Analysis

OpenAI API data residency is now self-serve for eligible organisations: the UK stores only, and newer models cost 10% more

1AYM is an OpenAI Select Partner.

Published . Vendor facts checked on against each vendor’s own pages, listed at the end. Plans, prices and availability change. Neither OpenAI nor Anthropic has reviewed this analysis.

The short version

OpenAI's release note dated 18 November 2026 lets the owner pick a data residency region when creating an API project. Its data residency guide and pricing page, read on 20 November, set out the regions, the price and the limits.

What changed and how to switch it on
From 18 November 2026, eligible API organisation owners pick a region when they create a project. Models released on or after 5 March 2026 cost 10% more on a data residency endpoint.
Regions: storage and processing
Six of the ten regions store and process data, three of them for gpt-6-luna only, and the UAE is reached through sales. The UK, Canada, Singapore and South Korea store only.
What it does not cover
System data, such as account and billing data, can be stored and processed outside the region. Data sent to third-party services such as remote MCP servers follows their own rules, and some caching and endpoint features have limits.

What changed in OpenAI API data residency, and how an owner switches it on

What was announced

OpenAI's release notes for ChatGPT Enterprise and Edu carry an entry dated 18 November 2026 about the API. It says owners of eligible API organisations can now choose regional data residency when they create a project. OpenAI's guide describes data residency as a project setting that lets you choose where the infrastructure OpenAI uses to serve you sits.

An owner can pick one fixed region, or create a new Global project with per-request data residency switched on. The release note says this set-up doesn't need a separate Zero Data Retention application.

The region picker shows whether each option supports storage only, or storage and processing. Storage is where content is kept. Processing is where inference runs, which is the model working out its answer to a request.

OpenAI's guide says eligible organisations turn residency on when they create a project in the API Platform. To become eligible, the guide says to contact sales.

An organisation owner opens Projects, selects Create, names the project and picks a region under Data and inference residency. The owner then creates an API key for the project, the secret that software sends with each request. That key is used with the region's own web address, such as eu.api.openai.com.

OpenAI's guide says a project's residency region can't be changed after it is created. To use a different region, you create a new project.

An endpoint is one web address in the API that does one job. OpenAI's pricing page lists a 10% uplift on regional processing (data residency) endpoints for models released on or after 5 March 2026. The guide puts the same 10% on data residency endpoints.

What it means for enterprises

By our reading, once an organisation is eligible, its owners can set up a regional project without another sales step. Whoever creates the project also picks its region, so a developer who holds the owner role can make that choice alone.

A token is the small unit of text the API bills by. GPT-6.2 Sol lists at $2 per million input tokens and $10 per million output tokens on the Standard tier. By our arithmetic, the uplift takes that to $2.20 and $11.

Take, as an illustration of ours, a workload of 10 million input and 2 million output tokens a day on Sol. It costs $40 a day at list price, and $44 on a residency endpoint.

Because the region is fixed, our reading is that moving a live workload means a new project, new keys and a new web address. Anything stored in the old project, such as files or batch jobs, would have to be created again.

Before you switch it on

Who picks the region
Decide who approves the region for each new API project before owners start creating them. Once the organisation is eligible, the choice sits in the project set-up screen.
Eligibility
Ask OpenAI sales whether your organisation is eligible, which is the only route the guide gives.
Budget the 10%
Add 10% to the forecast for each workload on a model released on or after 5 March 2026 that moves to a residency endpoint.
Region first
Pick the region before the project holds any data, because the guide says it can't be changed afterwards.

Sources: [1] [2] [3]

Which OpenAI API regions store data, and which also process it?

What was announced

The guide's residency promise covers Customer Content, a term defined in OpenAI's services agreement. Storage means that content is kept at rest in the chosen region whenever an endpoint needs to keep it, as batch jobs do. Processing means inference on it also runs there.

The six regions that offer storage and processing are the United States, Europe (the European Economic Area and Switzerland), Australia, Japan, India and the United Arab Emirates. In Australia, Japan and India, only gpt-6-luna is processed in the region.

The United Kingdom, Canada, Singapore and South Korea store data and offer no regional processing. The guide says that where a region lacks regional processing, OpenAI may process and temporarily store Customer Content outside it to deliver the service.

OpenAI says to contact sales for access to the UAE region, at ae.api.openai.com. The guide lists three models for processing there on Chat Completions, and the view we read named two of them, gpt-5.6-luna and gpt-5.5-2026-04-23.

Image support outside the US needs extra approval, and the guide says to contact sales for it.

