Anthropic and AWS turned on general availability of Claude Platform on AWS Tuesday, giving developers Anthropic's native API surface directly through their AWS accounts. The companies signaled the move in April when they expanded their compute partnership.
Customers get the full first-party feature set: Messages API, Agent Skills (beta), MCP connector (beta), Claude Managed Agents (beta), web search, and code execution. Claude Opus 4.7, Sonnet 4.6, and Haiku 4.5 are available at launch, per Anthropic's launch post.
Here's the catch worth flagging: you authenticate with IAM credentials and pay through your AWS bill, but Anthropic operates the service and data leaves the AWS security boundary for inference. The product page makes this explicit. Bedrock, by contrast, keeps Claude usage entirely inside AWS infrastructure for teams with data residency requirements.
OpenRouter is named as a launch customer in the blog post, with its testimonial citing "consistent performance on uptime, latency, and throughput," the kind of line you expect on a launch post. Pricing matches Anthropic's direct rates, and coverage spans 18 AWS regions per the AWS announcement.
The launch follows the April agreement, in which Anthropic committed more than $100 billion to AWS over a decade for up to 5GW of Trainium capacity. Anthropic's run-rate revenue is now above $30 billion, up from roughly $9 billion at the end of 2025.
Bottom Line
Anthropic operates Claude Platform on AWS itself, meaning inference data leaves the AWS security boundary even though billing and IAM stay within AWS.
Quick Facts
- Generally available May 12, 2026
- Available across 18 AWS regions
- Claude Opus 4.7, Sonnet 4.6, and Haiku 4.5 supported at launch
- Pricing matches Anthropic's direct platform rates
- Data processed outside AWS security boundary (company-disclosed)
