Wayve is betting that the same artificial intelligence technology powering self-driving cars can teach robots to manipulate objects and perform complex physical tasks. The London-based autonomous vehicle company has appointed Alex Toshev, who previously led multimodal foundation model research at Apple, as Research Director at Wayve Labs to lead a new robotics team. This move signals that Wayve views autonomous driving not as its ultimate destination, but as an early proving ground for broader physical artificial intelligence.

Why Is Wayve Expanding Into Robotics Now?

Wayve Labs, the company's frontier research unit launched in June 2026, is establishing a dedicated robotics team in Wayve's Sunnyvale office to develop what the company calls "foundation intelligence for robot manipulation". Rather than folding robotics work into its existing autonomous vehicle division, Wayve is keeping the effort separate, suggesting the company wants room to pursue longer-term foundational research without competing directly against near-term deployment pressures from automotive partners like Nissan, Stellantis, and Uber.

The robotics team will combine Wayve's foundation-model platform and infrastructure with a new technical direction spanning hardware, software, data collection, model development, training, evaluation, and deployment across different physical embodiments. This approach mirrors how Wayve has approached autonomous driving: building a general-purpose AI system that learns from real-world data rather than hand-coding specific behaviors for each scenario.

What Makes Toshev's Background Relevant to This Effort?

Toshev brings two decades of experience at the intersection of AI and physical systems. At Apple, he co-led MM1, the company's multimodal large language model family, which can process and understand both text and images simultaneously. He also led research on generalist agents that interact with applications through graphical interfaces. Before Apple, Toshev spent more than a decade at Google, including six years in Google Robotics, where he co-led SayCan, a project that introduced large language models into robotic planning and won the Best Innovation Paper at CoRL 2022, a top robotics conference.

"Toshev brought exceptional expertise across computer vision, machine learning and robotics to build a new effort at the intersection of robotics and foundation models," said Jamie Shotton, Chief Scientist at Wayve.

Jamie Shotton, Chief Scientist at Wayve

His SayCan research is particularly significant because it demonstrated how large language models, the same technology underlying ChatGPT and other AI assistants, could help robots understand and execute complex multi-step tasks. This work proved that foundation models could translate human instructions into robotic actions, a capability that could extend far beyond autonomous vehicles.

How Does This Reflect a Broader Shift in AI Strategy?

Wayve's robotics hire reveals a strategic reframing of the company's identity and ambitions. Rather than positioning itself primarily as an autonomous driving specialist, Wayve is increasingly presenting itself as a foundation-model business whose underlying technology can be applied to multiple physical domains. This positioning suggests that autonomous vehicles are one application of a more general technology stack, not the end goal.

The decision to recruit from Apple and Google's robotics programs brings validated large-scale deployment experience to Wayve. Toshev's work on SayCan and generalist agents at Google Robotics gives Wayve access to research already proven at major technology companies, potentially accelerating a robotics effort that would otherwise take longer to build from scratch. Rather than starting from zero, Wayve is acquiring talent and knowledge from companies that have already invested billions in robotics research.

Steps to Understanding Wayve's Expansion Into Physical AI

• Foundation Models as Universal Tools: Wayve is leveraging the same multimodal AI systems that power autonomous driving, which can process images, text, and sensor data, to teach robots how to manipulate objects and perform complex tasks in the physical world.
• Separating Research From Commercial Pressure: By establishing a dedicated Sunnyvale-based robotics team rather than folding it into existing autonomous vehicle operations, Wayve creates space for longer-term foundational research without competing against near-term deployment deadlines from automotive partners.
• Recruiting Proven Talent From Tech Giants: Hiring leaders from Apple's multimodal AI research and Google's robotics division brings validated expertise and accelerates development timelines by leveraging research already proven at scale by major technology companies.
• Expanding Beyond Automotive: The robotics initiative signals that Wayve views autonomous driving as one application of a broader physical AI platform, positioning the company to eventually serve industries like manufacturing, logistics, and construction.

The timing of this hire also reflects broader industry trends. As autonomous vehicle deployment accelerates with companies like Waymo and Cruise expanding their robotaxi fleets, the competitive advantage increasingly shifts toward companies with more general-purpose AI systems. A foundation model that can be adapted to multiple physical tasks, from driving to manipulation to navigation, offers more long-term value than a system optimized for a single application.

Wayve's move suggests that the next frontier in AI is not just autonomous vehicles, but autonomous agents that can perform a wide range of physical tasks. By hiring Toshev and building a dedicated robotics team, Wayve is positioning itself to compete not just with other autonomous vehicle companies, but with robotics startups and tech giants investing in general-purpose physical AI. The company's automotive partnerships with Nissan, Stellantis, and Uber provide both revenue and real-world data that can accelerate the development of these broader capabilities.