Autonomy for Any Vehicle, Anywhere

Last week, we announced $1.5 billion in new capital to accelerate the global deployment of Wayve’s embodied AI platform.

This includes the closing of a $1.2 billion Series D led by Eclipse, Balderton and SoftBank, with participation from Microsoft, NVIDIA and Uber, alongside investments from leading global automakers Mercedes-Benz, Nissan and Stellantis. Uber has also committed additional capital to support multi-year robotaxi deployments on its network, beginning in London in 2026 and expanding to more than 10 cities globally.

This funding round marks our transition from proving end-to-end AI works to deploying it at a global scale across passenger vehicles, ride-hailing fleets and global vehicle platforms.

In our first pitch (slides below), we set a simple ambition: Wayve will be the first to deploy autonomy across hundreds of cities.

At the time, few believed that to be possible. Today it is within reach.

Solving the Hardest Problems First

When Wayve was founded in 2017, the prevailing approach to autonomy narrowed the problem.

Start with simple environments. Rely on high-definition maps. Layer expensive custom hardware and rule-based systems to manage edge cases. Reduce uncertainty by constraining the world.

This approach enabled early driverless demonstrations. It also embedded structural limits on global scale.

Wayve chose a different path.

We focused on the core intelligence problem: teaching an AI system to make safe decisions in the open world. To operate through ambiguity, unpredictability and the long tail of human behaviour.

Instead of engineering around complexity, we chose to embrace and learn from it.

We pioneered end-to-end learning when it was widely dismissed. We built world models years before they became fashionable. We prioritised generalisation across many environments over driverless optimisation in a single domain. We started in dense urban environments because they represent the full spectrum of real-world complexity. If a system can drive in London, it can drive anywhere.

Almost everyone thought we were crazy. Solving the hardest problems first forced us to prioritize independent thinking, novel research paths from prevailing industry wisdom, and a level of persistence that we continue to emphasize at the company today.

The trade-off was clear: If you design for narrow uses first, you inherit constraints later. If you solve the hardest problem first, scale is baked into the system.

That principle shaped Wayve from day one.

Bravo to early investors, like Michael Dempsey at Compound, Seth Winterroth at Eclipse and Suranga Chandratillake at Balderton, who shared that conviction when it was far from consensus.

Today, that contrarian technical foundation enables us to chart a contrarian business model for autonomy. This model is only available because we designed our autonomy to generalise.

Three Paths to Autonomy

Autonomous driving is almost always framed as a single race. 

In reality, three structural models are emerging.

1. One ties autonomy to selling your own cars.
2. One ties autonomy to operating robotaxi fleets you build and control.
3. One builds autonomy as a software platform that runs across all vehicles, brands and geographies.

Each is viable, but their total addressable markets are different by orders of magnitude.

If autonomy is tied to selling your own cars, scale is limited to your production volume. If autonomy is tied to robotaxi fleets you build, own and operate, expansion progresses city by city.

If autonomy runs as software on standard hardware, improves through data and generalizes to new environments, scale is unconstrained.

Wayve has chosen the third model.

We license our AI Driver as vehicle-agnostic software that runs on onboard compute and embedded sensors. It does not rely on high-definition maps. Automakers retain control of their brand and driving experience while benefiting from shared intelligence in a continuously improving foundation model. Every mile driven strengthens the system for every vehicle running our software.

Scale is built into our architecture. And architecture determines economics.

Systems that depend on bespoke hardware, heavy infrastructure or operationally intensive rollout models carry structural cost burdens. Every new platform requires a redesign. Every new city requires engineering. Scaling becomes slower and more capital-intensive.

A software-first platform inverts the equation.

Low incremental hardware cost. Compatibility across brands. Continuous improvement through fleet learning.

This contrarian business model is only possible now because our AI generalises to new cities, vehicles and applications. This creates a unique flywheel, further deployment leads to better results at a cheaper price. Compounding economics. High margin software revenue. AV2.0 is a structural shift in how autonomy is built and deployed.

A Decade of Compounding Advantage

For nearly ten years, we have built a series of technical firsts around this architecture:

• First to put deep reinforcement learning on an autonomous vehicle (2017).
• First to integrate a world model directly into a driving system (2018).
• First to test a model trained only in simulation in the real world (2018).
• First to demonstrate end-to-end AI urban driving on unseen roads using only cameras (2019).
• First to run a multi-city generalisation test across 10 cities (2021).
• First to generalise across fundamentally different vehicle platforms (2022).
• First to build a generative AI foundation model built for driving (2023).
• First to introduce a vision-language-action driving model, LINGO (2023).
• First to show a model trained in London can generalise to driving in the US (2024).
• First and only to drive zero-shot in more than 500 cities across three continents (2025).

These milestones reflect a consistent thesis. We have not pivoted to this view. We have no technical debt to write off. We built AV2.0 from day one.

From Research Leadership to Global Deployment

Our first decade was proving the architecture. The next decade focuses on deployment at global scale.

With this Series D, we are accelerating the next phase of firsts:

• First to integrate autonomy into OEM production vehicles without retrofitting hardware
• First to deploy autonomy with multiple global OEM partners at the same time
• First to generalise autonomy across different sensor and computing architectures
• First to deploy robotaxi services in multiple countries
• First to deploy autonomy in 100 cities

The industry is shifting. Automakers are converging around software-defined vehicles. Foundation models are redefining what AI can do in the physical world. The economics of autonomy are becoming clearer.

The question is no longer whether autonomy works, but how it scales. This makes a platform model necessary, and this is the system we’ve built.

If you are an automaker integrating embodied AI across your portfolio, build with us. If you are a mobility platform scaling autonomy, deploy with us. If you are an engineer or researcher defining the next era of mobility, join us.

Autonomy available to any vehicle, anywhere.