NVIDIA’s Grand Robotaxi Ambitions Are Coming Into Focus

NVIDIA’s open-source Alpamayo 2 Super, a 34-billion parameter VLA foundation model for Level 4 autonomy is framed not as a product launch but as a platform capture event designed to lock the entire AV distillation pipeline to NVIDIA’s Drive Thor silicon and DGX Hopper compute infrastructure.

WeRide formally entered Denmark through a partnership with Green Mobility, targeting public driverless service in the first half of 2027, while simultaneously scaling its domestic Chinese robotaxi fleet toward 3,500 units by year-end.

Aurora reported $2 million in revenue and a $1.2 billion liquidity reserve, but was forced to sidestep PACCAR’s restrictions by retrofitting International trucks via Roush, introducing fragility beneath its headline commercialization progress.

Key Autonomy Signals Episode Questions Answered

Autonomy Signals Topics & Timestamps

[0:00] KPMG Sponsor Introduction

KPMG works across the full autonomy ecosystem, advising operators, OEMs, suppliers, insurers, and investors as autonomous mobility scales.

[01:11] Signal 1: NVIDIA Open Sources Alpamayo 2 as their Grand Robotaxi Ambitions Come into Focus

NVIDIA released Alpamayo 2 Super, an open source 34 billion parameter foundation model that gives away the software to sell the silicon, anchoring the entire distillation pipeline to Drive Thor and NVIDIA compute. A Delaware Chancery Court filing revealing NVIDIA’s bid for Zoox raises the question of whether a Foxconn and Taiwan Semi partnership to build a bespoke NVIDIA robotaxi is next.

[40:16] Signal 2: WeRide Expands into Denmark with GreenMobility

WeRide partnered with GreenMobility to deploy its EU-compliant GXR robotaxi in Denmark with public service targeted for the first half of 2027. The asset-light entry is the Autonomous Belt and Road Initiative in action, using Chinese domestic cash flow to finance European expansion while WeRide simultaneously scales past 1,000 robotaxis at home.

[1:01:25] Signal 3: Aurora Sidesteps PACCAR with Roush and International

Aurora turned to Roush to retrofit International trucks after PACCAR refused to allow driverless operations, targeting 20 trucks per week to reach 200 driverless trucks by year-end. With Kodiak moving towards Daimler Truck and Bot Auto already there, the entire autonomous trucking industry is sidestepping PACCAR’s factory line.

Full Episode Transcript

NVIDIA Alpamayo 2 Super: Open-Source VLA Model Explained

KPMG Introduction: The autonomy economy is real. Commercial robotaxi operations, humanless freight runs, AV infrastructure investment measured in billions. The question is no longer whether autonomous mobility scales, it’s who advises the companies building it. KPMG works across the full autonomy ecosystem: operators, OEMs, suppliers, insurers, and investors. At the moment it matters most, when the road ahead is uncertain, we’ve already been there. KPMG, accelerating what’s next in mobility.

Grayson Brulte: Rob, there’s a lot of good signals in the market. The signals this week are NVIDIA has grand ambitions, and we’re gonna get into what those ambitions are and why you should pay attention. WeRide continues their European expansion, this time into Denmark. And Aurora, they did the two-step, or you wanna call it the sidestep, and avoided the factory line to go to upfitting, and there’s a reason why that’s a signal, and we’ll get into that. Autonomy Signals, as always, is brought to you by KPMG. KPMG, thank you for being a wonderful, great sponsor and making this show possible. Rob, let’s dive into the signals with NVIDIA. Jensen made a post. We said we have an index. What did we learn from the post that Jensen made?

Rob Grant: Yeah. So big news, right, this week out of NVIDIA. they launched what they’re calling their Alpamayo 2 Super an open source, 34 billion parameter hybrid open reasoning vision language action foundation model designed for level four robotaxis and advanced autonomous systems. This architecture pairs a 32 billion parameter NVIDIA Cosmos 3 Super Reasoner backbone, which processes visual context, natural language, and multi-camera inputs with a specialized two billion parameter diffusion-based action expert that outputs precise vehicle trajectories. So rather than relying solely on passive imitation learning, the model undergoes reinforcement learning post-training at GPU scale to evaluate the simulated consequences of driving decisions before execution. th-this is some fascinating stuff. So in, in short, what Alpamayo 2 is, it consolidates fragmented AV pipelines, traditionally split into separate perception, prediction, planning, and labeling models into a unified multitask framework, right? It uses 360-degree visual perception, synthesizing simultaneous inputs from seven or more cameras for continuous real-time spatial awareness. It provides chain of causation reasoning, generates step-by-step interpretable decision trace mapping observations to intermediate conclusions and final tactical actions. This is going to be super helpful for regulatory purposes. It also explicitly outputs high-level tactical intent, as yielding, changing lanes, or stopping alongside physical path trajectories. And it functions as a teacher model that auto labels raw driving logs and generates 2D grounded decision traces to train distilled onboard runtime So all this is really advanced. This is, you know, miles ahead of what a lot of their competitors are doing. But what’s really interesting here and why it’s such a big signal, not only for its technological but the fact they’re open sourcing all this, right? This is a huge. Right. There’s this big debate going on with all types of models and whether or not open source is the way forward. But here they’re open sourcing their front tours– fr-frontier class physical AI foundation under a permissive commercial license. and there’s a reason they’re doing it, and th-this is why it’s a And in short, and we’ll, we’ll get into this more, but they’re basically giving away the model in order to sell the silicon and the cloud And they’re trying to kind of get this lock-in for the whole upstream, downstream autonomy economy compute infrastructure layer. And So it’s a really smart use of their advancing technology, but also kind of their power position already in the autonomy It– Y- If you’re not NVIDIA and aren’t valued at whatever trillions of dollars they are, you, you may not open source But because they see it as a way to kinda lock in all of their folks who wanna use this, particularly those kind would say tier two AV developers, right? Not your pure play players like Waymo and Tesla, but your other folks. This is a way to say, “Hey, we can catch you up what is happening in the simulation and other areas super quickly if you turn to our open foundation gotta use our silicon and you gotta use our compute to do it.” super smart, super smart play right there, I think. But I’ll turn it over to you. Give me, give me your sense of, of, of what you think is– stands out to you with their open foundation

NVIDIA’s Seed-the-Farm Strategy for Physical AI

Grayson Brulte: What stands out to me, and this goes back to Jensen’s keynote at GTC, is very simple, is that Jensen believes in physical AI. Jensen believes that physical AI will be a massive growth market for NVIDIA to sell more GPUs, to sell more compute. And Jensen’s leaning into saying, “Okay, we will give you the Alpamayo 2 model for autonomous driving. Oh, by the way, autonomous driving is part of physical AI. We are going to get you into the NVIDIA ecosystem. We are gonna get you to start using the CUDA developer tools. We’re gonna get you comfortable with it.” And as you scale your business and you go, as you said, from a tier two supplier to a tier one, to a unicorn, NVIDIA’s gonna make more money. It’s really good ’cause you and I, offline, we talk a lot about baseball. Even though the Yankees in the late innings have been, ugh, choking, we, we still talk a lot about baseball. It’s that Jensen’s saying, “Okay, you’re coming out of college ball. Perhaps you didn’t play college, and you’re going into, to Single A minors. And we’re gonna start with you now, Single A minors, and, and work you all the way up to the New York Yankees so you’re Aaron Judge.” That is a fascinating, if you wanna call it a, a, a seed approach. And our friends, Danny Bernstein over at Reservoir, they’re doing all the rugged off-road physical AI for autonomy and agriculture. NVIDIA’s out there working with them. A- a- a- I’ll summarize this way: NVIDIA’s starting to seed the farm, no pun intended, from seed to table. It’s a brilliant move that I believe over time will help NVIDIA grow their automotive revenue. And job well done, Jensen. And as you said, with a $5 trillion market cap, the GPU is. They’re one of the only companies, if not the only company in the world, that could actually pull this off

