OpenAI brings local AI processing to UAE

OpenAI has launched Inference Residency in the UAE, allowing eligible organisations to run artificial intelligence model inference on graphics processing units located within the country while using its enterprise and developer services. The move strengthens local control over sensitive AI workloads as government bodies and regulated industries expand their use of generative AI.

The service is available to eligible customers using OpenAI’s API, ChatGPT Enterprise and ChatGPT Edu. It extends the company’s existing UAE Data Residency offering by addressing not only where customer information is stored, but also where the computational work involved in generating AI responses takes place.

Inference Residency means GPU execution involving covered customer content can remain within the UAE. Prompts, files, conversations and embeddings derived from eligible content can therefore undergo model inference on GPUs located inside the country. Data residency, by comparison, determines where covered customer information is stored when the service retains it.

The distinction is significant for organisations handling commercially sensitive, institutional or regulated information. Enterprises increasingly want clarity not only about where their data rests but also about where it is processed when an AI model responds to a query, analyses a document or generates other output.

UAE customers seeking inference residency must have data residency enabled in the same region. The UAE joins the United States and Europe, comprising the European Economic Area and Switzerland, among the locations where OpenAI currently supports ChatGPT inference residency. Eligibility is not automatic, and organisations can engage with OpenAI’s sales operation to establish whether their deployment qualifies.

Farouk El Hamzawi, head of enterprise at OpenAI MENA, said organisations across the Emirates were adopting AI rapidly, supported by the government’s long-term technology strategy. He said the new capability would give eligible customers greater control over where their workloads were executed and provide stronger regional assurances for institutions scaling their use of AI.

The launch nevertheless carries technical limitations. GPT-5.2 is currently the model available for ChatGPT workspaces configured for UAE inference residency, a restriction that also applies to API inference in the country. GPT-5.6 is not supported for workloads configured to use UAE inference residency, although eligible users in the Emirates can use it when that residency configuration is not applied.

Image generation and internal search are not supported in ChatGPT under the UAE configuration. ChatGPT Work and improved memory are also unavailable, although standard saved memories remain supported. GPT-Live is not available for UAE inference-residency workspaces, while Advanced Voice can continue to operate where it is otherwise enabled. External web search may remain accessible unless administrators disable it through workspace controls or Lockdown Mode.

Inference Residency also does not mean every element of OpenAI’s processing takes place inside the Emirates. The guarantee applies specifically to GPU execution involving covered customer content. Authentication, routing, analytics, other CPU-based activities and some external integrations may still operate outside the selected region. System information such as billing details, account information and certain metadata is also outside the scope of the customer-content residency guarantee.

For API customers, residency controls are configured at project level. OpenAI requires additional approval for projects selecting the UAE region, while non-US regions are subject to further data-control requirements. The company has also differentiated regional storage from regional processing in its technical documentation, reflecting the more demanding infrastructure needed to guarantee where inference actually occurs.

OpenAI introduced UAE Data Residency for business users in November 2025, enabling eligible enterprise, education and API customers to keep covered information stored at rest within the Emirates. That offering addressed a key concern for institutions seeking stronger governance over information used with generative AI systems.

The expansion comes alongside a broader build-out of AI computing infrastructure in the country. Stargate UAE, announced in 2025 as OpenAI’s first international Stargate deployment, is planned as a one-gigawatt AI computing cluster in Abu Dhabi. The first 200 megawatts of capacity were targeted to become operational during 2026. Partners named in the project include G42, Oracle, Nvidia, Cisco and SoftBank.