Wayve trains AI driver models on CoreWeave Cloud with Blackwell GPUs – Unite.AI

CoreWeave announced on November 9, 2026, a multi-year agreement with Wayve to train its AI Driver core models on CoreWeave Cloud with NVIDIA Blackwell GPUs. The work spans training, evaluation and simulation for a system that learns from real-world driving data collected in more than 70 countries.

Wayve started with Weights & Biases and has since moved to CoreWeave Cloud. The expansion adds CoreWeave cloud computing to a development process that must show progress between software releases, support long training cycles, and provide enough observability to compare models and validate security at scale.

A data flywheel for driving models

Wayve is developing the AI ​​driver for uses ranging from personally owned cars to robotaxis. Silvius Rus, the company’s senior vice president of engineering, described how its global driving data fuels model development: “Our AI driver improves by running this data flywheel: collecting real-world driving experience, training on it, evaluating it, and implementing only what surpasses what’s already on the road.” He expects access to CoreWeave’s AI cloud to help accelerate commercial deployment of the AI ​​driver.

Simulation before road distribution

A CoreWeave technical overview of physical AI infrastructure identifies world model training, sensor simulation, and closed-loop evaluation as autonomous vehicle workloads. Autonomous vehicles generate terabytes of sensor data per vehicle per day. Managing that flow requires storage, networking, and training channels to work together efficiently.

Physical AI development continues post-deployment as new edge cases appear, necessitating repeated operational requirements for retraining, checkpointing, and fast data pipelines. Because failures can have physical safety consequences, most model development goes through phased simulation and validation before real-world use.

In a CoreWeave reference demonstration outside the Wayve deal, the company reports running 1.25 million CARLA autonomous vehicle simulations in about 12 hours on independent shards using one GPU each. The runs varied weather conditions, traffic densities, maps, intersections and seeds, illustrating the range of simulated conditions that can be generated for model development.

Reliability across the entire training stack

On the hardware level, NVIDIA says Blackwell’s reliability, availability and maintainability engine continuously monitors thousands of data points about the health of the hardware and software to predict and intercept sources of downtime. It is designed to early identify potential failures, locate problems and facilitate maintenance, features relevant to the long training cycle reliability requirements described in the CoreWeave announcement.

Across its broader physical AI platform, CoreWeave says its Kubernetes service orchestrates training and simulation jobs, with distributed and object storage that contains petabyte-scale sensors and simulation data. InfiniBand and RoCE networks move data directly between server GPUs, bypassing CPUs to reduce downtime while distributed jobs exchange sensor data, simulation state, and gradients.

Experiment tools add version lineage: Weights & Biases models track experiments, Registry links a checkpoint to the exact dataset and training run that produced it, and Weave supports traces, evaluations, and monitoring. These records provide engineers with a path from an issue encountered during testing to the model checkpoint, data set, and execution of the training involved. For a safety-critical driving system, such traceability supports comparison and validation alongside training capability.

The agreement makes the entire iteration cycle part of Wayve’s infrastructure expansion. Faster commercial implementation is tied to how efficiently integrated training, evaluation and simulation can transform the driving experience into validated improvements, while maintaining the reliability and traceability needed for safety-critical driving software.