be closed and all of its associated projects and resources w unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | Explicitly time-boxed: 'The Free Trial is a 90-day program'; at day 90 or credit exhaustion 'your Free Trial billing account will be closed and all of its associated projects and resources w |
| Notes | Explicitly time-boxed: 'The Free Trial is a 90-day program'; at day 90 or credit exhaustion 'your Free Trial billing account will be closed and all of its associated projects and resources w |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | cloud.google.com/free/docs/free-cloud-features | Explicitly time-boxed: 'The Free Trial is a 90-day program' at day 90 or credit exhaustion 'your Free Trial billing account will be closed and all of its associated projects and resources w |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "Explicitly time-boxed: 'The Free Trial is a 90-day program'; at day 90 or credit exhaustion 'your Free Trial billing account will be closed and all of its associated projects and resources w",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
google cloud public datasets (as an intake) Storage
fails: quotaemail?card?R
| Free tier | No application process exists. Google hosts only its own and partner datasets (Sentinel-2, Landsat, ARCO-ERA5, NEXRAD), and the only stated route for anyone else is an email address: 'If you unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | No application process exists. Google hosts only its own and partner datasets (Sentinel-2, Landsat, ARCO-ERA5, NEXRAD), and the only stated route for anyone else is an email address: 'If you |
| Notes | No application process exists. Google hosts only its own and partner datasets (Sentinel-2, Landsat, ARCO-ERA5, NEXRAD), and the only stated route for anyone else is an email address: 'If you |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | docs.cloud.google.com/storage/docs/public-datasets | No application process exists. Google hosts only its own and partner datasets (Sentinel-2, Landsat, ARCO-ERA5, NEXRAD), and the only stated route for anyone else is an email address: 'If you |
More fields (1)
{
"dead_end": {
"failure_class": "write-gated",
"why": "No application process exists. Google hosts only its own and partner datasets (Sentinel-2, Landsat, ARCO-ERA5, NEXRAD), and the only stated route for anyone else is an email address: 'If you",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
Google Cloud Shell Storage
fails: quotaemailno cardRretention-clamped
| Free tier | 'Cloud Shell provisions 5 GB of free persistent disk storage mounted as your $HOME directory ... on a per-user basis and is available across projects.' unit GB; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | Not specified in the quotas doc; the intended use is interactive admin, and there is no public-serving path. |
| Identity | email derived from the record's prose, not from an identity field, and no signup was executed: ely small. The 120-day deletion is real and is announced by email only. 5 GB per Google account, shared across all that account's projects. High account value (it's also a free shell wi |
| Limits | 5 GB hard; 50 h/week quota; VM terminates after 40 min idle (files outside $HOME lost); $HOME auto-deleted after 120 days of no access; web preview is auth-gated, not public. |
| Binding | retention-clamped a clock, not a quota: bytes leave whether or not the client is finished with them |
| Interface | sftp, scp, ssh, rclone sftp |
| Tooling | gcloud cloud-shell ssh, cloudshell scp, SFTP; the $HOME is a real ext4 disk you can rclone into. |
| Composes with | 5 GB per Google account, shared across all that account's projects. High account value (it's also a free shell with gcloud auth), so it chains well for *tooling*, not for capacity. |
| Reliability | Genuinely persistent and genuinely small. The 120-day deletion is real and is announced by email only. |
| Notes | 'Cloud Shell provisions 5 GB of free persistent disk storage mounted as your $HOME directory ... on a per-user basis and is available across projects.' |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | docs.cloud.google.com/shell/docs/quotas-limits | '5 GB of free persistent disk storage mounted as your $HOME'; 'default weekly Cloud Shell quota is 50 hours'; 'don't access Cloud Shell for 120 days, Cloud Shell automatically deletes your $HOME' |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "'Cloud Shell provisions 5 GB of free persistent disk storage mounted as your $HOME directory ... on a per-user basis and is available across projects.'",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (6)
