U.K.-based autonomous vehicle startup Wayve started life as a software platform loaded into a tiny electric “car” called Renault Twizy. Festooned with cameras, the company’s co-founders and PhD graduates, Alex Kendall and Amar Shah, tuned the deep-learning algorithms powering the car’s autonomous systems until they’d got it to drive around a medieval city unaided.
No fancy lidar cameras or radars were needed. They suddenly realized they were on to something.
Fast-forward to today and Wayve, now an AI model company, has raised a $1.05 billion Series C funding round led by SoftBank, Nvidia and Microsoft. That makes this the U.K.’s largest AI fundraise to date, and among the top 20 AI fundraises globally. Even Meta’s head of AI, Yann LeCun, invested in the company when it was young.
Wayve now plans to sell its autonomous driving model to a variety of auto OEMs as well as to makers of new autonomous robots.
In an exclusive interview, I spoke to Alex Kendall, co-founder and CEO of Wayve, about how the company has been training the model, the new fundraise, licensing plans and the wider self-driving market.
(Note: The following interview has been edited for length and clarity)
TechCrunch: What tipped the balance to attain this level of funding?
Kendall: Seven years ago, we started the company to build an embodied AI. We have been heads-down building technology […] What happened last year was everything really started to work […] All the elements that are required to make this product dream a reality [came together], and, in particular, the first opportunity to get embodied AI deployed at scale.
Now production vehicles are coming out with GPUs, surrounding cameras, radar and, of course, the appetite to now bring AI onto, and enable, an accelerated journey from assisted to automated driving. So this fundraise is a validation of our technological approach and gives us the capital to go and turn this technology into a product and bring it to market.
Very soon you’ll be able to buy a new car and it’ll have Wayve’s AI on it […] Then this goes into enabling all kinds of embodied AI, not just cars, but other forms of robotics. I think what we want to achieve here is to go way beyond where AI is today with language models and chatbots. To really enable a future where we can trust intelligent machines that we can delegate tasks to, and of course, they can enhance our lives. Self-driving will be the first example of that.
How have you been training your self-driving model these last couple of years?
We partnered with Adsa and Ocado to collect data to trial autonomy. That’s been a great way for us to get this technology off the ground, and it continues to be a really important part of our growth story.
What is the plan around licensing the AI to OEMs, to automotive manufacturers? What will be the benefits?
We want to enable all the auto manufacturers around the world to work with our AI, of course, across a wide variety of sources. More importantly, we’ll get diverse data from different cars and markets, and that’s going to produce the most intelligent and capable embodied AI.
Which car makers have you sold it to? Who have you landed?
We’re working with a number of the top 10 automakers in the world. We’re not ready to announce who they are today.
What moved the needle for SoftBank and the other investors in terms of your technology? Was it because you’re effectively platform-independent and every car will now sport cameras around it?
That’s largely correct. SoftBank has publicly commented on their focus on AI and robotics, and self-driving [tech] is just the intersection of that. What we’ve seen so far with the AV 1.0 approaches is where they throw all of the infrastructure, HD maps, etc., in a very constrained setting to prove out this technology. But it’s a very far journey from there to something that’s possible to deploy at scale.
We’ve found that — and this is where SoftBank and Wayve are completely aligned in the vision for creating autonomy at scale — by deploying this software and a diverse set of vehicles around the world, millions of vehicles, we can not only build a sustainable business, we can also get diverse data from around the world to train and validate the safety case to be able to deploy AV at scale through “hands off, eyes off” driving around the world.
This architecture operates with the intelligence onboard to make its own decisions. It’s trained on video as well as language, and we bring in general-purpose reasoning and knowledge into the system, too. So it can deal with the long-tail, unexpected events that you see on the road. This is the path we’re on.
Where do you see yourself in the landscape at the moment in terms of what’s deployed out there already?
