Anthropic launches in‑country Claude inference in India via Amazon Bedrock

Anthropic brings Claude inference to India

Anthropic has made its Claude family of generative‑AI models—Claude Opus 5, Sonnet 5 and Haiku 4.5—available through an India‑specific endpoint on Amazon Bedrock. When a customer in India calls the endpoint, the prompt and the model’s response are processed on servers located inside the country, a capability the company says it has been promising since August.

“Our customers, especially in regulated sectors and government, want to use Claude on their most critical work, with requests processed in the country. From today, they can,” said Irina Ghose, Managing Director, Anthropic India.

The announcement was made on 5 October 2024 and is aimed squarely at organisations that must keep data within Indian borders – banks, insurers, public‑sector bodies and large enterprises that face strict data‑handling rules.

Why local inference matters

Sandeep Dutta, President of AWS India and South Asia, explained that data residency can be the difference between a pilot project and full‑scale production:

“Responsible AI adoption at scale has become a strategic imperative for some of India’s most regulated industries. For them, data residency is the parameter that decides whether AI stays confined to pilots or reaches full production.”

AWS’s Satinder Pal Singh added that Bedrock’s built‑in audit trails and access controls satisfy risk‑and‑compliance teams, giving enterprises the governance they need before moving AI into production.

Early adopters and rollout plans

Anthropic named several early customers who tested the capability in private preview, including Reliance Industries, CRED, TCS, Infosys, Kotak Bank, Mahindra Group, Godrej Industries, Swiggy and Freshworks. TCS is equipping about 50,000 associates with Claude, while Infosys is integrating the model with its Topaz platform and has set up an Anthropic Centre of Excellence.

The company says the Indian market is its second‑largest globally, with 45.2 % of Claude usage coming from software‑development tasks – the highest share of any region.

Building a longer‑term presence

Anthropic opened its Bengaluru office in February 2024 and plans to keep expanding the team and local partnerships. The launch coincides with a surge in community activity: when the Claude Partner Network rolled out in March, India recorded some of the highest registration numbers worldwide, and a certification event in August saw 5,000 people become Claude‑certified.

In parallel, Anthropic is working with the Indian non‑profit Karya to improve Claude’s training data across ten Indian languages – Hindi, Bengali, Marathi, Telugu, Tamil, Punjabi, Gujarati, Kannada, Malayalam and Urdu – to boost performance for local users.

What’s next?

Ghose expects the momentum to grow rapidly, noting that local inference could unlock deployments that were previously on hold because of regulatory constraints. With Bedrock’s infrastructure handling the heavy lifting, Anthropic says enterprises can now focus on building applications rather than worrying about where the data travels.

For companies navigating India’s Data Protection Bill and sector‑specific rules, the ability to keep AI workloads on‑shore removes a major barrier to scaling generative‑AI solutions.

---

Key take‑aways

• Claude Opus 5, Sonnet 5 and Haiku 4.5 are live on an India‑hosted Bedrock endpoint.
• Requests are processed on servers inside India, satisfying data‑residency requirements.
• Early adopters span finance, IT services, e‑commerce and consumer tech.
• Anthropic’s India team is expanding, with a focus on language localisation and partner certification.