ChatGPT has its own residency settings, separate from the API. There too the UK is a storage region, and inference residency covers Europe, the US and the UAE. OpenAI includes data residency at no extra cost on ChatGPT Enterprise and Education.

What it means for enterprises

A UK buyer who needs the model to run in the UK can't get that from the gb. region today. The gb. region keeps stored content in the UK, and the guide says OpenAI may process that content outside the region.

An EU project processes in the European Economic Area and Switzerland. By our reading, the choice between UK storage and EU processing is one for the data protection officer and the legal team to make.

Gulf buyers have a UAE region with processing, reached through sales. The two models named in the view we read come from the GPT-5 line, while GPT-6.2 Sol has US and EU residency.

In Australia, Japan and India, a team that needs local processing is limited to GPT-6 Luna on Standard pay-as-you-go, without hosted tools.

Before you switch it on

Storage or processing
For each workload, write down whether it needs data stored in a region, processed there, or both.
UK workloads
If a UK workload must be processed in the UK, note that gb. stores only. Take the choice between gb. and eu. to your data protection officer.
Model by region
Confirm that the model you plan to use is processed in your region, because Australia, Japan and India process gpt-6-luna only.
UAE access
Gulf teams can ask OpenAI sales for the UAE region, and should confirm which models they will get there.

Sources: [1] [2] [4]

What OpenAI API data residency does not cover

What was announced

OpenAI's guide says system data may be stored and processed outside the region. It defines system data as account data, metadata and usage data without Customer Content.

The guide's examples include account information and the profiles of end users who use the services directly. Analytics, usage statistics, billing information, support requests and the schemas used for structured outputs count too.

Residency also doesn't cover data that leaves the region because of where an end user or your own systems are. Nor does it cover products, services or content from parties other than OpenAI.

The Remote MCP server tool connects a model to a server that another party runs. OpenAI says MCP servers are third-party services, and data sent to one follows that server's own residency policies.

On Chat Completions, the store=true setting can't be used outside the US. Extended prompt caching reuses the repeated part of a prompt. In a storage-only region, it may need OpenAI to process and temporarily store content outside the region.

On the Realtime endpoint, OpenAI says tracing is not yet EU data residency compliant.

The Decisions API is offered in every supported region, with regional processing in the US and Europe only. OpenAI notes that being offered in a region doesn't mean inference runs there.

What it means for enterprises

Account, billing and usage data from a residency project can still be kept outside the region. A data protection impact assessment, the record of a project's privacy risks, can record that.

A remote MCP server or another party's service called in a request can take data out of the region, because it follows its own rules.

The guide says extended prompt caching in a storage-only region may send content outside it. By our reading, projects in the UK, Canada, Singapore or South Korea that use it are the ones to check first.

Before you switch it on

System data
Record in your data protection impact assessment and records of processing that account, usage and billing data can leave the region.
MCP and other tools
List every MCP server and outside tool each workload calls, and check where each one keeps data.
store=true
Search Chat Completions code for store=true before you move it to a project outside the US.
Caching
Check whether extended prompt caching is on for projects in a storage-only region.
Realtime tracing
If you use Realtime in an EU project, check whether tracing is switched on.

Sources: [2]

Where the market is heading: OpenAI API data residency region by region

OpenAI's release note puts the choice of region in the project set-up screen, once an organisation is eligible. By our reading, the sales steps left are eligibility, UAE access and image support outside the US.

On ChatGPT, OpenAI includes data residency at no extra cost on Enterprise and Education. On the API, models released on or after 5 March 2026 cost 10% more on a residency endpoint.

Six of the ten API regions offer processing, and three of those six process only gpt-6-luna. By our reading, buyers outside the US and Europe who need local processing have a narrower choice of models today.

OpenAI API data residency by region, from OpenAI's data residency guide and API pricing page (checked 20 November 2026). Plans and prices change.
DimensionWeb addressStorageProcessingNotes
United Statesus.api.openai.comYesYesText, audio, voice and image
Europe (EEA and Switzerland)eu.api.openai.comYesYesImage needs extra approval from sales
Australiaau.api.openai.comYesgpt-6-luna onlyStandard pay-as-you-go; no hosted tools in region
Japanjp.api.openai.comYesgpt-6-luna onlyStandard pay-as-you-go; no hosted tools in region
Indiain.api.openai.comYesgpt-6-luna onlyStandard pay-as-you-go; no hosted tools in region
United Arab Emiratesae.api.openai.comYesYesContact sales; listed models include gpt-5.6-luna and gpt-5.5-2026-04-23
United Kingdomgb.api.openai.comYesNoContent may be processed outside the region
Canadaca.api.openai.comYesNoContent may be processed outside the region
Singaporesg.api.openai.comYesNoContent may be processed outside the region
South Koreakr.api.openai.comYesNoContent may be processed outside the region
PriceResidency endpointsThe guide puts a 10% uplift on data residency endpointsThe pricing page puts a 10% uplift on regional processing endpointsModels released on or after 5 March 2026; neither page says whether storage-only regions pay it

Sources: [1] [2] [3] [4]

What is not known yet

We read OpenAI's release notes, data residency guide, API pricing page and ChatGPT data residency article on 20 November 2026. This post is analysis of those pages, and it is not legal advice.

