singletenant.ai

Changelog

Every change to the vendor database, newest first. Data is reviewed manually; there is no automated collection. Corrections: simon@singletenant.ai.

  1. Updated

    Nebius data residency upgraded from Stated to Contractual: Data Processing Agreement section 6.1 now commits Token Factory dedicated endpoints to process Customer Content in the region the customer selects (public model endpoints follow the processing location shown in the API). Verified 2026-10-22.

    Nebius

  2. Updated

    Together AI data residency moves from Not verified to Stated: the dedicated endpoints page says customers choose the deployment region for data residency. Its terms still make no geographic commitment, and no regions are named, so regions stay not verified. Verified 2026-10-22.

    Together AI

  3. Updated

    Schema change: category is now functional only; geography moves to parent jurisdiction (ISO codes, Civo UK to GB), regions (country/bloc codes) and a new sourced Sovereign for field. Recategorised: Civo, Nebius, Scaleway to Dedicated GPU; OVHcloud to Managed self-host. Sovereign for: Civo GB; OVHcloud, Scaleway EU; Nebius none (no claim found). Verified 2026-10-22.

    Civo · Nebius · OVHcloud · Scaleway

  4. Updated

    Baseten egress restrictions added to the analyst note: deployments can limit outbound traffic to allow-listed IP ranges and domains via egress_restrictions in Truss, enabled by support (changelog, 12 October 2026). Sources: https://www.baseten.co/resources/changelog/egress-restrictions/ and https://docs.baseten.co/development/model/egress-restrictions. Verified 2026-10-19.

    Baseten

  5. Updated

    Provisioned Throughput commitment terms updated: orders can now change scope and increase term (28 September 2026) and support multiple pending orders and change requests (10 September 2026). Source: https://docs.cloud.google.com/gemini-enterprise-agent-platform/release-notes. Verified 2026-10-19.

    Vertex AI Provisioned Throughput

  6. Updated

    Scaleway analyst note now records that the serverless Generative APIs tier prices cached tokens separately. Source: https://www.scaleway.com/en/generative-apis/ (announced 13 October 2026: https://www.scaleway.com/fr/news/scaleway-accelere-pour-donner-a-leurope-les-moyens-de-ses-ambitions-dans-lia/). Verified 2026-10-19.

    Scaleway

  7. Updated

    This week's vendor changes: AWS Provisioned Throughput now in London (eu-west-2); Azure Data Zone adds APAC; Modal adds Dedicated Endpoints; Nebius offers a HIPAA BAA; Baseten no longer lists UK or Australia regions. Renamed: Vertex AI to Gemini Enterprise Agent Platform, Scaleway Managed Inference to Generative APIs Dedicated Deployment, Databricks drops Mosaic AI. Verified 2026-10-17.

    Amazon Bedrock Provisioned Throughput · Azure OpenAI Provisioned Throughput (PTU) · Modal · Nebius · Baseten · Vertex AI Provisioned Throughput · Scaleway · Databricks Model Serving

  8. Updated

    Every certification and residency fact regraded on one evidence rule: Contractual (contract terms, or a certificate, audit report or certifying body), Stated (the vendor's own pages) or Not verified. Residency is now an evidence level rather than yes or no. The rule is on the methodology page. Verified 2026-10-17.

    Amazon Bedrock Provisioned Throughput · Azure OpenAI Provisioned Throughput (PTU) · Baseten · Civo · CoreWeave · Databricks Model Serving · Fireworks AI · Hugging Face Inference Endpoints · Modal · Nebius · NVIDIA NIM · OVHcloud · RunPod · Scaleway · Together AI · Vertex AI Provisioned Throughput

  9. Updated

    Residency regraded: Vertex AI moves from not verified to Contractual (Google's Service Specific Terms commit to a chosen multi-region); OVHcloud moves to Contractual (Public Cloud terms let the customer choose the datacentre region); Together AI's "no guarantee" becomes not verified. Verified 2026-10-17.

    Vertex AI Provisioned Throughput · OVHcloud · Together AI

  10. Removed

    HIPAA BAA removed for Fireworks AI, Modal and RunPod. Each states a HIPAA programme, but none of their docs, trust centres or terms names a BAA. Verified 2026-10-17.

    Fireworks AI · Modal · RunPod

  11. Removed

    SecNumCloud removed from OVHcloud. The ANSSI qualification covers named private-cloud and bare-metal products, not AI Endpoints or AI Deploy. Verified 2026-10-17.

    OVHcloud

  12. Added

    ISO 27001 added for Together AI: certified by A-LIGN, an ANAB-accredited certification body. Verified 2026-10-17.

    Together AI

  13. Corrected

    Four articles corrected for the evidence levels: the jurisdiction walk, Why this site says Not verified so often, the UK buyer's position, and Sovereignty washing. Civo's UK residency is Stated, not contractual; Scaleway, OVHcloud and Vertex AI are Contractual. Corrected 2026-10-17.