$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
Google Cloud Storage Always Free tier Storage
fails: quotacardcardRper-ip-shaping
| Free tier | 5 GB-months of regional storage (US regions only) + 5,000 Class A ops + 50,000 Class B ops per month. unit GB; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | 100 GB/month outbound to all destinations excluding China/Australia — described in the Always Free table as '100 GB of outbound data transfer from North America'. Beyond that it is $0.12/GiB, metered. free_allowance: 100 GB per month |
| Identity | card derived from the record's prose, not from an identity field, and no signup was executed: e hot working set and egress beyond 100 GB/mo is metered at $0.12/GiB — a 500 GB/day serving load would cost ~$1,800/mo. 5 GB x N projects. Project creation is capped (~30 default be |
| Limits |
5 GB stored, 100 GB/mo egress, free tier restricted to us-east1/us-west1/us-central1; 5 TiB max object; anonymous writes are correctly refused. |
| Binding | per-ip-shaping one stream is shaped, and a user's throughput is the shaped rate times the number of streams |
| Interface | s3, rest, rclone s3 |
| Tooling | Native S3-interoperable XML API + JSON API; rclone 'gcs' backend; gsutil/gcloud storage; signed URLs; HMAC keys make it literally S3. |
| Composes with | 5 GB x N projects. Project creation is capped (~30 default before requesting increase) and soft-deleted projects still count for 30 days, so the practical ceiling is a few hundred GB, not TiB. |
| Reliability | Rock solid technically, but 5 GB is 0.05% of the hot working set and egress beyond 100 GB/mo is metered at $0.12/GiB — a 500 GB/day serving load would cost ~$1,800/mo. |
| Notes | 5 GB-months of regional storage (US regions only) + 5,000 Class A ops + 50,000 Class B ops per month. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
| Parsed numbers |
|
Evidence (4)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | cloud.google.com/free/docs/free-cloud-features | 'Cloud Storage | 5 GB-months of regional storage (US regions only) per month … 100 GB of outbound data transfer from North America … per month' |
| 2026-10-21 | cloud.google.com/storage/pricing | 'Data transfer to Worldwide Destinations (excluding Asia & Australia) (per GB) 0 to 10 TiB — $0.12 / 1 GiB' |
| 2026-10-21 | storage.googleapis.com/storage/v1/b/gcp-public-data-landsat | probe: GET -> 200 (public buckets are anonymously listable/readable — useful for reading, irrelevant for storage) |
| 2026-10-21 | storage.googleapis.com/gcp-public-data-landsat/probe.txt | probe: PUT (anonymous) -> 403 (no anonymous write path) |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "5 GB-months of regional storage (US regions only) + 5,000 Class A ops + 50,000 Class B ops per month.",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (6)
$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
Google Colab Storage
fails: quotainstitutionno cardR
| Free tier | No persistent disk at all. Notebook files live in Drive; the VM disk is per-session. unit mixed; metered on unstated |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | until the instance is recycled; not persistent |
| Egress | N/A — file hosting is outside the published allocation, so there is no serving path. |
| Identity | institution 'Virtual machines are deleted when idle'. iaas |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown wrapper-only |
| Tooling | google.colab Python package, Drive mount; no storage client because there is nothing to store to. |
| Composes with | None for storage. |
| Reliability | Confirmed both counts: file hosting is outside the published allocation AND zero persistence — the VM disk and the container are destroyed with the session. |
| Notes | No persistent disk at all. Notebook files live in Drive; the VM disk is per-session. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | research.google.com/colaboratory/faq.html | not published |
| 2026-10-21 | research.google.com/colaboratory/faq.html | 'Virtual machines are deleted when idle for a while'; 'notebooks can run for at most 12 hours' |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "No persistent disk at all. Notebook files live in Drive; the VM disk is per-session.",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (7)
$.evidence[0].provedwithheld$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.interfaceinterface-unknown$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted
google colab as a serving box Storage
fails: quotaemail?card?R