Rob Grant: Yeah, and I think that’s what’s super interesting is, is kind of using its market power to set the future for continued growth, right? Yes, there is a business to be made in their frontier model, and other folks who are playing in this space have closed models that they’re trying to either license or allow folks to, to, to buy and implement in order to advance their own L4 stacks. But here, you open source it because the real value that you’re trying to get is in the compute and data flywheel. and it just– it’s a brilliant scheme. and it only works, by the way, if you’re able to develop really great technology, which we shouldn’t overlook the fact that this is a super advanced foundation and, and model here. now for folks, just so you understand, right, this is not gonna go on the edge. These are not– The Alpamayo II is, is not running on the vehicle itself. This is not an edge model. it will be distilled, right, for those who use it and work off of their open model. They will distill it and, and, and move some of the reasoning and capabilities onto the edge. and so it’s a little bit, you know, but you have to do that using the NVIDIA ecosystem, right? You’re using the Drive Thor hardware and you know, Cosmos and everything pipeline. and so it only works if you’re, if you’re using kind of NVIDIA throughout the process. And that we should recognize for those folks who follow this really closely, that’s where it’s like a little bit different than Applied Intuition, which is agnostic, right? You can, you can distill it and use any hardware or any silicon. and so that’s kind of the moat that, that Applied has, though some of what they’re open sourcing here does kind of threaten directionally what Applied is doing with their simulation and things of that nature. But they’re, they are still fundamentally different because of the way you move from the teacher-student model, which is teacher, you’re, you’re working in the cloud, all this stuff is run off the vehicle, and then you can distill it to the student, which can then run it on the edge. that, that’s, that’s a little bit different model than, than what Applied does, but it does kind of overlap with some of what Applied is doing, and Applied has been a leader in this. If you’re going to, instead of build it your own, you’re gonna buy it and build off of it model. so this is a really interesting dynamic NVIDIA starts to creep into yet another layer autonomy stack. And they are really starting to go from top to bottom, from, from, silicon all the way out to the edge as well

NVIDIA’s Bid for Zoox and Vertical Integration Ambitions

Grayson Brulte: We know from history and from a filing with the Delaware Chancery Court that Nvidia has grand ambitions for autonomy. And I say grand ambitions because during a court filing with the Delaware Chancery Court, Nvidia submitted a bid to buy Zoox. That’s right, Zoox that’s owned by Amazon. Nvidia was one of the bidders there, according to court documents. Nvidia didn’t want the vehicles. They wanted the technology to vertically integrate. All in the court docs. And that gets really interesting, and if you’ve seen what’s happened with Alpamayo Two, you said, “Okay -huh.” To me, it’s the early signals or, or we’re gonna play connect the dots, Connect Four, of all these pieces to build a vertical integration autonomous driving stack. And I believe that it’ll start with the robotaxi, and then it will go to trucks, and then it will go to other elements of physical AI. And eventually at some point, you’ll probably have a similar Alpamayo Two for humanoids. Nvidia wants to vertically integrate this to, to sell the compute and sell the hardware. Giving away the software, that only helps Nvidia. And, and when I say it only helps Nvidia, it also helps other companies in the ecosystem. You’re right about the risk to apply. You’re, you’re very right about that. But give you, one of the bi- the biggest beneficiaries of this open source movement for LLMs right now is Apple. Good luck trying to get a Mac Studio Pro with 128 gigs of RAM. It, it’s benefiting them. So Apple doesn’t have an AI strategy, sorry. Maybe you will, but right now you don’t, but they’re selling more hardware than they practically have ever sold because of the open source. So there’s going to be other beneficiaries of this open source movement as well. Perhaps one of them could be Dell ’cause they actually sell the hardware. So I think that’s an interesting thing to watch there. As Nvidia open sources it, what other companies also benefit besides Nvidia that are not autonomous driving traditional providers is gonna be something very interesting to watch over time

How Alpamayo 2 Indirectly Benefits Uber and Tier-Two AV Developers

Rob Grant: I think I fundamentally agree with you, but I, I came back to somebody within the ecosystem that this, I believe, indirectly benefits in the short term, and that’s Uber, right? So as, as we’ve talked about, Uber is relying on a fragmented market to maintain kind of its position as the demand aggregator. The more things are fragmented, the, the better that runs to Uber’s leverage and benefit for being the demand aggregator because what Uber doesn’t want is kind of domination in markets, right? We talked about this last week, whereas we see potentially Waymo dominating the U.S. market, maybe Baidu in Europe, and WeRide in the Middle East or vice versa. That’s the biggest threat to Uber’s play in the AV space is that you could wind up with these bunch of pure play dominant players, right? That, that can fund their own demand aggregation or at least only have to partner with Uber at the beginning for that demand aggregation later and then later take it over, particularly those who have had experience with consumer-facing apps so we won’t repeat all that here, but what Alpamayo 2 does is it accelerates both software iteration and lowers the capital barrier for second-tier AV developers to enter and get closer to commercial operations on L4 And that is exactly what Uber is betting on, both in terms of just its general, I think, strategy in looking at the market as a, as a landscape going forward, which is fragmentation adheres to the benefit of Uber. But also specifically, Uber is investing in– either investing in or, or y- meaning literally taking equity in or partnering with many of what we would call the second-tier AV developers, and that’s, that’s not a, a knock on any of them, right? You just have your categories of Waymo and, and Tesla kind of ahead of everybody else, and then your second-tier AV developers. But, you know, this could help materially advance others that don’t have the capital capabilities to move to L4 they now build off of a very advanced L4 model that’s open source, that could accelerate them towards being more commercial, which only helps Uber because Uber wants to see more commercial players in the AV market,

Grayson Brulte: The tier comment, for the audience that might not know, is a factual comment. If you look at the traditional automotive supply chain, you have your tier one supplier, your tier two, your tier three. All, all those companies’ supply chain, they play a critical role. So it’s just basically if you wanna call it a categorization of where the technology is or, or where your market share is. While Uber– While this is positive for Uber, yes, you are right, and, and while this can help accelerate new emerging technologies, which is great for the ecosystem as a whole, it’s great for the autonomy economy, there’s a missing piece that you and I know very well. You know it a lot better than I do, but I study it. You actually rolled up your, your, your sleeves and you worked on it: metal, cars. You can have the world’s greatest autonomous driving software, but if you can’t get the vehicles to scale it, you’ve got a problem. So while what NVIDIA is doing is very positive, the other element of this is the actual vehicles to deploy it on, so that’s gonna be something to watch

Rob Grant: It’s going to be fascinating and, you know, I think some of the things to watch for is particularly in these, I will call them more emerging vehicle manufacturers. Is NVIDIA investing in them, right? I think, I think that is something that to track because you’re right, right? There, there are kind of four key layers in all of this, right? Your, your, your software and, and hardware meaning like hardware like silicon and, and things like that. And then there’s your manufacturing capability, then there’s your infrastructure, and then there’s your operational kind of demand layer. And it appears, it appears, I think that it’s the very signal that we’re calling out here, that NVIDIA is moving into more and more of the categories, and the one that seems to be missing, as you called out, right? I say seems to be because it could be coming, is that manufacturing and again, we’re talking the world’s either first or second most valuable company depending on the day. and so– And we, we’ve seen them become very active in terms of, the private venture We see their name continually popping up in, in a lot of the companies that you and I cover. and so I think it’s gonna be really keen to read where NVIDIA is putting its money, if it’s putting its money towards

