From Trucks to Robotaxis: Waabi’s One Driver with CEO Raquel Urtasun

And we scientifically can prove this. Zero days. We go from years to zero. It's very clear that if you build it, they will buy it. If you build a technology that fits the use case, fits what they need. Physical AI is 85%... touches 85% of the global GDP.

I'm doing fantastic and really a pleasure to be here with you, Harry, today.

So originally I'm from Spain. But, you know, I left Spain 27 years ago. So, I like to say I'm a citizen of the world, proud Canadian as well. But yeah, I can walk you through the background if you want in terms of what led me to Waabi.

Yeah. So I spent the last... I'm going to age myself here, but 27 years at the forefront of AI innovation. And you know, quickly, innovation was very important at the time because AI was not what it is today, right? There was a lot of technology that needed to be developed in order to really enable the transformation that we see today. Prior to Waabi, I really got exposed to self-driving I think 16 years ago as an application of the kind of technology that I was developing and fell in love with the problem. It had everything that I was looking for. On one side the difficulty as a scientist, right? I was attracted to really solving fundamental problems, the amazing benefits that can bring to society and obviously I think many of us have unfortunately somebody close to us that has had a traffic accident. So I was very motivated to really bring safety to our roads. And you know, then it became not just an exciting problem to work on and everybody will tell you this is an absolute obsession 24/7.

Yeah, there is definitely crossover. My background is in engineering as an undergrad but my background is really from AI innovation. And the reason why this is important is that if you have a lens into how to solve from an AI perspective, you actually build very different technology. And my core background really explains also the differentiation with respect to everybody else where our competitors really come from a robotics engineering background where the technology stack is very different. Where you basically try to divide a complex problem into small sub-problems and you try to engineer that versus for me it's all about how can you build a single AI system that can truly generalize that humans do and solve very complex tasks also beyond transportation. Really it's a physical AI platform with what Waabi is building.

Yeah. So if we rewind to I guess 2017, the industry looked very different at the time, right? Everybody thought technology is here, it's just a matter of engineering it and then you can deploy massive, you know, larger scale. And my experience at Google but also looking at everybody in the industry really led me to believe that we needed a better way. We needed to totally reimagine the way that we approach this problem. And start from a solution that could truly scale versus try to engineer bits and pieces and then try to scale from that. And I left Uber because I had a very different kind of technology stack in mind. This idea, this is, you know, pretty moment- I think that's important to put in perspective. But as an AI scientist it was clear to me that end-to-end was going to be the way in terms of a single neural network that can solve this problem. But simply a black-box architecture will not be enough. We needed to build systems that can truly reason so that they are verifiable, so that you can actually bring technology to market that you can prove is safe before you deploy, etc. So that was very important for me in terms of this will become key. And the second piece was about, you know, driving miles will never be enough no matter how big your fleet is you're never going to observe particularly the safety critical situations, the accidents where you need to learn from and it's unethical to create those, right? So the second core thesis was that it's fundamental to have a different data approach and simulation will be the way. And that one can build this new concept of world model, right? Waabi was the first world model, right? That you can build a simulator that is as realistic as the real world in a way that a simulator is also verifiable. And if with these two core pieces of technology then you will actually finally build a solution that can truly scale. And we've seen now five years after starting the company is that the benefits of this technology are amazing and how it's off the charts when you compare to what the other generation of technology can do.

Yeah. So I would say that we proved that AI-first, end-to-end is very important. We proved that simulation is very important. And you hear a lot of narratives in the industry today about that because there is no way around not having something that you can talk about those topics regardless of what your technology really is. And I think the technology that our competitors have is very different from what Waabi has. And they're still rule-based. If you look at tracking, our competitors, and the simulation systems that they utilize are still very primitive. And nobody has been able to prove scientifically, mathematically, that the simulator is as realistic as the real world. And I challenge everybody. We actually put a white paper out, we show all our methodology, we show our results with the spirit of really transparency to really move the industry forward. We all agree that simulation is important but if your simulator is not the same as the real world, it doesn't matter how much simulation you do. It's a good tool but it's not a tool where you can actually utilize for your safety case. And at the end of the day this is about bringing safe technology to market and that's a must-have. So definitely as I say, challenge everybody to show us your results and your simulator with numbers at the end of the day.

Yeah, I would say that the company today, the DNA is still... we have similar kind of values that came from the origins. Things like the really scientific rigorous mindset, being safety first, being data-driven, being very humble, being evidence-based and things like this will still permeate, are still there front and central for the company. But it's at the same time very different. We passed from an R&D phase that is over to a full commercialization phase that we are in today. And it's really exciting to really see the evolution of the company. Today we have a massive pipeline of customers that really cannot wait to get access to Waabi technology and trucks. We have commercial operations for the last four years in Texas. And we have a fantastic track record of always 100% on-time delivery to all of our customers together with no safety incidents whatsoever as we do commercial operations. So I will say that it's been really exciting to see the evolution of the company and as I said that R&D phase is in the past and now is where that commercialization and truly scale is happening as we speak.