    Scaleway · OVHcloud · Nebius · Civo · Amazon Bedrock Provisioned Throughput · Azure OpenAI Provisioned Throughput (PTU) · Modal · Vertex AI Provisioned Throughput

  14. Updated

    Every field of every profile re-checked against the vendor's own pages: category, summary, deployment models, regions, certifications, pricing, residency, parent jurisdiction and analyst note. Each profile now lists the source and check date for each field. Verified 2026-10-17.

    Amazon Bedrock Provisioned Throughput · Azure OpenAI Provisioned Throughput (PTU) · Baseten · Civo · CoreWeave · Databricks Model Serving · Fireworks AI · Hugging Face Inference Endpoints · Modal · Nebius · NVIDIA NIM · OVHcloud · RunPod · Scaleway · Together AI · Vertex AI Provisioned Throughput

  15. Updated

    UK added to Amazon Bedrock Provisioned Throughput regions. The Provisioned Throughput model table lists eu-west-2, which AWS names Europe (London). Verified 2026-10-17.

    Amazon Bedrock Provisioned Throughput

  16. Updated

    Asia-Pacific added to Azure OpenAI Provisioned Throughput regions. Data Zone deployments now cover a US, EU or APAC zone. Verified 2026-10-17.

    Azure OpenAI Provisioned Throughput (PTU)

  17. Added

    Dedicated Endpoints added to Modal's deployment models: models on isolated, autoscaling containers at standard compute rates. Verified 2026-10-17.

    Modal

  18. Added

    HIPAA BAA added for Nebius. Its HIPAA guideline makes PHI use subject to a signed Business Associate Agreement; the profile previously said no BAA was found. Verified 2026-10-17.

    Nebius

  19. Corrected

    Nebius pricing corrected: the dedicated-endpoint billing policy makes an endpoint billable while at least one replica runs, though the product page still cites per-token pricing. Summary no longer says per-token for dedicated endpoints. Corrected 2026-10-17.

    Nebius

  20. Corrected

    Analyst notes corrected to current vendor wording: Hugging Face now offers serverless (not dedicated) NIM endpoints; OVHcloud AI Endpoints lists more than 20 models, not 40+; RunPod Secure Cloud is T3/T4 data centres; Databricks names the product Model Serving. Corrected 2026-10-17.

    NVIDIA NIM · OVHcloud · RunPod · Databricks Model Serving

  21. Updated

    All 16 vendor profiles re-verified against primary sources. Each source was re-read for the claim, not just checked for a response. Moved pages were repointed for CoreWeave, OVHcloud, Scaleway, Civo, Nebius, Together AI and Vertex AI. The facts that changed have their own entries. Verified 2026-10-17.

    Amazon Bedrock Provisioned Throughput · Azure OpenAI Provisioned Throughput (PTU) · Baseten · Civo · CoreWeave · Databricks Model Serving · Fireworks AI · Hugging Face Inference Endpoints · Modal · Nebius · NVIDIA NIM · OVHcloud · RunPod · Scaleway · Together AI · Vertex AI Provisioned Throughput

  22. Added

    ISO 27001 added for Baseten and RunPod. Both trust centres now list ISO/IEC 27001:2022, and Baseten publishes the certificate. Verified 2026-10-17.

    Baseten · RunPod

  23. Updated

    Baseten regions narrowed to US and Europe. Its regional-deployments doc now lists only us and eu as selectable regions; the UK and Australia regions its docs named in July are no longer listed. Verified 2026-10-17.

    Baseten

  24. Updated

    UK added to Databricks Model Serving regions. The Model Serving region table lists EU (London) with GPU serving and provisioned throughput. Verified 2026-10-17.

    Databricks Model Serving

  25. Added

    Civo added as the first uk-sovereign vendor. UK-registered (Companies House 09568551, UK-resident sole PSC, no foreign parent); single-tenant dedicated NVIDIA GPU via VFIO passthrough; UK Sovereign Cloud states data stays in the UK under UK law. ISO 27001 and SOC 2. Verified 2026-07-18.

    Civo

  26. Corrected

    Jurisdiction analysis article re-walked for the v2.3 data: Civo added as the first UK-sovereign entry (UK parent and UK residency guarantee), the walk retitled to sixteen, and the closing note now points at both the EU sovereign and UK sovereign category filters. Corrected 2026-07-18.

    Civo

  27. Updated

    UK added to Modal's regions. Modal's region-selection doc lists "uk" (United Kingdom) as a broad region distinct from "eu" (EEA); it was omitted before the UK bucket existed. Checked 2026-07-18.