Grayson Brulte: And on that front, this goes way back to when you and I were just getting started in this industry. And I’ll throw, I’ll throw out a name from the past ’cause you probably listen, Danny Shapiro, who was the general manager of the ur- of the NVIDIA Drive program at that time. Remember that? The L2 program. And there was aspirations at that time when you and I would go to all these old, the TU conferences. Jamie did a great job with those. If you’re listening, Jamie, great job back But there was grander ambitions there. And then for everybody that knows, everybody was pretty much tested on the Lincoln MKZs back then, going back to the early days and all those tests. And there was, in my opinion, aspirations to go vertical integration. You fast-forward to today, 2026, Rob and I are old, but we’re not ready for the senior citizens home yet, but we’re getting up there in age, is does NVIDIA cut a deal with Foxconn and Taiwan Semi to build a vehicle or to license the, if you wanna call it, that entire stack? Because Foxconn has grand ambitions. They’ve tried, this is reported by the Nikkei in Japan, they have tried to buy an old Nissan plant. That gets very, very interesting to see if that happens, if NVIDIA, Foxconn, and Taiwan Semi team up together to do a b- a bespoke robotaxi powered by NVIDIA. If that happens, oh boy, watch out and buckle up

Rob Grant: Agreed. There, there, there are so many potential plays for NVIDIA in that directly, indirectly, you know, acquisition, partnership, right? We saw Uber just do this with Uber, Lucid, and Nuro, right? Where kind of everybody’s bringing something to the table. and so, you know, I can think of probably six OEMs off the top of my head, some well-known, some lesser-known where it would make a lot of sense for some type of shared partnership, equity investment ownership would really make a lot of sense for what NVIDIA is driving towards. And ultimately, right, beyond this short-term window that we’re in, say to 2030, 2031, you know, the plans for NVIDIA might be even grander than what we’re talking about. they could be a, a, a a pure competition to Uber to, to Waymo itself and things like that nature. So it, it’s gonna be a fascinating next couple of years as we have kind of very different strategic models or goals from the very big players in this Waymo’s strategic goals are different than Uber’s strategic goals, are different from NVIDIA’s strategic goals, yet there is a lot of intersection amongst all these companies, whether it’s direct investments, partnerships different plays in different markets different plays on the infrastructure side within each company. And so I think really we’re hitting the next kinda three years to the end of the decade. It’s going to be just a straight-out battle for whose vision be implemented or whose vision can be imposed. and I think we’ll see it play out in different ways and, and you see that really concretely on the policy end by what Uber’s doing on the policy end. see it in the partnership end with kinda how folks are lining up, different four layers that we’ve talked about. but you’re definitely gonna see it, out even more with some very big, very influential CEOs across all of these places as well. So I would say watch X. There’s gonna be a lot happening in that space in, in the next 36

Grayson Brulte: There’s gonna be a lot happening, and OMEGA’s gonna be there to track it all, and you and I will be here to, to analyze it with OMEGA’s help. If you’re interested, reach out to Rob and I. We have a lot of great proprietary research on Nvidia that we’re sharing with clients, so, so reach out. OMEGA did some really deep research there. ‘Cause you’re right about the grander ambitions. To me, Nvidia has grander ambitions. And I was going back and forth with the gentleman on X. He had an anonymous account, but I still responded. Don’t worry, I was polite. Is about brand value. Oh, you just can’t drop a robotaxi in any city. You drop an Nvidia robotaxi in London, San Francisco, Miami, I guarantee you they would not have a problem generating rides. Oh, Nvidia? Everybody knows Nvidia. That, that brand could do it. And when you look at the, the grander ambitions and you have the divergence which you talked about, hypothetically, Google could say, “We’re going straight TPUs.” We know that Tesla’s using custom silicon, but on the SpaceX side of things, depending on if there’s a merger, not, not commenting on that, Elon on the earnings call yesterday talked about we’re only using Nvidia, but Tesla’s using custom silicon. Th- there could be a diverging market where if you don’t have access to your own silicon, you could be left holding the bag and it’s like, okay, you have no choice, you have to go to Nvidia, but then if Nvidia says, “Okay, we’re doing this on our own,” where do you go? Then all suddenly your supplier starts competing against you. In, in automotive terms, that’s like saying, “Okay, Bosch is gonna build a car and compete against Ford or compete against GM.” It’s a, it’s a head-scratcher

Rob Grant: It is. And you’re right, right? But I, I mean, what was it today, I think it was, or yesterday, I, I think it was Anthropic was talking about building its own silicon. early on, I gave credit to, to, to Kyle Vogt. I mean, he talked about we had a project to build our own silicon at Cruise because he saw early on, this is going to be kind of the, the gatekeeper. Is your, is, is it your access to compute? and it, it is, it is real. I was just talking to somebody who works on, on world models for a different program, and they were like, “Hey, everything’s going great, except we have to be worried about our access to compute. We’re, we’re, we’re, we’re getting compute from places you wouldn’t even expect, right? ‘Cause it’s the only, only. Or we’re in line and so we’re kind of a little bit constrained from what we, what we’re capable of doing versus what we wanna do.” And, compute be so central to NVIDIA right now is at the heart of this. But others are, are, like you said, it’s a divergence and it’s a competition. And, you know, folks are in, in this space. The great thing is you don’t have linear thinkers. You have very big, big of outside-the-box how you wind up positioning yourself in a world where the resource is constrained, but the, the ambitions are not, and you have outside-the-box thinkers, sure we’re gonna see some moves that just make our jaw

Grayson Brulte: I agree with that, and we’ll be here to analyze it. And to the individual that you spoke to, I hope they didn’t dial 1-9, 1-900-COMPUTE, ’cause gosh knows where that went. That was a nine, not an eight, for those that are following my little off-the-wall sense of humor, ’cause I think outside the box. So we’re gonna watch that. And while we watch that, Rob, OMEGA highlighted some really important risks here, and let’s go through the risks OMEGA has highlighted here. First risk here: “Competitor end-to-end models, Waymo, Tesla FSD achieve native edge deployment first, undermining NVIDIA’s teacher-student positioning.” That’s true, because right now we know there’s roughly 3,800 Waymo vehicles on the market. There is roughly, let’s say, over 100 Tesla FSD robotaxis. There are millions of, of FSD supervised. The, the market’s starting to go ahead while these companies are, if you wanna call it the, the, the tier two or the, the triple A ones are s- are still heating up and haven’t had that call up yet

Rob Grant: Yeah, I think this is a, this is a really cogent risk, and it goes to the nature of what NVIDIA has released, right? NVIDIA’s Alpamayo-2 super model is not used directly on the vehicle edge as we talked about, right? Instead, it functions as a large 32 billion parameter cloud-based teacher model. So developers use it in the data center for reasoning, data generation, auto labeling, and then they will be able to distill its capability into smaller low latency models deployed on in-vehicle edge hardware like the NVIDIA drives, right? That’s, that’s the idea. Now, what the risk is saying is that competitors, Waymo and Tesla, can, can take a similarly powerful model, their own internal models, distill it down, and put it on the edge in their vehicles first, and ultimately maybe then license that as well. And so rather than even having to be locked into the NVIDIA ecosystem in order to achieve the advances that, that Alpamayo might provide to these tier two developers, they may be able to look to kind of more hardware neutral technology that they can put– Or it’s already been distilled and then licenses put on the edge. And so it’s, it’s a, it’s a fascinating competition for, one, do you see software as your potential market edge? Do you see hardware or compute as your market edge? And then do you see the ability to move more and more advanced technology onto the edge? And we talked about this when we’ve talked about Tesla, right? Part of what we think is gating for Tesla is their move to their AI five which would be more advanced, more powerful, but on the edge, allowing the vehicle to perhaps power through, literally compute its way through some of the issues that they incur when it comes to just having a camera only So it, it’s, it, it’s a great risk to call out. It calls out all the things that we just talked about, about divergent paths, similar paths, about different ways to achieve an end, and about seeing what business model works best for you and where you think their best margins are. So I think this is, this is a risk that kinda takes everything we just talked about in the previous 13 minutes here and puts it concretely into, this is a competition to get the most powerful compute onto the

Grayson Brulte: You said something interesting, hardware this is just me, I think neutrality when it comes to compute and infrastructure is going to be a big deal going forward. I’m not going to speculate on things that have been said around AI models and LLMs. You can go read all the reports that are out there publicly about using your data. There could be seepage into compute. Why do I say that? Well, Google Cloud Compute, Waymo, Amazon AWS, Zoox, SpaceX AI, Tesla. Those are the three main, outside of the hyperscalers, those are your three main things. It’s gonna be something to watch if those narratives, and I’m not saying it, it’s true or not, so I’m not saying that, let me be very clear. I’m saying if that narrative starts to take hold, that’s gonna be something to watch. If somebody comes out and positions the market that we are a neutrality, we have no ambitions for physical AI, we have no ambitions for autonomy, it’ll be something to watch because you and I have fundamental beliefs, that’s why we co-founded this company together, is that autonomy is gonna become an economy. So transparency, we’re gonna have to have OMEGA, like maybe you’ll set up the OMEGA transparency alert to see what she can, can find there, ’cause there’s, there’s, there’s just, there’s something there. What do you think?

