NHTSA Audits Cybercab While Waymo Launches With Lyft in Nashville

NHTSA opened an audit query on Tesla’s Cybercab self-certification, examining whether Tesla correctly determined that certain Federal Motor Vehicle Safety Standards are inapplicable to a vehicle with no steering wheel or pedals—a compliance conflict, not a safety defect investigation.

Waymo and Lyft went live in Nashville as the first market where Waymo’s autonomous vehicles are bookable through both apps, with Lyft subsidiary Flexdrive operating an 80,000-square-foot depot as the physical operations layer.

XPENG commissioned a dedicated humanoid robot production facility in Guangzhou, with its Iron Unit walking off the line under its own power, backed by a $900 million raise at a $6.3 billion post-money valuation. Across all three signals, the dominant theme is infrastructure and manufacturing readiness separating early movers from long-term winners.

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.

[01:02] Signal 1: NHTSA Opens Audit Query into Tesla Cybercab

NHTSA’s Office of Vehicle Safety Compliance opened audit query AQ26002 covering an estimated 1,000 Cybercabs hours after Tesla launched paid driverless service in Austin. The question is not whether the Cybercab is unsafe but whether Tesla self-certified around FMVSS provisions for steering wheels, pedals, and mirrors rather than seeking a Part 555 exemption as Zoox did. Rob weights it 65 to 35 that Tesla succeeds, and nothing in the query halts operations or caps production.

[32:00] Signal 2: Waymo Launches on Lyft in Nashville

Nashville is the first market where Waymo vehicles are bookable through both the Waymo app and the Lyft app, with Lyft subsidiary Flexdrive running physical operations for a fleet of over 100 vehicles across roughly 60 square miles. Waymo owns the vehicles and the fare, Lyft supplies demand and operates on a service layer fee, and Flexdrive becomes the strategic asset. The signal to watch is whether Waymo and Flexdrive partner in new markets without the Lyft platform.

[47:11] Signal 3: XPENG Commissions Humanoid Production Facility

XPENG commissioned a dedicated humanoid production facility in Guangzhou with an 80% automated build, backed by a $900 million raise at a $6.3 billion post-money valuation, the largest single-round private raise in China’s embodied AI sector. Iron carries 76 degrees of freedom and three Turing chips delivering 2,250 TOPS on device with mass production beginning by the end of 2026. The signal is a manufacturing readiness milestone, not a product reveal, with the risk sitting in a mothership posting a Q2 net loss of RMB 1.34 billion.

Full Episode Transcript

NHTSA Opens Audit Query AQ26002 on Tesla Cybercab

NHTSA Audits Cybercab While Waymo Launches With Lyft in Nashville KPMG ad: 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, it was a busy week with lots of signals. Well, this week, you know what happened? NHTSA said hello to Tesla as they went no steering wheel, no pedals. Well, they did that. Waymo is making music in Music City with Lyft. And lo and behold, the fast follow, sorry, I gotta call it that, Xpeng says, “We’re gonna make automotive grade humanoids now.” File it under not surprised. Let’s get into signal one, the NHTSA audit of Cybercab. Half of the tech nerds are like, “Oh, this is a bad thing.” But then Omega says, “No, this is par for the course.” What did we uncover with this signal?