There have been a bunch of really exciting proof points, but self-driving has largely plateaued for three years, and there’s been a lot of consolidation in the AV space. What this technology represents, what AI represents, is that it’s completely game-changing. It allows us to drive without the cost and expense of lidar and HD. That allows us to have the onboard intelligence to operate. It can handle the complexities of unclear lane markings, cyclists and pedestrians, and it’s intelligent enough to predict how others are going to move so it can negotiate and operate in very tight spaces. This makes it possible to deploy technology in a city without causing angst or road rage around you, and to drive in a way that conforms with the driving culture.
You did your first experiments back in the day, peppering the Renault Twizy with cameras. What’s going to happen when car manufacturers put lots of cameras around their cars?
Car manufacturers are already building vehicles that make this possible. I wouldn’t name brands, but pick your favorite brand, and particularly with the higher-end vehicles, they have surround cameras, surround radar and an onboard GPU. All of that is what makes this possible. Also, they’ve now put in place Software Defined Vehicles, so we can do over-the-air updates and get data off the vehicles.
What’s been your “playbook”?
We built a company that has all the pillars required to build. Our playbook has been AI, talent, data and compute. On the talent front, we’ve built a brand that’s at the cross-section of AI and robotics, and we’ve been fortunate enough to bring some of the best minds around the world to come work on this problem. Microsoft’s been a long-standing partner of ours, and the amount of GPU compute they’re giving us in Azure is going to allow us to train a model at the scale of something that we haven’t seen before. A truly enormous, embodied AI model that can actually build the safe and intelligent behavior we need for this problem. And then Nvidia, of course. Their chips are best-in-class in the market today and make it possible to deploy this technology.
Will all of the training data you get from the brands you work with be mixed together into your model?
That’s right. That’s exactly the model we’ve been able to prove. No single car manufacturer is going to produce a model that is safe enough on their own. Being able to train an AI on data from many different car manufacturers is going to be safer and more performant than just one. It’s going to come from more markets.
So you’re effectively going to be the holder of probably the largest amount of training data around driving in the world?
That’s certainly our ambition. But we want to make sure that this AI goes beyond driving — like a true embodied AI. It’s the first vision-language-action model that’s capable of driving a car. It’s not just trained on driving data, but also internet-scale text and other sources. We even train our model on the PDF documents from the U.K. government that tell you the highway code. We’re going to different sources of data.
So it’s not just cars, but robots as well?
Exactly. We’re building the embodied AI foundation model as a general-purpose system trained on very diverse data. Think about domestic robotics. The data [from that] is diverse. It’s not some constrained environment like manufacturing.
How do you plan to scale the company?
We continue to grow our AI, engineering and product teams both here [in the U.K.] and in Silicon Valley, and we just started a small team in Vancouver as well. We’re not going to ‘blitzscale’ the company, but use disciplined, purposeful growth. The HQ will remain in the U.K.
Where do you think the centers of talent and innovation are in Europe for AI?
It’s pretty hard to look anywhere outside London. I think London is by far the dominant place in Europe. We’re based in London, Silicon Valley and Vancouver — probably in the top five or six hubs in the world. London has been a great spot for us so far. We grew out of academic innovation in Cambridge to begin with. Where we are now to the next chapter is somewhat a road less well-trodden. But in terms of where we are now, it’s been a brilliant ecosystem [in the U.K.].
There are a lot of good things to be said around corporation, law and tax. On the regulation front, we’ve worked with the government for the last five years now on new legislation for self-driving in the U.K. It passed the House of Lords, it’s almost through the House of Commons, and should soon come into law and make all of this legal in the U.K. The ability for the government to lean into this to work with us […] we’ve really worked in the weeds for that and had over 15 ministers visit. It’s been a really great partnership so far, and we’ve certainly felt the support of the government.
Do you have any comments on the EU’s approach to self-driving?
Self-driving is not part of the AI act. It’s a specific vertical and should be regulated with subject matter experts and as a specific vertical. It’s not some uncoordinated catch-all, and I’m glad about that. It’s not the fastest way to innovate in specific verticals. I think we can do this responsibly by working with specific automotive regulatory bodies that understand the problem space. So sector-specific regulation is really important. I’m pleased the EU has taken that approach to self-driving.
Kommentar