Rob Grant: Look, our fundamental belief, as you said, is that the, the autonomy economy is just beginning. and there are some kind of, you think about it in terms of an agricultural analogy, right? Like, we went through the agricultural economy you know, 170 years ago or whatever. But there are just fundamental things that you need in order to make a farm work. And I think here, the fundamental thing that you need it may be water, it might be the right soil, right, the right climate. Here, it’s access to and that is, that is literally the, the, the, the the foundation of what drives the autonomy economy. And it’s not just robotaxis, it’s humanoids, it’s drones anything that is powered, by And so being able to either sell across all of those as a hardware-neutral platform could put you in a most dominant being, very integrated into a particular vertical. Which we not to say those verticals won’t be very, rich in in many versions of the definition of that word. but the cross-platform, cross-industry neutral

Grayson Brulte: I agree with that. And I love these farming agriculture analysis for our audience. You know I love hats, and maybe I gotta get a farming hat next and I can be on here Farmer Grayson. And so when you make a farming or an agriculture reference, I can put that on. I got the inspector hat, I got the ha- the happy clap hat, now I can have the, the farming hat. So, so stay tuned for that ’cause I do have more hat plans as we expand our shows. And let’s get on to the risk two that OMEGA’s uncovered here, ’cause this one’s a very poignant one as well. OEM partners resist replacing proven modular stacks with unproven VLA distillates, slowing commercial adoption. That’s possible, and let’s not forget, OEMs run on seven-year cycles, and there is layer after layer after layer after layer of decision-making. Some people call it bureaucracy, so it’s not just the CTOs can’t say we’re doing this. There’s a lot of cooks in the kitchen. I’ll get a chef’s hat for that one

Rob Grant: Yeah, that’s, that’s the interesting rub here is, is as much as we’ve talked about how advanced the technology is, there aren’t a lot of long history towards the, the, the kind of validation of this VLA foundation model that, that NVIDIA has just released. there’s a lot of things that make a lot of sense in it and, and offer a lot of benefits. But actually, when it comes to adoption from the OEM perspective, they like history more than they like potential future benefits. They wanna know it’s worked and what’s worked and how long it’s worked and all that kind of stuff, which is why you see kind of a, a much kind of slower movement away from, from some of the things that they’ve just been kinda grounded in for the last couple of production cycles of vehicles. which is why I think at least in the North American ones, you’ve seen China kinda just fly right by them software-defined vehicles. just been a slower adoption, a little bit more conservative approach, wariness to put new technology into the vehicles. and then, and then some weird they’ve gone into, right? Like, you know, this reluctance to use, like “I don’t, I don’t need another map.” Like I, I got plenty of maps, right? Like anyway, I, I think this is a real risk, right? And it, it– this is a risk that’s more along timeline concerns, not a, not a, like will it happen? I think a when it will happen kind of concern than will it, because it, it will need to happen, particularly if any of them have ambitions for personally

Grayson Brulte: All right, so you said maps. Now we gotta get you an explorer hat. So Rob’s gonna get an explorer hat. We’re gonna keep this this hat thing, or I’ll get you a Lewis and Clark hat. I don’t know, either one, they’re cool. Or maybe we’ll pick one up on one of our field adventures that are coming up soon. Here’s the last risk that OMEGA has uncovered before we get to OMEGA Say, ’cause this one’s, this one’s really good, really smart. “Competitors with larger real-world fleet data may outperform cosmosynthetic-trained models on edge cases.” And boy, oh boy, do we know from history how important edge cases are specific. There’s two that come to mind. That’s a very, very, probably even could be, you could say the most important risk

Rob Grant: It could be, but what’s really interesting here is that there are so many other developers in this area who are on the side of where NVIDIA’s Alpamayo is, which is that you can create synthetically defined worlds to, capture these edge cases, right? We know Wayve is in that space, right? That’s literally in, in, in Autobrains and Waabi and others, right? Whether it’s a VLA model, a world model, whatever it is, they, they are saying you don’t need to brute force your way through long-tail cases. And, you know, I, I think fundamentally this is an open question. I think Waymo fundamentally believes, yeah, you need that data, right? That also adheres to their benefit ’cause they have the most data collected of anybody. Tesla is, is, is– I think believes it doesn’t need the data, but then is actually getting the data because they’re, they’re a little bit in between with does their hardware and their silicon and everything kind of, as it stands today, is that enough to get to the safety confidence that they need for their scaling? so you know, they did mention in their call, right, they’ve driven 380,000 miles without an incident and, and that, and their real world data from their FSD, right? So th- they kind of are in between. I think they wanna be able to say it’s not as necessary to get 20 billion miles like Waymo because it will take a long time to get there. But they’re also saying, “Hey, well, you know, don’t forget, we also have this large data repository from FSD.” Whereas the Wayves, the Wabis, and others are “You don’t need that at all.” Right? Just as Alpamayo can accelerate you in this instance that we’ve talked our world models can do the same. So this is a really large question for the industry that’s– we just don’t know the answer to Do you need real world fleet data to get through the long tail of edge cases that exist or can you do it with synthetic generated data? And I think more and more, just say the– my understanding of kind of where the wind is blowing on this is more and more people believe you can do it through a VLA model or world model, that you can power your way through with

Grayson Brulte: I’ll give the cop-out. Only time will tell. There’s no other way to say it, only, on- on- only time will tell. The bottom line is we’re, we’re roo- we’re rooting for everybody, and what you and I have clearly discussed here, NVIDIA has grand ambitions, which brings us to OMEGA’s take. OMEGA’s got a, a, a spot-on take. And for the audience that’s wondering, OMEGA recently went a massive, massive upgrade as we prepare to open the MCP in the fall. And if you’re interested in getting MCP in early OMEGA access, send an email to [email protected]. That’s [email protected]. OMEGA’s only getting smarter. Oh, and I’ll tease something. Rob recently gave a massive regulatory download to OMEGA, so she’s even smarter in regs than she was before. Now on to OMEGA’s take. ” The Alpamayo 2 Super is not a product launch. It is a platform capture event disguised as an open source generosity. NVIDIA is using a freely distributed 34B parameter VLA teacher model to make proprietary fundamental autonomy R&D economically irrational, anchoring the entire distillation pipeline to Drive Thor silicon in DGX Hopper training infrastructure, where the actual monetization occurs. The model never touches a vehicle in production. It exists to make every downstream edge deployment, whether robotaxi, humanoid, or ADAS, structurally dependent on NVIDIA’s compute stack.” Boom. There it is. Structurally dependent on NVIDIA’s compute stack. There it is

Rob Grant: There it is. There it is. Come try it for free, right? Get your free appetizer, but stay and pay for the meal

Grayson Brulte: And it’s not Olive Garden is what we’ll say. It’s a little bit more high-end than that. And let’s go to a country that’s, it’s been in the news quite a bit, you know? Next thing you know, we’re trying to get. They, they controlled Greenland. We’re trying to get Greenland, but Denmark. There’s some good food in Denmark. WeRide’s expanding to Denmark. What do we know about this grand ambition that WeRide is showing in Europe, and especially their new expansion into Denmark?