Rob Grant: First of all, a, a big shout-out to our sponsor, KPMG. We appreciate you as always, even as I’m on the road here in Western Michigan here in Kalamazoo, where I, where I’m sitting right now across from the Western Michigan Broncos Stadium, and all the angry fans who got gypped by that one extra second that the Wolverines were able to add on at the end of last week’s game. So what we uncovered and what Omega uncovered about the n- what happened with NHTSA and the Cybercab. So NHTSA put out an audit query that they opened on September 3rd, 2026 and this covers an estimated 1,000 Cybercab vehicles, and it’s tied directly, as we’ve called all along, right? The trigger was the launch of Tesla’s commercial driverless Cybercab service. And so as soon as Tesla went public and opened it up and started charging for rides, NHTSA came down and said, “Wait a second, I wanna audit your self-certification.” And so that occurred just hours after the public launch. And now, you know, it’s really important to understand what the audit is probing, right? So this is very technical. Under the Code of Federat- Code of Federal Regulations, Title 49, Part 571, go ahead and look it up. You can do it on the internet. It’s a super interesting read if you’re into really boring regulations. But under this Title 49 CFR Part 571, manufacturers self-certify Federal Motor Vehicle Safety Standard compliance without needing NHTSA pre-approval. And I’ve seen a lot of talk over the week about, you know, why the U.S. is so different and NHTSA’s dangerous, but this, this is normal. This is what we’ve done in the U.S. for the last 70 years since NHTSA was formed Manufacturers make a car, they certify that they’re in accordance with all the federal motor vehicle standards, and if NHTSA has a question, NHTSA comes by after the fact and opens an audit. And so Tesla told NHTSA the Cybercab meets all applicable standards, or for some standards that they’re not applicable at all. And so that’s really an interesting kind of perspective into the rationale of what Tesla has done with its self-certification. but as we all know, Cybercab has no steering wheel, no pedals, conventional mirrors, and most legacy safety standard rules assume a h- human driver is physically present to move a steering wheel, to depress a pedal, to look into a conventional mirror, or need a conventional mirror to see behind it or to the sides. So the core question NHTSA is examining, just so everyone is clear, isn’t, “Is this car unsafe?” Right? That is not the question in front of it. It is not saying, you know, the question that we’re all here gathered as jurors, so to speak, “Is this car unsafe?” No, that’s not it. The question is whether Tesla’s self-certification depended on unilaterally deciding that certain federal motor vehicle safety standard provisions simply don’t apply to a vehicle with no manual controls, rather than seeking a formal exemption like Zoox did. So it’s really, really important to understand that aspect of it and concurrent with that, to understand that It is not questioning NHTSA’s neural network perception testing or s- auto- autonomous vehicle stack performance, right? None of that is in question. In fact, the audit sits in the Office of Vehicle Safety Compliance. If you look at the audit query, you will see on top in big black letters, “OVSC.” Now, they don’t explain on the document what that is, but if you’re familiar with the alphabet soup, that is the Department of Transportation, that stands for the Office of Vehicle Safety Compliance. And this is where audits and investigations about certification go to, as opposed to a performance issue, right? That goes to the Office of Defect and Investigations. That’s ODI. That’s a performance investigation. So this is where when, you know, you read such and such company had a recall, right? It could be, you know, Ford for something with its, you know, brakes or, you know, Stellantis for something with its airbags. They come from ODI. That’s a performance investigation. Here, this clearly sits within the Office of Vehicle Safety Compliance. And so, to finish up real quickly, the audit is going to request a ton of technical data and legal rationale related to how Tesla either determined that the standards weren’t applicable and thus didn’t submit evidence, or if they have, in some cases, maybe found an alternative way or alternative evidence to satisfy the standard, what that evidence is. And so, that is really what’s going on here. It is a standard process in the sense that this, this process has existed for 70 years. This is not something unique. It is not something that NHTSA had to put together on the fly because Tesla caught it off guard. It is literally what has been written into, like I said, Title 49 of the Code of Federal Regulations, my good old friend

Grayson Brulte: I love the alphabet soup, but to the listeners and viewers, this is the data that you get in Omega. It is rational, it is level-headed, it’s not hyperbole, it’s grounded in true fact reality. Now I have to, I gotta ding you a little bit ’cause you and I share a common love of the New York Yankees. You’re in Kalamazoo, Michigan. That’s the home of the great Derek Jeter. You’re saying all these other things, but you’re not talking about the great shortstop. I mean, come on now. What’s going on?

Rob Grant: Number two has been my captain, right? My captain, oh, captain. you know, he is the GOAT, so yes hail to Derek Jeter in the home of, of the captain himself, Kalamazoo, Michigan

Grayson Brulte: Now if you come home with a Kalamazoo grill, now I know you’re really cooking. No pun intended on that. So Omega gives this, I’ll call it sober reality of truth, of truly what’s happening w- with the NHTSA audit. Now what does this signal to the market now that we put the truth out there of actually what’s happening with NHTSA’s audit query into Tesla Cybercab?

