Amazon Web Services (AWS) is running into some bumps with its AI platform, Bedrock, as users express frustration over limitations when accessing Anthropic’s models. According to The Information, despite AWS’s hefty investment in Anthropic, customers have flagged issues like arbitrary usage caps and missing features such as prompt caching—capabilities they can get directly through Anthropic’s own API. This has led startups like Lovable and Praxis AI to switch over to Anthropic’s native interface instead.
Inside AWS, conversations have described the platform’s capacity situation as a “disaster,” though spokesperson Kate Vorys pushes back against that label. She explains the restrictions are part of standard industry measures designed to guarantee “fair access” amid what she calls “unprecedented demand.”
The partnership between Amazon and Anthropic centres on AI and cloud infrastructure. Amazon has invested up to $8 billion, making it Anthropic’s largest backer. Anthropic relies on AWS as its main cloud provider, running its Claude AI models on AWS’s custom chips like Trainium and Inferentia. The goal is to attract enterprise and government clients to Amazon’s Bedrock AI platform.
Amazon’s approach aims to boost AWS cloud revenue by supporting Anthropic’s AI workloads. It’s a cycle where investing in AI startups fuels demand for AWS services, which in turn grows Amazon’s revenue. This strategy echoes moves by Microsoft and Google, who have also poured billions into AI companies to drive cloud usage. For example, Microsoft has invested over $13 billion in OpenAI, which runs its models on Azure and integrates them into Microsoft products. Google has made similar plays with firms like Anthropic and Runway.
If you’ve ever wrestled with platform limits or missing features when working with AI services, this situation highlights the trade-offs between direct API access and third-party integrations. For AWS customers, it’s worth keeping an eye on how these partnerships evolve and what that means for flexibility and performance. Meanwhile, startups are making pragmatic choices to ensure they get the features and reliability they need.