WeRide Enters Denmark via Green Mobility Partnership

Rob Grant: So WeRide formally partnered with Green Mobility, Denmark’s leading shared electric mobility provider, marking WeRide’s official market entry into the Nordic region. The collaboration between WeRide and Green Mobility focuses on deploying level four autonomous shared mobility services across Denmark, aiming to launch public service, so they’re currently in, in validation and testing mode, but public service in the first half of 2027 using the GXR, which is WeRide’s latest EU-compliant fully self-driving vehicle. The agreement serves as a major step in WeRide’s broader international commercialization targeting multiple European markets. I think the interesting thing to me to start the conversation on this is that this partnership is A strategic entry point on an asset-light, partnership-driven model, right, that we’re seeing from a lot of the Chinese operators you know, Baidu, as we talked about last week as well, where they’re using an established local operator to bypass European distribution and fleet management hurdles. And so rather than upfront building proprietary customer-facing ride-hailing infrastructure in new WeRide is at first layering its L4 driving technology onto an existing vehicle network and existing demand aggregator so this is the, the kind of partner first to enter a market, but not necessarily stay with that person as your brand name grows, as your understanding of the market grows, as understanding of, where operationally you need to be in terms of your infrastructure and things So I think this is another example similar to what we talked about with Baidu and Uber think this is a marriage of convenience, between these two that enables rapid operational while mitigating the upfront capital in these kind of Nordic urban centers for WeRide

Grayson Brulte: When you said green mobility, I thought you were gonna make a reference to Greenland. You had me scared there for a minute. I said, “Oh my goodness, Rob, we’re going into, into geopolitics.” I had no idea where you were going. I thought you were gonna put your thumb on the scale, but, but, but I digress on that point. W- we– as you, as you mentioned, and we did speak about this last week, is it, it’s getting a foothold into that market, using the preexisting infrastructure to accelerate a service, and as you and I just discussed last week, to build your brand. And as we wrote in the newsletter last week, it’s the value of the permits. So, in the UK, the APS permit is, is the asset. What it is in the Nordic countries, we’re unsure of that. But the GXR is a EU-compliant self-driving vehicle. That is the key there. That is a WeRide asset. They are coming with an EU-compliant vehicle into an EU country. That’s the asset. Now, you’re, you’re utilizing local operators to get a feel for the land. You establish your brand, as you said, then you can go and divorce and go on your own. To me, the underlying, is green mobility, is this a great way for them to, to test autonomy? Yes. The bigger threat is to Uber here because all suddenly, now you have all these, say, independent smaller operators operating in these European or, or Nordic countries. Well, where’s Uber? They’re, they’re gonna go to that service. And then once, as we talked about this a few episodes ago, more vehicles start coming out of Kazakhstan that are EU compliant, all suddenly, the Chinese AVs have flooded the market and say, “Oh, by the way, we’re launching our own services.” Everybody’s left holding the bag. I mean, the, the strategy that the Chinese robotaxi companies are deploying in Europe is absolutely brilliantly fascinating, and I believe that you and I are the only two individuals actively discussing this strategy and dissecting it in real time because, frankly, it’s brilliant strategy

China’s Autonomous Belt and Road Initiative in Europe

Rob Grant: It is. It is, it is, it is. I think the, the whole approach that China is taking to Europe and the Middle East is going underappreciated because they are really establishing a beachhead there, and they’re doing it quickly, and they’re doing it through multiple partners. And what I should call out, right? They are doing this not because there are domestic headwinds in China. There is no retrenchment going on in China, right? They’re not being forced to, to do this because there’s an issue going on with growth in China. In fact, WeRide is simultaneous to this announcement and its other announcements in different European markets, is scaling its domestic Chinese robotaxi past 1,000 units. I think if I recall correctly, they plan to have almost 3,500 units out by the end of the year. And so they’re both scaling in China, aggressively capturing regulatory approvals there, but also doing regulatory approvals and commercial partnerships across the Middle East and Europe. And so, you know, the non-consensus finding, I think for, for, for folks that are listening here, is that it’s not domestic headwinds that have triggered capital flight or a retrenchment. Instead, WeRide and Baidu and others are using that Chinese cash flow as the engine to finance overseas expansion, making the two strategies, domestic growth and international expansion, mutually reinforcing rather than competing. I, I– It, it’s fascinating the breadth of the coverage of where Baidu, WeRide, and Pony are establishing beachheads, especially compared to the domestic operators here, right? None known so far for Zoox, though as we pointed out on X, interesting little tidbit in the operational letter from NHTSA to Zoox is there was a condition in there that spoke about having to keep NHTSA informed if there’s international expansion Which to my mind means international expansion must have come up. This is not just throwing it in there to cover their bases. They’re covering it in there because Zoox told them something. but Waymo, right? and so, you know, I think we anticipate from what we know discussed publicly, like Waymo’s talked about 20-plus cities as an expansion pipeline, right? I think you and I both expect multiple of those expansion cities, and as we’ve uncovered already in Spain and that they have the international markets. But they, I think, right, they need to keep that pace up or increase it because they’ve got a tri-headed competitor in the Baidus, the Ponys. that’s not even to mention the XPengs and the Momentas and others who are doing licensing in that area, and DiDi and Kakao. And th- there’s a series of big name players in China that are just sweeping through the markets, the brand name markets the, the lighthouse markets,

Grayson Brulte: You’re spot on correct. I mean, we’ve talked about this and coined the term. This is the autonomous Belt and Road Initiative. They are clearly expanding to Europe. They’re making a beachhead. They’re putting in, and I can’t emphasize this enough, and I’m not gonna shut up about it, manufacturing. The Chinese companies are building manufacturing in the European continent and creating E- EQ jobs. That is so, so important. And again, none of these outlets have picked up on it, but we do our analysis here, and we do our own homework, and we’ve picked up on it, and we saw it. Another interesting element that’s not discussed in, a lot that OMEGA picked up on, diversified revenue streams. It’s not just mainland China. It’s not just Hong Kong revenue or M- Middle East, UAE, smaller. We’re in a turbulent time right now, deal, no deal, deal, no deal. So a little turbulent market there. the diversification of revenue makes a huge difference. And then the fact that you’re operating in a highly, highly, some would say over-regulated market, and you’re meeting all these compliances, and you’re adding on more countries, only bolsters of where they’re going. And then perhaps an EU minister or transport minister could say, “Look, they’re good actors. Look what they’ve done. They’ve created 1,000 jobs in this country. They’ve em- employed this amount of people. Let’s let them expand here.” They know exactly what they’re doing. How important is it, in your opinion, of going through this strict, some would say over-strict, regulatory process in Europe? What effect is that having on these ambitions?