What the NHTSA Audit Means for Tesla, Zoox, and the Broader Market

Rob Grant: Yeah, I think there are a couple of takeaways, right? And, and some will, will– would have already jumped out to many of our frequent listeners. One, wow, this is a lot different than Zoox, right? What, what, what allows Tesla to take this? Is that unfair, so to speak? I’ve heard that question po- put at me a few times. This seems unfair. Well, it is not unfair. Tesla has made an affirmative decision to take a different route than Zoox. And if we remember, if we go back in the history books, three years ago, or right around this time, Zoox was going down the same route as Tesla. They were going to self-certify, and they made a similar argument. Now, you know, these audit queries are not open to the public so I don’t know exactly the arguments that Zoox was making. I have a good sense, having talked to a few folks that I’m really close to over there. But I don’t think the arguments are identical, so it’s not like Tesla is just trying to rehash Zoox’s old arguments and, and looking for a different result because it’s a different administration, which shouldn’t go omitted, so different administrations have different priorities, right? And a lot has happened in terms of NHTSA actions. so one, this is not unfair to Zoox, but it does have a big impact on Zoox in two ways, and, and others, right, who want to build a purpose-built vehicle like the Cybercab. One, if Tesla is successful, it will not be subject to any manufacturing cap. It can produce as many Cybercabs as it can willingly eject out of the Gigafactory. And then two, this would potentially allow Zoox to go back and seek its own self-certification, knowing, you know, from the Tesla precedent that this is now kind of possible, that, that maybe they should go back and explore that route so that they can get relief from the exemption cap that they’re currently under with the petition that was granted for them. Secondly, for Zoox and Waymo and others currently competing with Tesla robotaxi in different markets, what needs to be made clear is there is no. nothing in the audit query that says that- Tesla has to halt any operations, right? Again, this is not a recall. This is coming from a different part of NHTSA. And so while the audit query is open and the data gathering and the investigation continues on that, Tesla can continue to put the Cybercab out into commercial operations. It can do it at any volume that it, it currently has, and it can do it in any geography that it, it wants to. Now, right now, we know that that’s limited to, to Austin, but there’s nothing in this audit query that suggests that they couldn’t expand it beyond there. Now, Tesla, if it decided to put out greater numbers or go to different geographies, would be taking a bit of a risk because it doesn’t know the outcome of the audit query, right? One potential outcome of an audit query could be that NHTSA issues a recall, as it did with Zoox three years ago, and say, “Hey, you gotta slow down here. We don’t agree with your self-certification. Your path to putting this vehicle into commercial use is through the petition process.” so that is a risk. Now, do I– do we think that’s, you know, a high level of risk? I would say it’s a moderate level of risk. I would say it’s m- more likely than not that Tesla succeeds in the avenue it’s chosen. But the risk is not zero that Tesla could come back and say, “Hey, no, you have to go the Zoox route.” And so if Tesla has, you know, put out 4,000 Cybercabs before this determination is made, they’re not going to be able to commercialize those 4,000 vehicles anymore. They may be allowed to put them into use for free, as Zoox did for a period of time under a different section of the law a special exemption under 49 USC 30114, but they wouldn’t be able to charge. And so you may wind up with Tesla having produced many vehicles that it can’t commercialize. That’s the risk for Tesla. But otherwise, there’s nothing here that limits Tesla’s ambitions other than the risk it’s willing to take on that an adverse decision comes out of this audit query. And so that’s, that’s the two things that, one, kind of helps Zoox. If Tesla’s successful, Zoox can follow it and maybe get out from under the exemption cap it’s under. But two, if Tesla’s successful and if Tesla’s willing to take the risk, Tesla can move much more quickly to build up a fleet and a geography of cities much quicker than Zoox, Zoox can, for sure

Grayson Brulte: When you look at this, since the Department of Transportation has been very vocal and there is public documentation from DOT where they’re not gonna judge the ADAS, the automated driver system, had to use a technical term there, what impact does it have on Tesla since these vehicles are manufactured in a factory? I’ve been in the Cybercab, you’ve been in the Cybercab. We both have had the opportunity to walk the factory, and I’ve said this to individuals and I said it to, to Jason Calacanis who, who came on Autonomy Markets, which will be released on Saturday. It is a safety cocoon when you see the way that they built it with the gigacasting. Does that help Tesla? And I, and I bring that up because as you and I uncovered on our fieldwork, Zooxs are not factory-built. They’re upfitted. Does that make a difference as it relates to this because the way that Tesla has engineered the Cybercab for safety?

