What happened
Anthropic launched in-country Claude inference for India through Amazon Bedrock, routing requests across AWS Mumbai and Hyderabad. IBM-Yotta launched a sovereign agentic AI platform, and IndiaAI Mission has approved 237 compute projects with 93.18 lakh GPU hours and a 1.1 EFLOPS system at NIC Delhi.
Why it matters
Regulated enterprises can now keep AI inference inside India, easing data-residency and compliance friction — though the domestic compute pool remains constrained.
Remember this
The sovereignty debate has moved from where data is stored to where AI inference runs.
More context & who is affected
Anthropic now runs Claude inference from inside India via Amazon Bedrock, routing requests across AWS Mumbai and Hyderabad regions rather than the US.
The sovereignty argument has shifted from where data is stored to where AI inference happens — the key ask from regulated enterprise buyers handling financial, health or government data.
On infrastructure, IBM-Yotta launched a sovereign agentic AI platform combining watsonx Orchestrate with Yotta's Shakti Cloud and Studio. IndiaAI Mission has backed 237 projects with 93.18 lakh GPU hours and a 1.1 EFLOPS system at NIC Delhi.
- IndiaAI projects approved
- 237 237 projects approved for subsidised compute under IndiaAI Mission.Business Standard
- GPU hours sanctioned
- 93.18 lakh 93.18 lakh GPU hours sanctioned across approved IndiaAI projects.Business Standard
- High-performance compute
- 1.1 EFLOPS 1.1 EFLOPS AI compute system being established at NIC Data Centre, Shastri Park, Delhi.Business Standard
How it came together
- Sep 2026 — Anthropic announces in-country Claude inference via Amazon Bedrock across AWS Mumbai and Hyderabad.
- 29 Sep 2026 — IBM and Yotta launch the Sovereign Agentic AI Platform for Indian enterprises.
- 6 Oct 2026 — Business Standard reports the pivot from data localisation to AI localisation, citing IndiaAI's 237 projects and 1.1 EFLOPS system.
Enterprise Decision Makers
Regulated enterprises in India can now run Claude inference in-country, easing data-residency and compliance pressure.
Tech Executives · Policy Makers
IndiaAI Mission's 237 approved projects and 93.18 lakh GPU hours grow the domestic AI compute pool, though supply remains tight.
Enterprise Decision Makers · Tech Executives
IBM-Yotta's sovereign agentic AI platform gives enterprises a domestic stack for agents, models and governance together.
Reporting fromBusiness Standard ↗AWS ↗
Sources & verification2 sources · 4 claims
These are the records behind this briefing. Linked reporting is not the same as independent confirmation.
- Business StandardReporting
From data localisation to AI localisation: Why inference is moving to India ↗
Published Oct 6, 2026 · Accessed Oct 6, 2026 - AWSOfficial
Amazon Bedrock expands Claude model availability to in-country inferencing in India ↗
Published Sep 25, 2026 · Accessed Oct 6, 2026
Claims and linked sources
Anthropic launched in-country Claude inference for India through Amazon Bedrock, routing requests across AWS Mumbai and Hyderabad regions.
IndiaAI Mission has approved 237 projects for subsidised compute, sanctioned 93.18 lakh GPU hours, empanelled 15 compute providers and set up a 1.1 EFLOPS system at NIC Delhi.
The AI sovereignty debate in India has shifted from data localisation to AI localisation — where inference runs now matters as much as where data is stored.
IBM and Yotta launched the Sovereign Agentic AI Platform combining IBM watsonx Orchestrate with Yotta Shakti Cloud and Studio for onshore enterprise AI.
