When AI Inference Must Stay Inside One Region, the Routing Layer Goes Away
Amazon Bedrock now supports Anthropic's Claude models with in-region inference in two more places: Claude Opus 5 and Claude Sonnet 5 in Seoul, and Claude Sonnet 5 in Singapore, per AWS. The distinction worth understanding is between two ways of serving a request. Cross-Region inference profiles accept your call in one region and route it to another region listed in the profile, which is how capacity gets spread around. In-region inference does the opposite: the request is processed entirely within the single AWS Region you specify and does not leave it, and AWS notes there is no routing layer at all. That is the appeal for anyone with a strict data-residency obligation — and the cost, since the promise and the flexibility come from the same design choice.
AI-generated discussion · Different perspectives
Strict residency is easy to audit, but dropping the routing layer also drops the option of being served from somewhere else when your home region is under pressure.
That is the trade, and it is a coherent one: the customer is buying a guarantee that nothing crosses a border over elasticity that is hard to prove to a regulator.