| Free tier | Both disqualifiers land at once: file hosting and media serving are in the provider's published list of restricted activities, and the VM is deleted on idle. unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | Both disqualifiers land at once: file hosting and media serving are in the provider's published list of restricted activities, and the VM is deleted on idle. |
| Notes | Both disqualifiers land at once: file hosting and media serving are in the provider's published list of restricted activities, and the VM is deleted on idle. |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | research.google.com/colaboratory/faq.html | Both disqualifiers land at once: file hosting and media serving are in the provider's published list of restricted activities, and the VM is deleted on idle. |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "Both disqualifiers land at once: file hosting and media serving are in the provider's published list of restricted activities, and the VM is deleted on idle.",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
google colab free-tier runtime disk (as scratch storage) Storage
fails: quotaemail?card?R
| Free tier | VMs are deleted when idle and runti unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | VMs are deleted when idle and runti |
| Notes | VMs are deleted when idle and runti |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | research.google.com/colaboratory/faq.html | VMs are deleted when idle and runti |
More fields (1)
{
"dead_end": {
"failure_class": "not-storage",
"why": "VMs are deleted when idle and runti",
"retry_when": "never; it is not storage",
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (14)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
google container registry (gcr.io) as a separate free tier Storage
fails: quotaemail?card?R
| Free tier | Not a separate quota. gcr.io-domain repositories are billed as Artifact Registry ('Repositories that support the gcr.io domain' are listed as storing artifacts subject to the same pricing), unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | Not a separate quota. gcr.io-domain repositories are billed as Artifact Registry ('Repositories that support the gcr.io domain' are listed as storing artifacts subject to the same pricing), |
| Notes | Not a separate quota. gcr.io-domain repositories are billed as Artifact Registry ('Repositories that support the gcr.io domain' are listed as storing artifacts subject to the same pricing), |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | cloud.google.com/artifact-registry/pricing | Not a separate quota. gcr.io-domain repositories are billed as Artifact Registry ('Repositories that support the gcr.io domain' are listed as storing artifacts subject to the same pricing) |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "Not a separate quota. gcr.io-domain repositories are billed as Artifact Registry ('Repositories that support the gcr.io domain' are listed as storing artifacts subject to the same pricing),",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
google docs / sheets as unlimited storage (uds, base64 embedding) Storage
fails: quotaemail?card?R
| Free tier | Closed. Collaborative files (Docs, Sheets, Slides, Drawings, Forms) created or edited after 2021-06-01 / 2022-05-02 now count toward quota. The stewardmcgown/uds repo (4.4k stars) still exis unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | Closed. Collaborative files (Docs, Sheets, Slides, Drawings, Forms) created or edited after 2021-06-01 / 2022-05-02 now count toward quota. The stewardmcgown/uds repo (4.4k stars) still exis |
| Notes | Closed. Collaborative files (Docs, Sheets, Slides, Drawings, Forms) created or edited after 2021-06-01 / 2022-05-02 now count toward quota. The stewardmcgown/uds repo (4.4k stars) still exis |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | support.google.com/accounts/answer/9312312 | Closed. Collaborative files (Docs, Sheets, Slides, Drawings, Forms) created or edited after 2021-06-01 / 2022-05-02 now count toward quota. The stewardmcgown/uds repo (4.4k stars) still exis |
| 2026-10-21 | knowledge.workspace.google.com/admin/storage/google-workspac | cited as the basis of the rejection: Closed. Collaborative files (Docs, Sheets, Slides, Drawings, Forms) created or edited after 2021-06-01 / 2022-05-02 now count toward quota. The stewardmcgown/uds repo (4.4k stars) still exis |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "Closed. Collaborative files (Docs, Sheets, Slides, Drawings, Forms) created or edited after 2021-06-01 / 2022-05-02 now count toward quota. The stewardmcgown/uds repo (4.4k stars) still exis",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
google groups as a file store Storage
fails: quotaemail?card?R