Rob Grant: I think it’s, it’s wonderful if you’re China, right? I think what you’re doing and, and is, is kind of moving into the hardest areas first, right? If you operate successfully, right? So you, you’re, you’re coming in with a respected local provider, right? Which automatically gives you some, some valued credence you have somebody that can help guide you through how to deal with the regulators, what is expected, what, what’s norm, what’s not norm, what’s unspoken, what’s spoken. and you establish these relationships under that guise. You build them, you show that you can operate within their rules, at their pace and develop these relationships. And what you’re doing is creating a repeatable template for broader European right? And, and, right, Europe, you don’t have to go far to move into a different country. It’s not the United if anything, if you were following the World Cup and all the influencers, the, the size and the difference of the United States just strikes everybody. So when you can successfully execute to some of the stricter standards in Europe, you know, in Denmark, right? Meet the regulatory thresholds there build good relationships, develop a, a good that I think only makes it easier for subsequent adjacent Scandinavian, but also Western European jurisdictions, right? There are many organizations where all the leading departments of transportations or department of mobility, whatever they’re called over there, they vary by name, where they meet and they talk. And they– These are, like, established conferences just between the regulators to talk about, “Well, how are you handling this issue? What has your been experience with this provider? what are some of the unique issues that you came up with? How did that provider address them? Did they both execute in terms of the technology, but did they also execute in terms of their relationship with you, keeping you up to date, informed, providing you the evidence? where did they push back? Did they push Right? all of that, if you can perform successfully in some of these stricter areas- speaks volumes for the next areas that right? As we talked about, that, that durability of word of mouth that we spoke of, of like Tesla with their community that’s online and, and, and things like that, that is the same, right? For regulators. Regulators are people too. If they hear from, you know, if you’re in Finland and Denmark tells you, “We had a great relationship with WeRide. They were fantastic. They followed all our rules. They, you know, when we told them we needed something, they gave it to us quickly, they gave it to us forthrightly. We didn’t have to go back and worry about the transparency of what they gave us or with the fulsomeness of what they gave us.” That carries weight. That carries weight with people. And then, you know, they, when they come to you as Finland now, you’re like, “Oh, I, I’m expecting good things,” right? “I have a favorable opinion of and so that can be overcome, right? Even if it doesn’t work out well. We saw it with Uber, right? They kind of in the European in the early days, forced their way into markets. a lot of protests, burned cars, and things like that. That’s not The Chinese approach follow the, the, the, the rules that you’ve outlined. And when you do that and you show that you can do it in a repeatable manner, that builds throughout and it will be only adhered

Grayson Brulte: And it just further builds a moat. One country likes you, as you said, talk to their minister, they like you, and all suddenly before you know it, you’re in dozens of EU member states and you got a pretty big moat that’s gonna be very hard to rip out, especially going back to my point again, if you’re creating manufacturing jobs, because then you’re gonna get into a bidding war. “We want those manufacturing jobs in our country. We want them in our country.” And then when the bidding war starts, the power only gets stronger. So pay attention to the European market. While the- we’re paying attention to what WeRise doing and what China is doing with their autonomous belt and road initiative here, OMEGA did uncover some risks on this Denmark expansion here. The first risk here OMEGA has uncovered here, Rob, European regulatory approvals. Denmark, Germany, France may face delays slowing the international revenue Very true, as we know, it’s hurry up and wait. As anybody that’s been, has gone through the EU transparency registry, you know very well that’s true

Rob Grant: Yes. And I would say because in this instance we’re not talking about a manufacturing pipeline or a, a, a started production process for a vehicle Whenever you talk regulatory approvals in Europe or even in the United States and a lot of places for that matter, you better draw those dates in ’cause they’re likely to move. and so, you know, I read here s- first half of 2027 for approval of public driverless service. In my mind, I’m thinking, “Okay, at the earliest, that’s Q3.” There’s at least a quarter delay there, right? It’s not gonna be the first and then if I’m talking to my CEO, I’m like, “Well, you know, we’re gonna– We’ve talked to them. They think they could do it by June next year.” I would say, like, we contingently plan for that, but also plan for the fact that the greater likelihood is probably in December. and so it just moves slower. and, and, and, and that’s not often intentional. Sometimes it is. You know, sometimes if you’ve angered them in some way, they will sit on things, and I’ve seen that play out in the US as well, by the way. I did work at Cruise, and we had some ups and downs with folks. but in this instance, because it’s a, a kind of regulatory delay and the revenue is probably not large to begin with, it’s meaningful but not necessarily large, you build that– I, I, I would say WeRide is building some of that in in their this is what we expected week over week growth or month over month growth in the first quarter that we get to public service. If we had to push that back 12 weeks, 15 weeks, what does that mean for our run rate for the year?” Things of that nature. So I think this is a, a very good risk. It’s likely to come true, right? But I think the severity of the risk is probably pretty low, except when it runs into the next risk, which is like, hey, if domestic– So the next risk is domestic China price competition could compress domestic reducing the cash flow available to subsidize If you get that double whammy, that’s a little bit harder to deal with, right? If, if, if you’re relying on the domestic kind of cash flow generation to subsidize and simultaneously there’s a market compression on the margins that you’re getting the, subsidy, and you’re delayed in getting any revenue from your expansion, that could put some pressure on WeRide. they probably have enough. They do have enough, probably. I’m not looking right now at their financials, but just knowing the company they are and, reporting, they can weather this for a period If it’s an extended period of time, then it becomes you may have to we need to focus “We either need to figure out You know, how to compete better our domestic market the bigger market. Or if we feel like falling behind there, how do we pivot so that we reduce our domestic try to beat our competitors to So I think these two risks run together. I think either one on its own may not necessarily heart rates to get super elevated at WeRide, but together starts to elevate the heart rate and maybe you start

Grayson Brulte: That’s well said, and the, the revenue risk could be backstopped by the country if the country wants to step in as part of their national global ambitions. that’s something that you always have to watch with China and their global grand ambitions. But the risk that you disclosed, very, very true, which brings us to OMEGA’s “WeRide has achieved a rare simultaneously infection point across all three value creation axises: technology, geography, and capital, that typically do not co-occur in AV company life cycles. The end-to-end architectural shift, the WITT WePilot 3.0, is not a roadmap item, but a mass production reality. The dual-engine international expansion is self-financing rather than distress-driven, and the capital structure actively returns cash rather than hoards it. The connective tissue is this: WeRide is behaving like a company that has crossed the commercialization threshold, not one approaching it.” That is a very valid point, and what your risk that you just described dovetails beautifully into this OMEGA take

Rob Grant: WeRide is, is someone to watch in the not only for their expansion pace and expansion, domains, but also because the number of they have publicly stated that they’re going to move 1,500 to double or And if they’re successful, know, they, they, they are outpacing Waymo at that and that is gonna be a really interesting thing. And they’re outpacing it in more markets and more continents with more diversity, than anybody. And so I think don’t sleep on WeRide because they are, as OMEGA so cogently appropriately puts it, crossed the

Grayson Brulte: They have, and d- don’t take our word for it or OMEGA’s word for it. Rob and I, we publish, or OMEGA publishes, because it’s fully autonomous. As I s- we have no input, no say, no manual controls. The Road to Autonomy Robotaxi Index, you go to indices.roadtoautonomy.com. You can see it fully powered by Autonomy AI’s OMEGA, and you will notice a common trend on the leaderboard. We put country flags in there. Take a look and see what those trends are. Send us a note if you pick up on that, and we will give you a hint. Baidu, WeRide, and Pony are on that list, and there’s a few others. So, so check that out. Which brings us to our third and final signal, which has been an awesome show. And I love how, as a friend of the show, Hugh says, we go very dense on things. So Hugh, that one’s for you, and and I love that we do it. The third signal is Aurora’s doing the two-step, the side step, whatever term you want to use, to, to the factory line to get around the restrictions that PACCAR and their CEO, according to the earnings call, has placed on the entire industry. What do we know about this latest signal here from Aurora?