Eligibility
OpenAI's pages don't say what makes an API organisation eligible, beyond contacting sales.
UK processing
None of the pages says whether or when the UK, Canada, Singapore or South Korea will get regional processing.
UAE access
The guide says to contact sales for the UAE region, and doesn't say how access is granted or priced. The view we read named two of its processing models and not the third.
Uplift on storage-only regions
The guide says data residency endpoints carry the 10% uplift, and the pricing page says regional processing endpoints. Neither says plainly whether a storage-only region pays it.
Zero Data Retention
The release note says the set-up needs no separate Zero Data Retention application. The sources we read don't explain how Zero Data Retention and data residency relate.
Announcement
We found no announcement post on openai.com/news, and the guide shows no date for its last update.

Sources: [1] [2] [3]

For engineers: domain prefixes, per-request residency, keys and headers, model and tier support, endpoint limits and the TLS path

Domain prefixes and project keys

The ten domain prefixes are us., eu., au., jp., in., ae., gb., ca., sg. and kr.api.openai.com. A fixed-region project is created by an organisation owner under Projects, then Create, then Data and inference residency. Create an API key in the new project and send requests with it to that region's prefix, for example eu.api.openai.com.

The guide says the region can't be changed after creation. By our reading, a move to another region means a new project, new keys and a new base URL. Stored objects such as files and batches would need re-creating.

Sources: [2]

Per-request residency on a Global project

For a new Global project, leave Data and inference residency set to Global and select Enable per-request data residency before Create. The project stays Global. Send each request that needs regional processing to a regional domain prefix with that project's key. The endpoint and model must support regional processing.

Existing Global projects that already support per-request residency can keep using their keys.

Sources: [2]

Keys and headers in the guide's example

The guide's code example uses one client with the Global project's key for global and US requests. EU processing and storage need a key from an EU project. The example builds a second client with the EU key and base_url https://eu.api.openai.com/v1, and that client omits the OpenAI-Project and OpenAI-Organization headers.

Sources: [2]

Model and service tier support

GPT-6.2 Astra and GPT-6.2 Sol support US and EU data residency on Standard, Fast and Batch, and Ultrafast with US data residency only. GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Sol and GPT-6 Luna support EU residency on Standard, Flex and Batch. GPT-6.1 Sol and GPT-6.1 Terra support US and EU residency only.

Fast mode supports EU residency for GPT-6.1 Sol, GPT-6.1 Terra and GPT-6 Luna. It is not available with EU residency for GPT-6 Astra or GPT-6 Sol. Ultrafast for GPT-6.1 Sol supports US and EU residency and global processing. GPT-6 Astra Ultrafast supports US residency and global only.

In Australia, Japan and India, gpt-6-luna is processed in region on Standard pay-as-you-go only. It runs through Responses and Chat Completions, with text and image input and text output. Requests need a project in that region, and hosted tools are unavailable for regional processing there.

Sources: [2]

Endpoint limits

/v1/chat/completions can't set store=true in non-US regions. Extended prompt caching on Chat Completions, Responses and Decisions, in a region without regional processing, may require OpenAI to process and temporarily store Customer Content outside the region. On /v1/realtime, tracing is not currently EU data residency compliant.

GPT-Live supports data residency in the US and Europe, and its sessions are eligible for Zero Data Retention. Stored sessions, where enabled, are kept for 30 days, and delegated backend models and tools have their own data controls.

The Decisions API (/v1/decisions, on gpt-6-luna) is available in every supported region, with regional processing in the US and Europe only.

Sources: [2]

Supported services and the TLS path

The guide lists 27 supported services. They include Responses, Chat Completions, Batches, Embeddings, Files, Fine-tuning, Images, Realtime and GPT-Live. Decisions, vector stores, File Search, Web Search, Code Interpreter, the Remote MCP server tool, Scale Tier and Structured Outputs are listed too. Structured Outputs is covered excluding the schema, which counts as system data, and file uploads are covered when sent as base64.