| Free tier | There is no file storage. Limits are message-activity limits: "Individual message size 25 MB", "Total message size 2.4 GB per hour" per group, "Total inbound emails per group 300 per 5 minut unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | There is no file storage. Limits are message-activity limits: "Individual message size 25 MB", "Total message size 2.4 GB per hour" per group, "Total inbound emails per group 300 per 5 minut |
| Notes | There is no file storage. Limits are message-activity limits: "Individual message size 25 MB", "Total message size 2.4 GB per hour" per group, "Total inbound emails per group 300 per 5 minut |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | knowledge.workspace.google.com/admin/groups/understand-group | There is no file storage. Limits are message-activity limits: "Individual message size 25 MB", "Total message size 2.4 GB per hour" per group, "Total inbound emails per group 300 per 5 minut |
| 2026-10-21 | webapps.stackexchange.com/questions/30506 | cited as the basis of the rejection: There is no file storage. Limits are message-activity limits: "Individual message size 25 MB", "Total message size 2.4 GB per hour" per group, "Total inbound emails per group 300 per 5 minut |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "There is no file storage. Limits are message-activity limits: \"Individual message size 25 MB\", \"Total message size 2.4 GB per hour\" per group, \"Total inbound emails per group 300 per 5 minut",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
google photos Storage
fails: quotaemail?card?R
| Free tier | 15 GB shared across the whole Google account (Gmail/Drive/Photos). Original quality counts against it, and 'Storage saver' re-encodes and resizes anything over 16 MP. Content deleted after 2 unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | 15 GB shared across the whole Google account (Gmail/Drive/Photos). Original quality counts against it, and 'Storage saver' re-encodes and resizes anything over 16 MP. Content deleted after 2 |
| Notes | 15 GB shared across the whole Google account (Gmail/Drive/Photos). Original quality counts against it, and 'Storage saver' re-encodes and resizes anything over 16 MP. Content deleted after 2 |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | support.google.com/photos/answer/6220791 | 15 GB shared across the whole Google account (Gmail/Drive/Photos). Original quality counts against it, and 'Storage saver' re-encodes and resizes anything over 16 MP. Content deleted after 2 |
More fields (1)
{
"dead_end": {
"failure_class": "retention",
"why": "15 GB shared across the whole Google account (Gmail/Drive/Photos). Original quality counts against it, and 'Storage saver' re-encodes and resizes anything over 16 MP. Content deleted after 2",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
Google Photos Library API (append-only upload) Storage
fails: quotainstitutionno cardR
| Free tier | No quota of its own — but every item uploaded counts against the user's Google Account storage at original quality. 200 MB max per photo, 20 GB max per video. unit GB; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | Media bytes served via baseUrl, 75,000 requests/project/day; no documented per-GB charge but the underlying storage is metered. |
| Identity | institution derived from the record's prose, not from an identity field, and no signup was executed: nal quality. 200 MB max per photo, 20 GB max per video. API uploads skip the 'Storage saver' free-compression path, so they consume quota fastest of any Google surface. Strictly |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | rest, rclone rest |
| Tooling | Photos Library API (photoslibrary.appendonly scope), uploads endpoint + batchCreate; no rclone backend. Unofficial wrappers are thin. |
| Composes with | 20 GB per video x N videos is the only interesting shape, but it does NOT escape the 15 GB account quota — this is billing-avoidance-free. Not a multiplier, a quota consumption path. |
| Reliability | API uploads skip the 'Storage saver' free-compression path, so they consume quota fastest of any Google surface. Strictly worse than Drive for bulk data. |
| Notes | No quota of its own — but every item uploaded counts against the user's Google Account storage at original quality. 200 MB max per photo, 20 GB max per video. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | developers.google.com/photos/library/guides/upload-media | 'All uploads are stored at original quality, counting towards the user's Google Account storage if they exceed 25MB per user' + max 200 MB photo / 20 GB video |
