Skip to content
RegionSouth Asia

LowV Technology Geopolitics & AI12 November 2026, Thursday

OpenAI launches in-country inference in India, easing the data localisation barrier

On 12 November OpenAI launched in-country inference and data processing in India for its API services. According to Inc42, eligible organisations can keep supported workloads inside the country; India is the third market to get the service after Australia and Japan.

Location: MUMBAI

According to Inc42, the service initially covers only the GPT-6 Luna model and two API interfaces. According to FoneArena on 12 November, customers must open a new API project that designates India as both the storage and the processing location. The first target audience is the 3 sectors most tightly bound by data localisation rules: banking, finance and insurance. Inference is the stage at which a trained model runs to generate responses. For banks subject to localisation rules, the 2 critical steps, processing and storage, now happen inside the country.

OpenAI's head of enterprise sales for India said on 12 November that some companies could not use AI in critical work because their data had to stay in the country. According to Inc42, more than 10 crore, or 100 million, people in India use ChatGPT every week. In 2026 the company formed a partnership with Tata Group for domestic data centre capacity.

According to Inc42, Anthropic announced in August that it would bring in-country inference to India via Amazon Bedrock. Local hosting had been the distinguishing advantage of Indian start-ups such as Sarvam AI; the arrival of 2 US companies could erode that edge. As of 12 November, neither the number of customers covered by the service nor its pricing had been disclosed; these details could not be verified.

Talay assessment

Bottom line

In-country inference turns India's data localisation demand from a barrier for US model providers into an infrastructure investment. Coverage is limited to a single model today, but the door is opening for banks and public bodies. The most likely path is that coverage extends to other models in 2027 and domestic start-ups' hosting advantage wears away.

Likely effects

  • Banking and public sectorPositive1–6 months

    Banks and public bodies held back by localisation rules can bring AI into critical workflows through in-country inference.

  • Domestic AI start-upsNegative1–6 months

    The in-country hosting advantage of domestic firms such as Sarvam AI narrows as two large US providers enter.

  • Data centre demandPositive6 months+

    In-country inference lifts demand for domestic capacity projects such as the Tata partnership.

Possibilities, ranked

  1. 1
    Gradual expansion60%

    OpenAI extends in-country inference to more models and services within a few months.

    Watch: Models added to the India region in OpenAI developer documentation

  2. 2
    Regulatory limits25%

    India attaches extra conditions to US providers' in-country inference on sovereign AI grounds.

    Watch: MeitY or RBI guidance on foreign model providers

  3. 3
    Limited take-up15%

    A single model and the zero data retention requirement keep enterprise adoption slow.

    Watch: Enterprise deals Indian banks announce with OpenAI

Probabilities are calibrated judgement based on the sources, not measurement, and are revised as new information arrives. Not investment advice.

Market reaction

Indicators affected

  • Weekly ChatGPT users▲ Over 10 crore
  • Models covered▼ 1

Sources

  1. Inc42 — OpenAI brings API inference residency to India
  2. FoneArena — OpenAI API now processes and runs locally in India
Share