Amazon weighs buying Decart to strengthen its AI chip stack

Wall Street Journal reports that Amazon is in talks to acquire Decart, an AI startup, in a deal that could be valued around 7 billion dollars. Reuters relayed the report and framed the transaction as still under discussion, not a completed acquisition.

The target is strategically interesting because Decart is not only another generative-video company. Its Decart Optimization Stack is marketed as software that helps AI teams run models across Nvidia GPUs, AWS Trainium and Google TPUs with better performance and utilization. The company also builds real-time world-model products including Lucy and Oasis.

For Amazon, that makes Decart a possible infrastructure lever. AWS has been pushing Trainium as an alternative to Nvidia-heavy deployments, but customers still care about portability, latency and developer friction. Owning a chip-optimization layer could make Amazon’s AI stack more credible for teams that want cheaper or more flexible compute.

The caveat is important: this is a reported negotiation, not an announced deal. But if it closes, it would be another sign that cloud providers are buying up the software layers around AI hardware, not just building larger data centers.