| 2026-10-21 | developers.google.com/photos/overview/api-limits-quotas | 'The quota limit for requests to the Library API is 10,000 requests per project per day'; media bytes 75,000/day |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "No quota of its own — but every item uploaded counts against the user's Google Account storage at original quality. 200 MB max per photo, 20 GB max per video.",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (6)
$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
Google Play Books (personal upload library) Storage
fails: quotainstitutionno cardRper-object-cap
| Free tier | Up to 1,000 uploaded files per Google account (2 GB max per file per the Partner Center spec; the consumer uploader historically enforced 100 MB). 1,000 x 2 GB ≈ 2 TB theoretical per Google account. unit MB; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | 2 GB max per file (2000000000 bytes) |
| Retention | not stated |
| Egress | No egress metering for reading, but there is no public direct-download URL: files are served through the authenticated Play Books reader. Not CDN-cacheable, not shareable as a link. |
| Identity | institution derived from the record's prose, not from an identity field, and no signup was executed: Play Books; it is only blocked by the absence of an upload API. 1,000-file cap per account; 2 GB/file (Partner Center) vs 100 MB in the consumer uploader; EPUB/PDF only; uploads are |
| Limits |
1,000-file cap per account; 2 GB/file (Partner Center) vs 100 MB in the consumer uploader; EPUB/PDF only; uploads are re-processed into Google's internal format, so the stored artifact is not byte-identical to what you sent. |
| Binding | per-object-cap the per-object cap sets the shard size and therefore the object count; it bounds the total more tightly than any headline quota |
| Interface | unknown wrapper-only |
| Tooling | Web upload only (play.google.com/books -> Upload files). No public API; third-party tools are downloaders for purchased books, not uploaders. |
| Composes with | 1,000 files x 2 GB = ~2 TB per free Google account, x unlimited free Google accounts. This is the Drive-service-account pattern applied to Play Books; it is only blocked by the absence of an upload API. |
| Reliability | Files are converted on ingest and served only inside the reader; no bulk export, no public URLs, and there is no supported automation path. |
| Notes | Up to 1,000 uploaded files per Google account (2 GB max per file per the Partner Center spec; the consumer uploader historically enforced 100 MB). 1,000 x 2 GB ≈ 2 TB theoretical per Google account. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
| Parsed numbers |
|
Evidence (4)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | support.google.com/books/partner/answer/9261355?hl=en | 'Maximum file size 2 GB per file' |
| 2026-10-21 | support.google.com/books/partner/answer/3424254?hl=en | 'Each file must be less than 2 GB'; EPUB/PDF only |
| 2026-10-21 | support.google.com/googleplay/answer/11012086?hl=en | consumer upload flow is web-only ('go to your Play Books library ... Upload files') |
| 2026-10-21 | webapps.stackexchange.com/questions/92325/seeing-the-count-o | 1,000-book cap report; corroborated by https://ebookfriendly.com/how-to-upload-own-ebooks-to-google-play/ ('up to 1,000 own files') |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "Up to 1,000 uploaded files per Google account (2 GB max per file per the Partner Center spec; the consumer uploader historically enforced 100 MB). 1,000 x 2 GB ≈ 2 TB theoretical per Google account.",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (6)
$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.interfaceinterface-unknown$.per_object_cap.notesnot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
google sites (new sites) as free hosting/storage Storage
fails: quotaemail?card?R
| Free tier | Sites and all embedded content count toward the owner's Drive storage — it is not an exempt surface. No independent quota exists. unit none; metered on none |
|---|---|
| Price | none stated |
| Per-object cap | not recorded |
| Retention | not recorded |
| Egress | not recorded |
| Identity | email (placeholder) not published |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | unknown |
| Tooling | not recorded |
| Composes with | not recorded |
| Reliability | Sites and all embedded content count toward the owner's Drive storage — it is not an exempt surface. No independent quota exists. |
| Notes | Sites and all embedded content count toward the owner's Drive storage — it is not an exempt surface. No independent quota exists. |
