Bengaluru: Anthropic has expanded its artificial intelligence infrastructure in India by making in-country inference for Claude models available through Amazon Bedrock, allowing eligible requests made through the India endpoint to be processed within Indian regions.

The development is particularly important for banks, government organisations, healthcare institutions and large enterprises that handle sensitive information and require stronger control over where AI workloads are processed.

Claude Requests Can Stay Within India

Under the new India geographic inference setup, Claude requests can be processed between AWS infrastructure in Mumbai and Hyderabad. This allows organisations to use additional computing capacity while keeping inference within India.

The available models include Claude Opus 5, Claude Sonnet 5 and Claude Haiku 4.5. AWS says the geographic setup routes requests only between the designated Indian regions.

This gives Indian enterprises an alternative to global processing when local data-handling requirements are important.

Major Advantage for Regulated Industries

Data residency has become a major consideration for organisations adopting generative AI.

Financial institutions, government bodies and other highly regulated businesses often need to understand where confidential prompts and generated responses are processed. Local inference can make it easier for these organisations to design AI deployments around their data-governance requirements.

Anthropic India Managing Director Irina Ghose said local processing had been a specific requirement from customers wanting to use Claude for critical workloads.

Growing Demand for Claude in India

India has emerged as an important market for Anthropic.

The company says India is its second-largest market for Claude.ai, with software development accounting for 45.2% of work-related tasks that users bring to Claude in the country. Adoption has also expanded across banking, payments, IT services, healthcare, science and public services.

Indian technology companies are also increasingly incorporating Claude into their operations. TCS is deploying Claude to tens of thousands of employees, while other major technology and digital businesses are using the models for software development, engineering and enterprise applications.

Enterprise Security and Governance

Claude deployments through Amazon Bedrock provide enterprise-oriented controls such as audit trails and access controls, which can help organisations monitor AI activity and establish governance processes.

However, local processing does not automatically guarantee compliance with every legal requirement. Companies still need to configure their cloud environment properly and assess their own rules concerning data access, retention, security and privacy.

India Could See Faster Enterprise AI Adoption

The availability of local inference could help organisations that had previously delayed large-scale AI deployments because of data-location concerns.

AWS has indicated that data residency can be a determining factor in whether AI projects remain at the pilot stage or move into full production. Early customers have also cited performance, latency, resilience and regulatory considerations as reasons for using local inference.

For businesses handling sensitive information, the ability to keep AI processing within the country could therefore remove one important barrier to wider adoption.

A Significant Step for India’s AI Ecosystem

Anthropic’s local Claude infrastructure strengthens competition in India’s rapidly expanding generative AI market.

The development also reflects a broader shift in enterprise AI: companies are no longer evaluating AI models only on their intelligence and capabilities. Data residency, security, governance, reliability and regulatory compliance are becoming equally important factors.

With Claude inference now available within India through Amazon Bedrock, Indian enterprises have another option for deploying advanced AI while maintaining greater geographic control over their workloads.