Why the Audit Will Be Narrow and How Pending Rulemaking Could Resolve It

Rob Grant: For this particular query, I don’t think it will make a big difference. I think this query is both by design and by, I think stated purpose of the administration going to be narrow. I think it is going to be very narrow on the question of if Tesla’s arguing that certain safety standards are inapplicable because the vehicle is purposely built without foot pedals and steering wheels and, and side view mirrors and things like that so that those standards are inapplicable. That’s the question that, that NHTSA and the Office of Vehicle Safety Compliance will look at in a very narrow sense. they will not, I think, expand to take into consideration both things that might harm Tesla, and I’m not saying that I find it harmful, but they– that others might argue, ” Hey, you need to take account of their perception system and what happens when it gets occluded,” right? Or things that you could argue benefit it, which is, you know, your, the manufacturing process and the safety cocoon that you’re talking about. So I think it’s gonna be a very narrow examination. Doesn’t mean that it will be a short examination. It could last, you know, weeks to months. and in some ways I, I, I might argue that the longer it goes, the more favorable that could be to Tesla. Now, a, a short, “Yes, you’re fine,” is obviously a great answer, right? That, that would be the answer that Tesla wants. But I don’t think it’s gonna be a short answer in either sense, but I don’t think we should be worried that it’s a longer wait either only in the sense that NHTSA’s currently got a couple of proposed rules in front of it, including one to amend the current safety standard on brake pedals that would specifically allow for a vehicle without a brake pedal. And so if that rule were to become official instead of a proposed rule, say, at the end of the year in December that would favor NHTSA, right? That would take off one of the questions of the audit query because now the vehicle is in compliance and, and pretty clearly in compliance because the rule was just changed to make it in compliance And so I think that’s one of the things that is a really kind of interesting wild card here is that the rules that NHTSA has proposed to change, a few of them, including on FMVSS 111, which is rear view and side view mirrors, they’re. And 135, which is the braking standard. They’re, they’re up for consideration for change. And so the question is, does, is there a way to dovetail these two paths such that you get both answers out of one rule change or does, you know, NHTSA, and I, I, I don’t know how NHTSA feels on this yet, I haven’t, I haven’t necessarily made any inquiries there to date Are they a little perturbed? Like, “Hey guys, you, you couldn’t wait three stinking months to get this done? You had to rush out in a big public way and kind of put us in a, in a, in a bind here?” you know, so we’ll, we’ll see kind of where the administration feels. While I feel that emotional response is is probably a bit fa- of a fallacy, I mean, NHTSA is sober-headed, there probably is a little bit of that sense of like, “Come on, like you can’t wait 12, 12 stinking weeks till we get this done?” But you know, Elon is Elon and you know y- there’s no time like the present to get things done, I suppose.

Grayson Brulte: There’s no time like the present, and if you look at the communications that’s been shared on X, and if you look at some of the public statements by Tesla executives, it becomes very clear, in my opinion, that this was very communicated to NHTSA. NHTSA knew this was coming. I wouldn’t be surprised if Secretary Duffy knew this was coming. So I don’t think it, it came as a surprise, ’cause I, I look at that signal and I also look at Tesla is showing confidence to the market when they put out, “Help us grow our robotaxi network.” We filled it out. A lot of folks wanna do it. We talked to, you know, our good friend Sergey over at Koop, who’s the proud sponsor. He’s back of this week in the autonomy economy. So Sergey, thanks so much, and the wonderful team at Koop for, for being our delightful newsletter sponsor, is he’s getting thousands of inquiries to do this. And so the, the confidence is, i- i- is building there with all this, I’m gonna call it bubbling confidence. There we go, bubbling confidence. I- is– should we read that as a signal, or is this, are we just reading too much into the the hype that’s surrounding the noise?