| Evidence rung | reported |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | knowledge.workspace.google.com/admin/sites/google-sites-requ | Sites and all embedded content count toward the owner's Drive storage — it is not an exempt surface. No independent quota exists. |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "Sites and all embedded content count toward the owner's Drive storage — it is not an exempt surface. No independent quota exists.",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (15)
$.chainingnot-recorded$.confidence_notewithheld$.egress.termsnot-recorded$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_noteprocess$.per_object_cap.notesnot-recorded$.per_object_cap.statednot-recorded$.retention.notesprocess$.retention.statednot-recorded$.toolingnot-recorded
Google Vertex AI Model Garden — custom model import Storage
fails: quotacardcardR
| Free tier | No hosting quota because there is no hosting: 'You must store your model artifacts in a Cloud Storage bucket, where the bucket's region matches the locational endpoint you're using.' Capacity is your GCS bill. unit mixed; metered on stored-bytes |
|---|---|
| Price | none stated |
| Per-object cap | not stated |
| Retention | not stated |
| Egress | Paid GCS egress. |
| Identity | card card |
| Binding | none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit |
| Interface | rest rest |
| Tooling | aiplatform.Model.upload(...) from the Python SDK — clean, but it uploads a pointer plus artifacts already in GCS. |
| Composes with | None for storage. GCS free tier is 5 GB, which the brief already records as a dead end. |
| Reliability | Not a storage provider. Console requires a Google Cloud billing account. |
| Notes | No hosting quota because there is no hosting: 'You must store your model artifacts in a Cloud Storage bucket, where the bucket's region matches the locational endpoint you're using.' Capacity is your GCS bill. |
| Evidence rung | reported One page read moves it to documented. |
| Checked | 2026-10-21 (inherited); stale on 2026-11-04 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | docs.cloud.google.com/vertex-ai/docs/model-registry/import-m | opened: 'You must store your model artifacts in a Cloud Storage bucket'. |
| 2026-10-21 | console.cloud.google.com/vertex-ai/publishers | PROBED: returns a 'Sign in' page (41 words, no content) rather than the catalogue. Confirms the gate is a Google identity plus billing. |
More fields (1)
{
"dead_end": {
"failure_class": "quota",
"why": "No hosting quota because there is no hosting: 'You must store your model artifacts in a Cloud Storage bucket, where the bucket's region matches the locational endpoint you're using.' Capacity is your GCS bill.",
"retry_when": null,
"first_recorded": "2026-10-21"
}
}Unknown or withheld values (6)
$.extras.dead_end.retry_whenprocess$.free_tier.notesprocess$.per_object_cap.notesnot-extracted$.per_object_cap.statednot-extracted$.retention.notesnot-extracted$.retention.statednot-extracted
Firebase Studio Compute
deademailno cardphoneD
| Free tier | Quoted: "Firebase Studio is sunsetting on March 22, 2027. As of June 22, 2026, new workspace creation and user signup are disabled." The plan was 3 workspaces per user at no cost. unit none; metered on wall-clock; none |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | not stated |
| Identity | email no signup is possible at all: 'As of June 22, 2026, new workspace creation and user signup are disabled', so there is no requirement left to meet and the 3-workspace quota is closed to new accounts. What the pre-closure signup asked for - email, phone, card or ID - is not described in any evidence row: unknown. No hold is stated. |
| Binding | none New accounts cannot create a workspace at all since 2026-06-22, so the 3-free-workspace quota is closed; existing workspaces keep running until 2027-03-22. |
| Interface | unknown inferred from the tooling the provider documents; not probed |
| Tooling | not stated |
| Composes with | Nothing new. The migration targets named by Google, AI Studio and Antigravity, were not evaluated and are the obvious next check for an always-on free development environment from Google. |
| Reliability | no reliability event recorded |
| Notes | This is the most commonly recommended free cloud IDE that no longer accepts users, and listicles written before June 2026 still describe three free Nix-based workspaces. Recorded as dead with the two dates, because a reader needs to know both that it is closing and that it is already closed to them. Apps already deployed to Firebase keep running after the sunset, which is a different product. |
| Evidence rung | documented |
| Checked | 2026-10-22 (live); stale on 2026-11-21 |
