What the evidence says
AWS Machine Learning says customers in India can access Anthropic Claude models on Amazon Bedrock while processing data in India Regions, alongside the previously supported global cross-Region inference option. The announced India regional endpoint is served through geographic cross-Region inference.
The account describes that inference arrangement from the Mumbai and Hyderabad Regions. It also identifies several ways to get started: the Amazon Bedrock console, Anthropic’s Messages API, the Amazon Bedrock InvokeModel API, and the Converse API.
The source further states that Amazon Bedrock does not store model inputs or outputs by default. Taken together, the evidence supports a specific announcement about access, a regional processing path, available entry points, and a stated default storage posture.
Operator implications
The announcement makes endpoint choice a concrete part of evaluating Claude workloads in India. Teams can examine whether the India regional endpoint fits the request path they intend to use and whether console access, Messages, InvokeModel, or Converse aligns with their integration approach.
Three ideas should remain distinct in that evaluation: where data is processed, how requests are served through geographic cross-Region inference, and the stated default for input and output storage. The evidence connects these elements in the announced offering, but it does not turn them into one broad operational or governance conclusion.
Mumbai and Hyderabad are part of the described inference design, so routing is relevant to the practical shape of this option. That does not establish a particular result for an individual application; it simply frames the announced path as more than a model-selection decision.
Limits and open questions
This account is not independently confirmed. The source does not establish measured latency, capacity, pricing, comparative performance, the precise Claude model set available, or routing behavior in every operating condition.
It also does not establish workload-specific outcomes, implementation results, or whether the announced arrangement meets any organization’s particular requirements. The supported conclusion is limited: Amazon Bedrock has described an India geographic cross-Region inference option for Claude access, with processing in India Regions, documented console and API entry points, and a default statement on input and output storage.