On November 12, OpenAI announced API inference residency in India: when eligible businesses and developers call GPT-6 Luna, the request is received, the model computes, and the result returns entirely inside India, and customer content is stored in the country too. India becomes the third market to get OpenAI inference residency, after Australia and Japan.
This fills the piece data residency was always missing. Until now, OpenAI's data residency in India only covered "where it is stored": customer content could sit on servers in India. Inference residency covers "where it is computed": the whole process of the model receiving a prompt and generating an answer stays in-country too. Together, a single API call never leaves Indian infrastructure from start to finish. At launch, only GPT-6 Luna is supported, through the Responses API and the Chat Completions API. Ordinary ChatGPT users are unaffected, and hosted tools are not supported yet.
Three hard conditions apply:
- The organization must enable Zero Data Retention (ZDR), under which OpenAI retains no API request or response data, and approval from OpenAI is required;
- Only standard pay-as-you-go processing is supported — annual plans and volume discounts are out of scope;
- A new API project must be created, with India selected as the region for both storage and processing, and requests routed through the India regional endpoint. OpenAI's documentation shows a 10 percent premium for regional processing of GPT-6 Luna.
Regulated industries such as finance stand to benefit first. Nitin Bawankule, OpenAI's head of enterprise for India, said some Indian businesses have not fully deployed AI for important work precisely because they need the information involved to be processed within the country. For those customers, inference residency unties a compliance knot: if data never leaves the country, audits get easier. Anthropic's Claude brought in-country inference to India more than a month ago, and OpenAI following now means India is becoming the main battleground where model providers compete for compliance-driven markets.
Two things are worth watching next: whether OpenAI extends inference residency to more models, and whether it takes it to more markets with data localization requirements. In February, OpenAI and the Tata Group announced plans to build domestic AI data center capacity together; inference residency is the next move in that plan. For now it remains a narrow opening: one model, direct API access, eligible customers only. Ordinary users and the ChatGPT experience will not change because of it.