For us., eu., au., in. and jp.api.openai.com, OpenAI uses Cloudflare Regional Services, so TLS termination and HTTPS decryption happen inside the selected processing region. The guide's list doesn't include ae. or the storage-only prefixes.

Sources: [2]

Where 1AYM fits

If you are choosing an OpenAI API region for workloads that hold personal or regulated data, bring in 1AYM. Deciding which control each AI workload needs is part of our AI governance implementation work, sold as our Enterprise AI Platform & Agentic Workflows engagement. Where a region is a requirement, data residency is architected in, and we can say which region each store sits in.

We build identity and permissions the system enforces, and an audit trail produced as the controls run. We work to UK GDPR and the Data Protection Act 2018, plus whatever your sector adds. We work inside our clients' ISO and SOC 2 compliant environments, including banks and other highly regulated organisations.

To sort which workloads need storage in a region and which need processing there, start with our AI Opportunity & Feasibility Sprint. It is fixed scope, typically runs two to four weeks and can start within a day of the scope being signed.

1AYM takes small fixed-scope statements of work as well as larger builds, and we work with clients across the UK, the US and the Gulf. 1AYM has a UAE entity, so UAE work can be contracted locally.

Sources

  1. [1]OpenAI Help Center, ChatGPT Enterprise and Edu release notes: Self-serve regional data residency for API projects, 18 November 2026 (checked 20 November 2026)
  2. [2]OpenAI API documentation: Data residency (checked 20 November 2026)
  3. [3]OpenAI API documentation: Pricing (checked 20 November 2026)
  4. [4]OpenAI Help Center: Data residency and inference residency for ChatGPT (checked 20 November 2026)

Frequently asked questions

Who can help us choose an OpenAI API data residency region?

1AYM. Where a region is a requirement, we build data residency in and can say which region each store sits in. Book a 30-minute call or email enquiries@1aym.com.

Does OpenAI process API data in the UK?

OpenAI's guide lists no regional processing for the UK. The UK region, gb.api.openai.com, stores Customer Content at rest in the UK, and OpenAI may process that content outside the UK. The Europe region processes in the European Economic Area and Switzerland.

What does OpenAI API data residency cost?

OpenAI's pricing page adds a 10% uplift on regional processing endpoints for models released on or after 5 March 2026. The guide words it as data residency endpoints. On GPT-6.2 Sol that makes $2.20 input and $11 output per million tokens, by our arithmetic.

Can we change an OpenAI API project's region later?

No. OpenAI's guide says a project's residency region can't be changed after creation. To use another region, create a new project and new API keys for it.

Is the UAE region available on the OpenAI API?

Yes, through sales. ae.api.openai.com offers storage and processing, and OpenAI says to contact sales to access it. The models listed for processing there on Chat Completions include gpt-5.6-luna and gpt-5.5-2026-04-23.

What does OpenAI API data residency not cover?

It doesn't cover system data, such as account, usage and billing data, or data sent to third-party services such as remote MCP servers. Nor does it cover data that leaves the region because of where your users or systems are.

About the author

Tayyeb Mahmud is the founder and Principal Architect of 1AYM. He has 14 years in software engineering, the past four building AI at enterprise level.

Claude Certified Architect – Professional · LinkedIn profile of Tayyeb Mahmud

Further

Latest analysis

  • ChatGPT SSO exemptions for external members · On 19 November 2026 OpenAI gave ChatGPT Enterprise, Edu, Healthcare and Business owners an SSO setting for each external member. New external members follow the workspace's SSO policy by default. Existing ones in workspaces that required SSO may hold an exemption, and owners and admins can check each member in the Admin Console.
  • Gemini 4 Neon pricing and availability · Google released Gemini 4 Neon on 18 November 2026 and placed it between Hydrogen and Carbon. It costs $3 per million input tokens and $15 per million output tokens, more than Carbon until 30 November and less from 1 December. Its default thinking level is higher than Carbon's, and it is in GitHub Copilot.
  • GPT-6.2 Astra and Sol: price and where they run · OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on 17 November 2026. Astra now costs $8 and $40 per million tokens, the same list price as Claude Fable 5.2, and Sol's price is unchanged. This post covers the price, where the models run, the admin steps and what OpenAI's safety findings mean for agents.

Choosing an OpenAI API region?

Half an hour is enough to sort your workloads by storage and processing needs, pick a region for each and agree who approves new projects.

The engagement

AI governance implementation

  • Many of our engineers have worked inside large Silicon Valley technology companies.
  • Fixed-scope work can start within a day of the scope being signed.
  • Day rates start at £750 to £1,250, depending on seniority and the project.

See our work for a government-accredited EdTech in the Middle East

Last reviewed · 1AYM