Evidence (3)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | firebase.google.com/docs/studio/pricing | sunset on 2027-03-22, new workspace creation and signup disabled from 2026-06-22, and 3 workspaces per user at no cost, quoted |
| 2026-10-22 | firebase.google.com/docs/studio/troubleshooting | the same notice in the FAQ, plus the statement that the no-cost plan is limited to three workspaces per user |
| 2026-10-22 | firebase.google.com/docs/studio/migrating-project | existing workspaces can be worked in until the sunset date and must be migrated to Google AI Studio or Google Antigravity |
More fields (1)
{
"wave2": {
"class": "inverse-dead",
"free_tier_kind": "none",
"shape": null
}
}Unknown or withheld values (9)
$.egress.meteredegress-unknown$.egress.termsnot-stated$.interfaceinterface-unknown$.per_object_cap.notesprocess$.per_object_cap.statednot-applicable$.reliability_notenone-observed$.retention.notesprocess$.retention.statednot-applicable$.toolingnot-stated
Google Programmable Search / Custom Search JSON API Crawling
deadcardcardDrequest-rate
| Free tier | 100 queries a day for existing customers only unit requests; metered on request-count |
|---|---|
| Price | none stated |
| Per-object cap | not applicable: the unit is a query |
| Retention | the service is discontinued on 2027-01-01 |
| Egress | not applicable: JSON answers |
| Identity | card billing had to be enabled past 100 queries a day |
| Limits | |
| Binding | request-rate |
| Interface | rest REST with a key and a search engine id |
| Tooling | not stated |
| Composes with | not established |
| Reliability | no reliability event recorded |
| Notes | Two closures in one page: closed to new customers now, discontinued in 2027. The free tier is real and unreachable, which is the fails-identity shape in a vendor that never asks for an identity. |
| Evidence rung | documented |
| Checked | 2026-10-22 (live); stale on 2026-11-21 |
Evidence (1)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | developers.google.com/custom-search/v1/overview | the Custom Search JSON API is closed to new customers; the pricing applies only to existing customers until the service discontinuation on 2027-01-01; 100 queries a day free, USD 5 per 1,000 beyond |
Unknown or withheld values (2)
$.reliability_notenone-observed$.toolingnot-stated
LunaProxy, PIA S5 Proxy and 911 S5 (911.re) IP and proxies
deadanonymousno cardD
| Free tier | No live offer. lunaproxy.com and piaproxy.com return NXDOMAIN, and 911.re serves a seizure notice in place of a product page, all read 2026-10-24. Nothing was priced because nothing answers. unit none; metered on none; The survey named all three as live candidates. None of the three has a page left to price, so there is no tier to compare and no free allowance to state. |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | not applicable: an address product, not stored bytes |
| Egress | unstated: no live page carries a traffic position. lunaproxy.com and piaproxy.com return NXDOMAIN and 911.re answers with a seizure notice, so whatever these three said about traffic out was never read and is not recorded here. free_allowance: none published |
| Identity | anonymous n/a: no registration path exists, so no identity gate was reached. Nothing was signed up for and no account was created. |
| Binding | none Nothing in the vocabulary fails at 45 GB a month because nothing is sold: two names do not resolve and the third serves a seizure notice, so no rate, cap, quota or shaping exists to hit. What removes these vendors is existence, which the seven-value vocabulary has no kind for and the verdict already names. |
| Interface | unknown Nothing is reachable: two of the three names have no DNS record and the third serves a seizure notice. The protocol these vendors sold was SOCKS5 and HTTP(S) per IP, which is what the names said (PIA S5, 911 S5) and what the one surviving stored copy of pia-proxy.com advertises ('Pay per IP over HTTP(S) and SOCKS5'); that describes the product, not a path a client can use today. |
| Tooling | Nothing to reach. A missing wrapper is not the obstacle: there is no service behind the name to wrap, so no client, SDK or protocol choice changes the outcome. |
| Composes with | Nothing chains. A vendor whose names do not resolve cannot carry the gateway's 45 GB a month, and the record's value is negative: it removes three names from the survey's candidate list so the next session does not spend a wave on them. |
| Reliability | Two of the three names are gone at the DNS level and the third serves a seizure banner; the notice on 911.re reads 'This domain has been seized by the Defense Criminal Investigative Service, the Federal Bureau of Investigation, and the Bureau of Industry and Security, Office of Export Enforcement' as part of a coordinated action 'taken against the 911 S5 residential proxy service'. |