Yeah. So there are three axes of scalability that are tremendously important from a commercial perspective. This is not just about a science experiment or technology. It is truly a massive commercial advantage. So let me walk you through these three things. On one side is about scalability or generalization across environments. What do I mean by this? If you look at the industry, it has been built on the premise for trucking that is a hub-and-spoke model. You have these hubs next to the highway and then between hubs freight will be moving autonomously and then humans will do the end part of the journey, going from those hubs to the end customer. And that was done in order to simplify the autonomy problem, right? Because for class 8 trucks, tractor-trailers, historically has been very difficult to handle any surface streets. So with this technology, we proved already two years ago that we can actually drive in generalized surface streets. And what that means is that we can move freight the way that freight moves today with humans. There is no need for terminals. We can do door-to-door for the customers. And we've seen that to be extremely well-received by the customers, which is, this is what I need, I don't like terminals. This is the solution that I've been waiting for, I want to adopt your solution. And that has been unanimously across the majority of the Fortune 500 companies that we see. So that's been great to see in terms of as it relates to environment. Another aspect that is very important is kind of this ability to scale or generalize across what we call embodiment or platforms. What do I mean by this? When you look at customers, they typically have a mixed fleet of different OEMs. When you look at geographically how trucks look like in Europe, in Asia versus North America, they are very different in shape. So what Waabi has is this ability to generalize from one embodiment to a totally different embodiment. And we proved this recently, we showcased this to the public, which is that without requiring any engineering, without requiring any retraining, without even requiring simulation, we can actually train on one platform and it analyzes perfectly from day one, nothing to be changed on a totally different type of truck with different dynamics, with different sensors, with a different shape. That is massive when you think about being able to penetrate globally the markets as well as go across OEMs.

Yeah. To Volvo VNL. That's absolutely... these are completely... and for someone like me, not a trucking expert, these are two different trucks, right? Totally different trucks. Humans can generalize. No autonomy stack can do that. But that was a massive proof.

The last bit I will say is that it relates to geographies, right? That's typically also a lot of what we talk about. And this is this ability that our customers have, either nationwide networks, most of them, the majority of them, or even international. So it's very important for them that it's not just about Texas, it's about across all the different geographies including all the surface streets that they have. So what Waabi can do is actually open new lanes nearly instantaneously.

From what we've seen, two years typically to actually do so. And we can prove that we can do this also zero-shot, which is amazing when you think about the scalability and what the customers can actually get. So if you take all these different pieces into account, is amazing where we are in terms of providing a much better product that is more reliable, safer as well, can scale much better.

Yeah. So we have a safety driver on board today. What we are waiting for is for the final validation of the Volvo platform. So just for context here, we have a partnership, a fantastic partnership with Volvo where they're actually providing a special type of truck that has already the redundancies so that we can actually launch a driverless product. And then Waabi provides the sensors, the compute as well as obviously all the AI that drives the trucks. So we believe, the same as all the OEMs out there, that in order to launch safely a level 4 driverless truck, you need to do it with a fully validated platform by the OEM. Right? They've been building this for 100 years. They do this extremely safely, robustly. And we are waiting for Volvo to finish their validation. They have now publicly announced when that's going to be which is Q1 of 2027 and then you will see us launching driverless because the technology from Waabi's perspective is ready to get us there.

So as I mentioned, if a fully validated truck was ready today, then Waabi will be launching driverless much earlier than Q1 of 2027. So we're waiting for that final validation. And this is very important because we do not compromise safety as some of our competitors are doing very differently. They are putting off-the-lot retrofitted trucks that haven't been fully validated. We don't believe that's safe. Waabi will not do that. Waabi is happy to wait for the OEM. We don't have pressure from our investors, etc. Instead we are waiting to do this the right way with a solution that really scales because at the end of the day you don't want to scale with something that is not safe.

Yeah. So as I mentioned, the R&D has been really done. We are in the... 2027, I mentioned Q1 '27, that's when we'll be ready. That's where you will see fully driverless operations with Waabi. And the good news of really having the scalable approach that I mentioned is that we can scale super, super fast. Right? And you will see a massive hockey stick already in '27 with Waabi. And that's what truly matters is when you can really provide a product that both customers as well as your ability to really achieve the goals that they have and then provide what they truly need. So it's going to be really quick. So we're waiting for that validated platform and then scale happens almost immediately after.

Yeah, fantastic question. So in terms of scale availability, what you're going to see is quite a few customers. There is definitely some customers that are much more advanced than others in terms of when can I have access to your trucks, I bought them yesterday. This is really the ROI is so massive for the customers when you think about utilization, almost 3x utilization of the assets, when you think about safety, etc. So there is no... the demand is not the issue at all. So the first two to three years is going to be mostly bound by the abilities of the OEM to produce trucks. And it's not going to be by Waabi's technology. It's not going to be by the customer demand. Now as I mentioned, Volvo will produce this out of the factory line. So that actually gives you the scale. So that scaling is going to be, I would say, in the industry faster over the next two to three years than a lot of the models that people have put out of the industry. So it's coming now and it's coming very fast from that perspective. But once more, I think this is very important is that that ability to scale across geographies, that ability to drive on surface streets and get to the customer locations, end points, is a must for being able to scale. By doing a few routes, it's going to be really hard to scale. To scale you're going to saturate those routes. So that's where you will see Waabi is scaling much faster than some of the competitors. All competitors, by the way.