    Modal

  28. Updated

    UK added to CoreWeave's regions. CoreWeave's newsroom confirms two UK data centres, Crawley and London Docklands, operational since 2024 and hosting its NVIDIA GPU platform. Checked 2026-07-18.

    CoreWeave

  29. Updated

    UK added to Baseten's regions. Its regional-environments documentation names "US, EU, UK, and Australia" as available regions; with the new UK bucket, UK is recorded separately rather than omitted. Source re-checked 2026-07-18.

    Baseten

  30. Updated

    Regions now use five defined buckets (US, UK, Europe, Asia-Pacific, Global), documented on the methodology page. The Europe bucket was narrowed to the EU/EEA excluding the UK; UK placement is tracked separately as its own bucket and is being added per vendor where a primary source names a UK location. See methodology for the definitions.

  31. Updated

    CoreWeave profile note now cites the primary source for the ISO 27001 exclusion: its security page aligns programs "to industry-standard SOC2 and ISO 27001 frameworks" rather than claiming certification, so ISO 27001 stays off the profile. Updated 2026-07-13.

    CoreWeave

  32. Corrected

    Jurisdiction analysis article updated to match verified region data: Baseten and Vertex AI move into the US-parented-with-EU-region walk now that their regions are verified; NVIDIA NIM and Together AI remain not verified. Corrected 2026-07-13.

    Baseten · Vertex AI Provisioned Throughput · NVIDIA NIM · Together AI

  33. Updated

    Baseten regions populated (US, Europe, Asia-Pacific) from the regional-environments docs, which name US, EU, UK and Australia as configurable in-region routing targets for data residency. Verified 2026-07-13.

    Baseten

  34. Updated

    Vertex AI Provisioned Throughput regions populated (US, Europe, Asia-Pacific) from the Vertex AI locations page, which lists supported model regions across all three; PT runs wherever its models are available. Verified 2026-07-13.

    Vertex AI Provisioned Throughput

  35. Updated

    Databricks Model Serving regions populated (US, Europe, Asia-Pacific) from the Model Serving limits documentation, which lists supported regions per cloud. Verified 2026-07-12.

    Databricks Model Serving

  36. Updated

    Hugging Face Inference Endpoints regions populated (US, Europe) from the FAQ, which names AWS us-east-1 and eu-west-1, Azure eastus and GCP us-east4. No Asia-Pacific region documented. Verified 2026-07-12.

    Hugging Face Inference Endpoints

  37. Updated

    Modal regions populated (US, Europe, Asia-Pacific) from the region-selection documentation, which supports region pinning across these continents. Verified 2026-07-12.

    Modal

  38. Updated

    RunPod regions populated (US, Europe, Asia-Pacific) from the Global Networking data-centre list. Verified 2026-07-12. Coarse continent buckets; RunPod runs further data centres beyond that set.

    RunPod

  39. Corrected

    Databricks SOC 2 Type II evidence link corrected. It previously pointed to the HIPAA compliance page; it now points to the Databricks SOC compliance page, which names the SOC 2 Type II report. Found during the post-launch integrity audit.

    Databricks Model Serving

  40. Updated

    RunPod analyst note updated: the affiliate relationship is now documented on the site's disclosure page, published today alongside the privacy policy, terms of use, and methodology pages. No referral link is live yet; when one goes live it will be marked on the page it appears.

    RunPod

  41. Added

    Hyperscaler dedicated tier added: Amazon Bedrock Provisioned Throughput, Azure OpenAI provisioned (PTU), and Vertex AI Provisioned Throughput. Compliance certifications checked per service against each provider's published services-in-scope pages; facts that could not be verified per-service are recorded as not verified rather than assumed.

    Amazon Bedrock Provisioned Throughput · Azure OpenAI Provisioned Throughput (PTU) · Vertex AI Provisioned Throughput

  42. Added

    Managed self-host tier completed: NVIDIA NIM and Databricks Model Serving. All facts from vendor documentation, trust pages, published agreements, and pricing pages fetched at review.

    NVIDIA NIM · Databricks Model Serving

  43. Added

    EU sovereign tier added (OVHcloud, Scaleway, Nebius) along with Hugging Face Inference Endpoints. Facts verified against primary sources on 2026-07-07; data-residency guarantees recorded as not verified where no explicit product-level commitment was found.

    Hugging Face Inference Endpoints · Nebius · OVHcloud · Scaleway

  44. Added

    Fireworks AI and CoreWeave added to the dedicated GPU tier, researched from vendor docs, trust centres, and pricing pages.

    Fireworks AI · CoreWeave

  45. Updated

    All four launch vendors migrated to schema v2: every populated fact now carries a primary-source link and the date it was last verified. Facts re-checked against vendor trust centres, pricing pages, and terms; claims that could not be re-verified were removed or marked not verified.

    Baseten · Modal · RunPod · Together AI