| Notes | Three candidates for why the names stopped answering, before any of them was tested. (1) The brands were shut with the rest of IPIDEA's marketing estate in the January 2026 action. (2) The domains lapsed on their own and the brands moved to new names. (3) A resolver-side filter returned NXDOMAIN falsely, which is a real failure mode here because the host resolver is a filtering one and it also reported lunaproxy.com as non-existent. Refuted for (3) by asking Cloudflare's resolver directly, which returned Status 3 for lunaproxy.com and piaproxy.com; a filtered answer would have come from the filter, not from the .com zone's authority. THE FACTS THIS RECORD CORRECTS: the survey sentence says pia-proxy.com returns NXDOMAIN. It does not. It resolves behind Cloudflare and returns 522, with no origin; The likely cause is that the two names were read as one. UNVERIFIED: the survey also records two lookalike 'continuation' sites carrying identical copy. Two PIA-adjacent names do resolve, pia-proxy.com and piaproxy.io, and both sit behind Cloudflare with no answering origin, so neither served a page to compare. piaproxy.net is a different company entirely, Private Internet Access, and pias5.com, pia-s5.com and piaproxy.co have no A record. So the pair may be exactly those two 522 hosts, or something else, and this record does not say. What would settle it: a stored snapshot for piaproxy.io to compare against the pia-proxy.com copy, or a registration-date read for both names. SINGLE SAMPLE: every DNS check here is one query per name from one resolver on one day, and the 522s were single fetches; a second resolver and a second day would turn a one-day observation into a trend, which for a domain is worth little either way. No record in another topic names these three as live, so there is no cross-topic correction to file; the correction is to the survey sentence itself. What the record exists to prevent is the thing the note in the assignment names: a dead vendor with no record gets re-attempted by whoever has the idea next, and re-attempting these three costs a search, a fetch and a decision before it reaches the same wall. |
| Evidence rung | documented The death itself is measured: Cloudflare's resolver returned NXDOMAIN for lunaproxy.com and piaproxy.com, and 911.re served its seizure banner, all. Nothing would raise the death above measured; what would raise the attribution is a registry record of the deletion or a second independent account of the same action. |
| Checked | 2026-10-25 (live); stale on 2026-11-24 |
Evidence (6)
| date | source | what it proved |
|---|---|---|
| 2026-10-24 | cloud.google.com/blog/topics/threat-intelligence/disrupting- | not published |
| 2026-10-24 | probe:lunaproxy.com | Cloudflare DNS-over-HTTPS answered Status 3 (NXDOMAIN) with no A record, so the name has been removed from the .com zone rather than merely failing to answer. |
| 2026-10-24 | probe:piaproxy.com | Cloudflare DNS-over-HTTPS answered Status 3 (NXDOMAIN) with no A record. |
| 2026-10-24 | probe:pia-proxy.com | Not NXDOMAIN: resolves to 104.21.6.121 and 172.67.134.213 behind Cloudflare, and a fetch returned HTTP 522, meaning the edge has no origin behind it. The only readable text under this name was a stored third-party snapshot, titled 'PIA S5 Proxy — Residential SOCKS5 IPs, 350M+', which is a copy of the product page rather than a live page. |
| 2026-10-24 | probe:piaproxy.io | Resolves to 162.255.119.134 and returned HTTP 522; two read attempts, one at a 25 second deadline and one at 60, produced no page text, so no copy could be compared against pia-proxy.com's. |
| 2026-10-24 | 911.re/ | HTTP 200 with a page titled 'This Website Has Been Seized', whose body names the seizing agencies and states the action was 'taken against the 911 S5 residential proxy service'. |
More fields (3)
{
"rung": 1,
"rung_note": "Session, and only as a floor: what these three sold was a session-bound exit address, and nothing above rung 1 was ever established for them here, since no page was served to measure an account, fingerprint or behaviour control against.",
"interface_enum_note": "The schema's interface enum has no token for a proxy endpoint or a tunnel, so this record follows the convention the egress records use: a proxy client or tunnel is 'cli', a vendor API is 'rest'. Previous value: ['none']"
}Unknown or withheld values (3)
$.evidence[0].provedwithheld$.interfaceinterface-unknown$.per_object_cap.statednot-applicable