CoreWeave signs multi-year deal to train Wayve's AI Driver

The agreement extends an existing relationship rather than starting one from scratch: Wayve began on Weights & Biases, a tool it used before scaling into CoreWeave Cloud. That progression matters because the workload in question is a frontier training job with physics and safety attached, as Wayve puts it — the kind that needs reliability across long training runs and enough observability to prove one model version beats the last.

Both sides frame the deal around that demand. Wayve's Silvius Rus describes a data flywheel of gathering real-world driving experience, training, evaluating, and deploying only what outperforms what is already on the road, and says CoreWeave's purpose-built cloud helps accelerate commercial deployment. CoreWeave's Gavin Jackson casts it as bringing one of the most demanding physical AI programs onto a platform built for exactly that, as AI moves from the screen into the physical world.

CoreWeave leans on benchmark credentials in making that pitch: it cites record-breaking MLPerf results in inference and training, three consecutive Platinum rankings in SemiAnalysis ClusterMAX as the only AI cloud to do so, and the number-one ranking for inference speed and price-performance for Moonshot AI's Kimi K2.6 and Kimi K2.7 Code in independent benchmarking by Artificial Analysis.