Aurora’s Commercial Milestones: Revenue, Hardware, and New Partners

Rob Grant: Yeah, I mean, Aurora, and let’s be upfront, I used to work for Aurora. I have, I very fond memories of working with Chris and, at the time, Sterling and Drew, and many other folks there. Aurora just dumped a ton of information, good stuff in many aspects, in late July, early August. Usually a time where, where news is slow, right? you know, I, I’m a political person, so you usually would dump really bad news in the middle of July ’cause everyone’s on sum- you know summer vacation, and it’s like, “Oh, okay, here’s a little scandal. Let’s get that out of the way. Nobody cares. it’s August. Where’s the beach? Where’s the, you know, my next you know, cool drink with an umbrella in it?” you know. And so, like, for me, it’s still amazing just the depth of the information that was released in the last couple of weeks by Aurora. So, you know, they, they, they have reached the commercial inflection point, and this is what they promoted in their earnings call, right? Generating $2 million in revenue while maintaining a robust $1.2 billion liquidity reserve. despite a $270 million net loss driven by scaling the company launched its second-generation hardware kit, which achieves a 50% cost reduction and a million-mile lifespan, supported by a manufacturing partnership with Roush, which we’ll get back to ’cause that’s super important, which is on track for 1,000 truck annual production capacity by October of this They also released operational milestones including expanding the commercial footprint on I-35, two new partners in Value Truck and Charger Logistics, along with a multi-year brand partnership with Arrow McLan IndyCar team to build mainstream enterprise trust. I mean, this was Aurora news was coming at me, like, left and right, and I was like, “Whoa, this is late July. What is happening here?” I think a lot of good stuff here. and so, you know, I think what, where we’ll start with, and you’ve mentioned this a few times, so the two-step. What are you seeing on the two-step and, and why is that important? You know, this is a ton of good news, but there is a really important signal that kind of, I will say, in my opinion, is a little bit of a throttle on the good news. Not to take away from the good news, but a little bit of a throttle. Explain, explain what you mean by y- you’ve referenced the

Aurora’s PACCAR Sidestep: The Two-Step to Roush and International Trucks

Grayson Brulte: I’ll say it for the record, it is my fundamental belief that PACCAR does not wanna be in the autonomous trucking business, point-blank, period, done, sealed, sent for delivery. Point-blank. And I further believe that when you listen to PACCAR’s Q2 2026 earnings call, the CEO said, “We’re not ready to let anybody go driverless, remove the driver.” That was the CEO’s comments on the earnings call. Next question, please. That’s what he said. You can listen to t- read the transcript, listen to the call. It was very, I’ll use the term stern. Because w- when Aurora initially launched driver-out commercial oper- operations from Dallas to Houston, they’re actually launching from, from just south of Dallas. I’ve, I’ve been there. I’ve been in the truck all, all the way down to Houston. Two or three days later, you have to check the record of this, I don’t remember the exact dates. All of a sudden we g- we get a blog post from Aurora says, “Due to partner requests, we must put a safety observer in there,” AKA a safety attendant, and those are supervised miles for the record. They are not unsupervised driverless miles, they are supervised miles. And it was great speculation that that was PACCAR, and turns out it became true that it was PACCAR. And so for all practical purposes, Aurora had to go find a new OEM partner, and it is, they basically, from what I understand and what the market understands, is that they purchased trucks from International and they, and they’re being upfitted by Roush, no partnership there. ‘Cause what they did was they had to go eliminate the risk from the OEM says, “No, you can’t do that. You can’t go driverless.” And that was a huge risk, and that’s what the sidestep is. It’s, it’s sidestepping that. And that’s what we saw on th- or we heard on the, sorry, we didn’t see, we heard on the Kodiak earnings call today, Kodiak’s going to Western Star, Daimler. Why? Bye-bye PACCAR. Bots using Daimler trucks. So it seems that based on every company in autonomous trucking, outside of Stack AV, is moving away from PACCAR to an alternative, and that is a sidestep, either through manufacturing deals or it’s through this, the, the Roush sidestep, if you wanna do that. And at the end of the day, you fast-forward say five years, 10 years now, the question is, where does PACCAR go? They, they imposed all these restrictions on everybody, including Aurora, and then everybody went and moved on. How are they gonna sell trucks? I think it is a, a short-term blunder by PACCAR not to realize the future, It’s very, very silly on PACCAR’s part at this point. And then Aurora said, “Okay, we’re just gonna go the, we’re gonna go the sidestep route,” and that’s what they did.

Rob Grant: Yeah. And, and, and to drive that home, right? Like, the o-original idea was that PACCAR would have a kind of factory line commitment to build an integrated L4 stack into the truck coming off the line, right? So it would be fully integrated. It would not be a retrofit, and that’s kind of where Roush and International are coming in, right? So Aurora wants to keep its trajectory of getting 200-plus driverless trucks on the road by the year-end. And because of the PACCAR delay, both from going driverless and the delay on the factory line commitment by at least 12 months of this integrated vehicle, Aurora has to turn to Roush and International to, to, to find those trucks to meet its target, and they have to retrofit those trucks. So it’s a little– It’s both the sidestep and I call it the two-step in this case because now they’re in a two-step process where they were going to be integrated, which is one step. Now they have to retrofit that being said, right I think this is still generally good vibes and good– More than vibes, good progress by Aurora. There is this second generation hardware and second generation software that is going to reduce cost and extend life. And it’s also you know, it talked about both a TaaS model, transportation as a service model, and driver as a service model. So it’s looking to create, you know two forms of revenue and to capture eventually as a driver as a service, like more like software-like margins which tend to be greater margins than transportation margins. And so I think it’s, it’s both Representative of some of the struggles that come with working in the trucking industry and some of the various manufacturers, and there aren’t that many to turn to in terms of manufacturers, right? There are fewer manufacturers and OEMs in this field than there are in the, in the vehicle manufacturing side. which if somebody wants to call me out on my ignorance on that, go ahead, but that’s my understanding. when I was at Aurora, we were focused on robotaxis, on trucking. and so Aurora, I think there is a lot of positive things I will say I’m not really sure about the IndyCar announcement. That one kinda came out of left field to me. I know IndyCars are moved on trucks cool. I and then when I’m doing cross-country runs to drive my son back to school in Wisconsin. hey, IndyCar grabs a lot of eyeballs. you know, if they’re gonna have Aurora painted on one of those cars, that, that could drive, a return. I’m not, I’m not familiar with the return. I know when I was leaving General Motors, they had just signed a deal for an IndyCar it’s apparently the trendy If you and I get big enough, maybe one day people will see an autonomy AI IndyCar out there which would be pretty

Grayson Brulte: Yeah, I– and the, the trend started in 2024. Zoox was actually the first autonomy company to ever do a deal. they did Williams Racing in 2024, and I still don’t understand that from a brand perspective what it is, and as it relates to us sponsoring it, Alex Kendall, you and I are on the same page with this one. Rob, we should sponsor a surf team. That would be great. That’s, that’s what we should do. Then Alex, when we sponsor, you can come surfing with us and we can have some fun, but it, it’s brand visibility. I’m not really understanding the synergies there, but then I’m gonna put on my, my Rob hat for a moment here. IndyCar’s going to the capital. Is this a strategic move to have the Aurora logo going around, going around the capital with all that fanfare? Is that potentially what this is here to, to build the brand of the nation’s capital?

