WAYVE

Wayve started in Cambridge in 2017, out of the university's engineering department. Two researchers, Alex Kendall and Amar Shah, had a heretical idea.

Every other self-driving company was bolting on lidar sensors and building giant high-definition maps of every street. Kendall and Shah thought that was a dead end.

Humans drive with two eyes and a brain, not a laser scanner and a pre-loaded map. So they built a car that learns to drive from cameras and deep learning alone.

In 2019 they put a car on the streets of Cambridge that had never seen the roads before and let the AI figure it out. That demo became the company.

Wayve does not want to build cars or run a taxi fleet. It wants to sell the driver.

The plan is to license its self-driving AI software to carmakers, who put it into their own vehicles. Think of Wayve as the brain and the automakers as the body.

Because its system learns from camera data instead of relying on hand-built maps, the pitch is that it can work anywhere, on any car, without remapping every new city. The customer is the car company.

The product is an AI that already knows how to drive.

Wayve's product is its self-driving AI, an end-to-end system that takes in camera views and outputs driving decisions. The company calls its approach Embodied AI.

It has also built driving foundation models, including work it named GAIA, that can imagine and generate realistic driving scenes to train the system on rare situations. There is no consumer gadget to buy.

The deliverable is software that a carmaker installs so its vehicles can drive themselves, first as advanced assistance and eventually as full autonomy.

The bet is called AV2.0, and it is deliberately contrarian. Old-school self-driving, AV1.0, leans on lidar and detailed maps that cost a fortune to build and update.

Wayve skips all of it and trains one AI model on huge amounts of driving video, so the system generalizes to new roads on its own. It is basically the Tesla approach, but sold to any carmaker instead of one.

The proof points are the demos. Wayve has driven its AI cars around UK cities and even dropped them into new countries to show the model adapts.

Working footage, not slide decks, is what pulled in SoftBank and Nvidia.

Self-driving is where ambitious startups go to burn billions and miss deadlines. Wayve is no exception to the risk.

Its learning-based approach is elegant, but a neural network that drives is hard to fully explain or guarantee, and regulators want guarantees. Rivals like Waymo have spent more than a decade and enormous sums and still operate in limited areas.

Wayve has raised over a billion dollars and now has to turn clever research into a product carmakers will actually ship. The hardest question in the field still hangs over it.

When is a car that taught itself to drive safe enough to trust with no human?

Wayve's cap table got serious fast. Eclipse Ventures led the $20 million Series A in 2019, then co-led the $200 million Series B in 2022 with D1 Capital, pulling in Microsoft and Baillie Gifford.

Then came the headline. In May 2024, SoftBank's Vision Fund led a $1.05 billion Series C, with chipmaker Nvidia joining as a new backer and Microsoft returning.

It was the largest AI funding round ever raised by a UK company. Uber later added a strategic investment.

When SoftBank, Nvidia, and Microsoft all write checks into the same car, the industry pays attention.