OpenAI brings GPT-6 Luna processing to India

The move extends OpenAI's existing data residency arrangements in India. Until now, the company's India residency offering covered the storage of customer data. The 12 November announcement adds inference, the stage at which a model receives a prompt and generates a response, to the in-country scope. For eligible customers, the full cycle of an API request, from receipt through computation to the return of a result, now remains within Indian infrastructure.

What OpenAI announced

Under the new arrangement, eligible businesses and developers in India can run GPT-6 Luna workloads through the Responses and Chat Completions APIs with processing and computation located in India. Supported customer content is stored in the country as part of the same arrangement. The change is described by the company as live at launch for eligible customers, who can begin by creating a new API project, selecting India for storage and processing, and using the India regional endpoint.

The announcement forms part of the "OpenAI for India" initiative, the company's programme of India-focused activity, which includes a data-centre partnership with the Tata Group. The reports do not specify the number, locations or capacity of the data centres involved, and OpenAI did not disclose an investment figure.

From storage to full inference

The distinction between storage residency and inference residency is central to the announcement. Data residency, as previously offered in India, meant that customer content was stored on servers located in the country. Inference residency goes further: the computation itself, the model run that turns a prompt into an output, is also performed in the country. The request therefore does not leave Indian infrastructure at any stage, according to the company statement.

India becomes the third market in which OpenAI offers inference residency, joining Australia and Japan. The three markets share the same conditions. Regional processing in each of them requires Zero Data Retention to be enabled and Standard pay-as-you-go processing to be used, and hosted tools are not currently available for regional processing in any of the three markets. These conditions were stated by the company and reported across the outlets covering the announcement.

How eligible customers start

Access is limited to eligible businesses and developers, and to supported workloads. A customer begins by creating a new API project, selecting India for both storage and processing, and directing requests to the India regional endpoint. The reports do not specify the eligibility criteria in detail, and it is not clear from the published accounts whether all existing Indian API customers qualify automatically or whether eligibility is assessed case by case.

The scope of the launch is narrow in one further respect. At launch, the in-country processing applies to GPT-6 Luna only. Other OpenAI models were not named in the announcement as part of the India processing capability. The company did not state a timetable for extending regional processing to additional models.

The conditions and limits

Three conditions define the practical scope of the announcement. First, regional processing requires Zero Data Retention to be enabled, the setting under which the company does not retain API request and response data. Second, it requires Standard pay-as-you-go processing. Third, hosted tools, the company-provided tools that can be invoked during an API call, are not currently available for regional processing in India, Australia or Japan.

The scope is also limited to direct API usage. The announcement does not change the experience of regular ChatGPT app users, whose requests are not covered by the new arrangement. The change concerns businesses and developers who call OpenAI models through the API and who require processing to remain inside the country.

Why India, and why now

The announcement is aimed at data-localisation requirements, the rules and policies under which certain kinds of data must be processed and stored within a country's borders. Financial services firms are named in the reporting as a category of customer for which such requirements are particularly strict, because regulators in the sector commonly require sensitive data to remain inside the jurisdiction. Nitin Bawankule, Head of Enterprise for India at OpenAI, was named in the coverage as the executive associated with the announcement.

India is one of the largest markets for OpenAI by user numbers, and the company has been expanding its presence there through the "OpenAI for India" initiative. The Tata Group data-centre partnership, cited as background in the reporting, is part of that expansion. The reports do not state whether the in-country processing announced on 12 November runs on infrastructure built under that partnership.

Why it matters

For Indian businesses subject to data-localisation rules, the difference between storage residency and inference residency is practical rather than technical. A company that stores data in India but sends it abroad for processing may still fall short of a requirement that the data remain within the country throughout its handling. Full in-country inference closes that gap for the workloads covered by the announcement, at least for customers who meet the stated conditions.

The conditions, however, narrow the benefit. A customer that relies on hosted tools, or that uses a model other than GPT-6 Luna, cannot yet use the regional processing capability. The requirement for Zero Data Retention and Standard pay-as-you-go processing may also exclude customers on other processing arrangements. The announcement therefore represents a defined first step rather than a general availability of in-country processing for all Indian API use.

What to watch next

The reporting to date leaves several questions open. OpenAI has not disclosed a timetable for adding models beyond GPT-6 Luna to the India regional processing capability, nor for making hosted tools available for regional processing. It has not published a list of eligible customers or detailed criteria, and it has not stated whether capacity will expand to further markets.

The development is also likely to be measured against competitors. Other providers of frontier models have made their own data-residency commitments in regulated markets, and the value of OpenAI's announcement will depend in part on how quickly the conditions are relaxed and the model coverage widened. For now, the verifiable position is that GPT-6 Luna API workloads for eligible Indian customers can be processed and stored entirely in India, subject to Zero Data Retention and Standard pay-as-you-go processing, without hosted tools, through the India regional endpoint.