Scaling Embodied AI for Autonomous Driving with Microsoft Azure

As our autonomous driving technology transitions toward real-world deployment, the demands placed on our AI products continue to grow. Scaling embodied AI from research into production requires infrastructure capable of supporting large-scale training, rigorous validation, and reliable operation across vehicle platforms, regions, and use cases.

Over the past several years, our collaboration with Microsoft has played an increasingly central role in meeting those demands. As our technology, teams, and customer programs have scaled, so too has the complexity of the AI workloads required to support them. To reflect this progress, we have extended our strategic partnership with Microsoft and are continuing to deepen our work together as we enter the next phase of deployment.

From R&D to deployment

Wayve’s AV2.0 approach to autonomous driving is built around embodied AI that learns directly from real-world data and experience. As our tech development moves into production environments, the requirements on infrastructure change significantly. Beyond training models, deployment demands continuous validation, customer model customization, automated evaluation loops, simulation, and operational readiness across diverse driving conditions.

Recent milestones, including our first customer agreement with global automaker Nissan and plans to launch robotaxi trials with Uber starting in the UK and expanding to other markets, signal this shift toward real-world deployment. These programs increase the need for production-grade cloud infrastructure that can support our software products as they scale across customers, regions, and use cases.

Building at scale on Azure

Azure has underpinned Wayve’s large-scale training and validation of embodied AI models since 2020. On Azure, we have been able to scale our generative world models, such as GAIA-3, alongside our own foundation models trained on one of the world’s most diverse driving datasets.

This infrastructure is a critical enabler as we work with automotive and mobility partners to bring advanced driver assistance and automation into real-world use. It allows us to iterate quickly while maintaining the levels of reliability and safety required for deploying embodied AI in consumer-facing applications.

Microsoft is learning from Wayve’s frontier AI workloads

Autonomous driving represents one of the most complex real-world AI challenges, combining large-scale learning, safety-critical operation, and deployment in constantly changing environments. With Wayve’s frontier AI workloads pushing infrastructure boundaries, our collaboration with Microsoft has helped meet those demands with the introduction of new Azure platform capabilities. These learnings shape the compute platforms required to deliver production-ready autonomous driving systems at scale.

Bringing autonomy safely to everyday consumers

Ultimately, the goal of autonomous driving is not technical achievement alone, but safe and dependable systems that people can trust in everyday use. Reaching that goal requires embodied AI designed for production from the outset, capable of operating reliably across vehicle platforms, geographies, and real-world conditions.

Alex Kendall, Co-founder and CEO of Wayve:
“Autonomous driving ultimately succeeds or fails on trust. Building AI that behaves predictably and can be deployed safely at scale requires serious AI infrastructure, and working with partners like Microsoft helps ensure we have the compute and systems needed to bring these technologies into real-world use.”

Dayan Rodriguez, Corporate Vice President, Manufacturing and Mobility, Microsoft:
“Autonomous driving is one of the most demanding real-world tests of AI. Wayve is pushing the boundaries of embodied intelligence, and Azure is built to support that level of complexity at scale. Together we’re helping turn advanced AI research into production-ready systems that are safe and ready for everyday use.”

Learn more about the Wayve and Microsoft collaboration in a feature on Microsoft Source.