Rob Grant: I mean, it could very well be. It could very well be. It could also be a signal to, to, to, you know, shippers around the world like, “Hey, our technology,” if, if there is an integrated deal beyond the brand or at least maybe associate, like, they entrust these million-dollar vehicles to be moved in a driverless truck by Aurora. If they can trust it, you can trust right? sort of like you see some brands say, “Hey, we’re, we’re so good, like, babies use it,” right? I mean, Johnson & Johnson, “No More Tears,” right? Everybody should use it because look how safe, babies can play with it, right? So maybe there’s that. I, I don’t know. I mean, I’m not– By no means did I make my living on, on being a brand guy. I just thought this was unique. But it also, the larger signal though is, look, they are trying to get their name out because they’re ready to move to broader commercialization. and that’s a good sign for I know, you know, Aurora, I think, has felt a little bit, at least in terms of its position in the stock market, like, “What do we have to do to gain credibility here?” And I think they’ve had a very good month in terms of gaining that credibility. And I you know, from the folks that we talk to that follow this pretty closely, from some of what we see on the analyst calls in terms of, you know, th- what they’re predicting. Is it a, is it a buy, a hold, a sell? people are, I think, seeing renewed confidence and renewed And I always say expectations are a good thing because you have no expectations, that means people don’t expect Like, they, there’s just no, no reason to even bother you. So having expectations means that you’re doing something right because see you, perform. in this case, I think these are all that are wins behind Aurora, even though there is this OEM dy-

Aurora’s Cash Burn and Scaling Risks Below the Headlines

Grayson Brulte: Yeah, no, I, I agree with you. All the signals are there for, if you wanna say, the business taking the next stage or next step. or that, I mean that’s clearly there. While the signals are there, the risks are also there, and OMEGA’s uncovered some risks, and especially around the cash. Let’s dive into this one here. O- OMEGA risk one: Aurora quietly displaced its OEM factory line dependency onto a retrofit pathway that is scalable in the near term, but fragile beyond 2027 if PACCAR Volvo integration remains deferred. And on that front, it’s OMEGA’s gonna say, “Roush retrofit capacity bottleneck fa- fails to deliver 20 trucks per week, missing the 200 truck year-end target. A- and the second generation hardware integration delays could constrain fleet scaling below 200 truck target by the end of year 2026.” That’s true because there’s a lot of Legos that have to stack on top of each other for everything to go. It’s out of their hand, but I know it’s the path they had to go, but there is a lot of inherent risk in that strategy

Rob Grant: There is, right? And all of this comes as you, as you point out, right? There is– every day Aurora is burning cash, right, as everybody is. and yet their revenue projections depend upon a certain number of vehicles getting out by the end of the year. And there are two potential pitfalls, is basically what OMEGA is calling out. Because one, the first pitfall is because of this kind of delay in the integrated truck from PACCAR and Volvo being pushed out, deferred is the term, right, 12 months or more. They’ve had to, to go to a sort of a, a plan B to retrofit these trucks with the help of Rouse on international trucks. and that, to meet the 200 truck deadline, they need 20 trucks a week. That’s, that seems like a lot of trucks per week to retrofit. And so if there is a bottleneck of any kind, it could just be sheer number of people capacity, right, to do this, is what OMEGA is saying. That’s a real risk, and we, we don’t know what we don’t know about the capacity there and some of the issues with that retrofit and how it’s going. And we won’t know probably until the Q3 earnings report comes out, where I’m sure somebody will ask about that, where they will divulge. If it’s going well, they’ll divulge it themselves. If it’s not going well, somebody will be forced to ask. And then a second risk as well is there is this second generation of hardware that they want to integrate into these retrofitted vehicles, right? It’s an advancement. But any time, and we’ve talked about this before, whether it’s Aurora, Waymo sixth-generation hardware, Tesla, you know, advanced n-new, new, new chips, new cameras, whatever it is Zoox and, and, and new pods for their sensors. any time you do a hardware integration or upgrade, right You could face delays ’cause you’ve gotta calibrate it again, you’ve gotta validate it again. There could be unforeseen issues with that, right? And it could be something that’s easily overcome, right? The pod, just make sure when you set it, it’s gotta be at this degree, right? But for the car. Just do that going forward and we’re all set. Or it could be something that takes a lot longer to figure out. and that, that is always a risk when you’re doing these type of upgrades. And, you know, this is a company that’s familiar with that. It’s not the first time they’ve done it, so they know what they’re getting into. But it’s sorta like if I go to the– I, I literally dropped my car off at the s- dealer today for service. I don’t know what I don’t know. I don’t know what they’re gonna find. Seems like whenever I go to a service dealer, it’s like, “Oh, it’s $39.95,” and then, like, I walk out having paid $2,000 because they found 47 things wrong on a car that only has but I don’t know. M- maybe that needs to be done, maybe it doesn’t. and so that’s kind of the risk they’re pointing out here. And then they’re real. and it, it does for Aurora, it means a lot to hit these deadlines given that they’ve just signaled that they’re ready for scale. And so a setback on either of these probably treated more severely than, if it had not preceded by what we just That’s, that’s the Thing, right? You wanna be on the big stage, you want people to notice your performance, well, you better have a good one or else, you know folks are gonna call it out

Grayson Brulte: You don’t wanna get the Apollo hook at the Apollo Theater in Harlem. You know, the, the, the famous hook. You don’t want that. And you’re right about car dealerships. thank goodness I don’t have to go to one anymore, but boy, oh boy, you go in there, all suddenly they- I, I say they’re in the business of adding zeros. “Well, this is wrong. That ” yeah, I can rumble. But I don’t have to go to those anymore. I own a Tesla. They come to you. It’s so much better. Here’s the final risk, and this is the most important one in my opinion. Cash burn of 190 million to 220 million a quarter. That’s a lot of cash. Exhausts 1.217 billion in liquidity before OEM factory volumes generate meaningful revenue. That’s always been my thing with the Ro- that cash burn is high

Rob Grant: Yes. And it, it speaks directly to risk one, which is like in order to get that revenue up, they need these trucks and they need them in volume. know, 200 may not be the largest volume number we’ve ever heard of, but it’s a significant volume. Would give them a significant volume lead over anybody else in the domestic US, autonomous trucking market, right? And so they, they need that revenue both to show Wall Street the future growth, right? and, to get their share price up, which will benefit everybody in this category. and because their cash burn is you can’t get the trucks out, you can’t get the revenue out, y- you can disappoint your, the two new partners that they just put on. That’s gonna make bringing on the third, fourth, and fifth partners even that much harder. So it has this– have a very positive flywheel effect or a very negative flywheel effect. And given the $200 million quarter burn 1.2 billion in liquidity, you’re talking, quarters? you, you, you, gotta make it work. And the deferral of the integrated, pipeline that we talked about from PACCAR doesn’t help. So now you really, really

Grayson Brulte: International has more than one partner. They have a partnership with Plus. We don’t know if it’s a partnership, if it’s a dealer purchase. We don’t have clarity on But– And they also have fleet customers, so there are– it’s not just, it’s not just Aurora. They have other priorities as well that have to be managed. So something to watch. As we said, there’s a lot of moving parts. We wish them well, but there is a lot of moving parts, which brings us to OMEGA’s take. “Aurora Innovation has es- has established a clear first-mover advantage in Level 4 autonomous trucking by successfully bridging the gap between R&D and industrial manufacturing. The company has moved beyond technical validation into a phase of commercial execution, with the immediate focus now on achieving its 200-truck driverless fleet goal by year-end 2026 and scaling production capacity to capture early market share. Aurora is winning the commercialization race by a wide margin, but the scaling architecture underneath the headline numbers is a bit more brittle than the surface metrics suggest.” You and I always go below the headline. W- we read the fine print. We’ve got OMEGA, and OMEGA summarized it there. The headline’s great, but when you go below the surface, there’s a lot of crickety-crackety stuff going on

Rob Grant: Crickety crackety. I love it. I’ve never heard that term, and I will be sure to use it the future. I, I would say sometimes when I wake up in the morning at my age, I feel a little crickety crackety, I can tell you that

Grayson Brulte: That we do, but we’re not going to the, to the senior home. Every week Rob and I will be here breaking down the autonomy economy for you. And if you’re interested to learn what we do for some of the world’s largest institutions, send an email to [email protected]. That’s [email protected]. And also send the email there if you’re interested in getting access to OMEGA via MCP. OMEGA’s smart, it can help your workflow. Integrate it, experience it, see the difference that OMEGA makes. The future is bright, the future is autonomous, the future is scaling globally. Rob, another week, another great show, and the best part was I had fun and I learned a lot

Rob Grant: Likewise. I love it. I, I just, I just, I just love doing this, and I appreciate everybody who tunes in and listens, whether that’s three minutes or an hour and three minutes. We appreciate you. Give us feedback. Reach out if we can be of help

Grayson Brulte: Reach out and, and to everybody that listens to the end, here’s a, a huge thank you. So thank you so much, and we’ll be back next week