Rob Grant: I think it’s a signal. I think it’s a signal that they’re confident. I also agree I imagine there were conversations right? We kn- we’ve known for a long while, and we’ve, we’ve told our clients that, that, that come in and ask us for a long while, “Hey, we’ve got, we’ve got an evidentiary record that Tesla’s gonna go this way. So, like, don’t get caught up in anything else. Like, they, they are going to self-certify.” And we’ve said that to our clients for, I don’t know, since the beginning of the year, practically. and I, I do think the, the, the putting out of the, the, the fleet notice, like, for people who wanna be fleet owners, I think they feel very good with where they’re at. I also think, right, like, I don’t think they’re gonna do this ’cause I don’t think they’re going to have to do it, but they could retrofit some of these vehicles with steering wheels and brake pedals, right? so they’re not kind of up a creek, so to speak, without a paddle. if, if the kinda nuclear decision were to come down and be like, “Hey, you’ve gotta go through the, the petition process like Zoox,” I don’t believe that they will have you know, to, to put to the side the Cybercab. I think they will be able to retrofit a certain number of them to keep them in use to gain mileage and, and all that kinda stuff, consumer experience as well. But I, I don’t think that is option A, B or A or B for Tesla right now. I think that’s, like, a nuclear option if they have to go that. I think, I think if I’m thinking of them from the perspective outside of just the legal arguments, the engineering arguments there’s a commercial aspect to this as well, which is, like, kind of touching on what a lot of recent mobility technology has done, which is get it in the hands of as many consumers as possible because it’s really hard to take it away from them. You know, regulators and legislators have. It, it’s just it’s not impossible. We’ve seen it happen in a few places. but it’s really hard if consumers really want a product, and particularly if there are certain segments of consumers, like those that have disabilities you know, those who are mobility challenged folks that rely on cheap transportation to get to jobs as an economic ladder. If these are the folks that are like, “Hey, I am relying on this method to help me out.” and so if you can get it into their hands as much as possible and just the sheer number of, of people who can use your product, right? If you show that it’s kind of beloved, it just makes it a little bit harder, that commercial success for regulators to be like, “Okay, we’ve gotta unwind all of this.”

Cybercab’s Accessibility Features as a Commercial and Regulatory Asset

Grayson Brulte: We know for a fact, we’ve seen them, and there’s been multiple photos published in public where some of the Cybercabs do have steering wheels with pedals. Those are the, the mules, the, the testing. But if you look at it from a disabled individual, David Moss, special field reporter for us, he produced a field report which you can watch on our YouTube channel, with his father in Austin riding in Cybercab. His father’s blind, and he was able to get in no problem, and the door did not close on his father, unlike the unfortunate banana. and, and, and I gotta say, for David, how the heck you end up on the front page of the New York Post, God bless. I don’t want that to be me, but we’re so happy that that was not our segment that got you on there, but well done there. And then what David was telling us in the field report and in a private debriefing that he gave us, which I’ll share with the audience, is that if there’s an individual in a wheelchair, the way, the, the height of the Cybercab lines up beautifully for an individual to transfer from the wheelchair to the vehicle. And so, and all the braille inside, things that you and I necessarily wouldn’t look for, but if you have an individual that’s disability, this thing was, was built for individuals that truly need mobility is what we’re gathering from that field report.

Rob Grant: Yeah, I think, I think everything from the big things that you covered there to some of the, the little touches that we’ve seen when we s-sit inside the Cybercabs in, in static demos or in some of the rides that, that, that, that we’ve been able to privileged to take. I think that attention to detail, right is important because it, it provides. Even though this isn’t a petition process, right? In the petition process, there is a public interest argument that has to be made in order to get a petition. And I think, you know, what I’m speaking to, that commercial public interest side, having that and having support, even though this isn’t a public process like the petition process is, I think that is als- It, it’s just helpful to have, right? It’s hard to kind of measure the direct effect because, you know, the o- again, this is a, a very different office than the office and the secretary who ultimately wind up signing off investigating and signing off on a petition. But I think treating it in a manner similar to the petition process is only helpful for Tesla. And, and so those attention to detail, particularly for you know, disabled users is, is super valuable for them

OMEGA’s Risk Assessment on the Cybercab Self-Certification Audit

Grayson Brulte: Well, Omega’s broken down the data and in Omega’s opinion, Tesla’s doing this the right way. They’re being transparent. They’re going through a well-established process. It’s been around for 70 years. There are risks, and Omega has highlighted the following risks here. Risk one, NHTSA finds self-certification unsupported and requires part 555 exemption petition for the Austin fleet. Where do you think we are on a scale of zero to o- 100 of that potentially happening right now?

Rob Grant: Yeah. I mean, if, if 100 is the grade that is, like, this is absolutely certain to happen, I would put it at around, I would give it about 35. like I said, I do think it’s more likely than not that Tesla will succeed in this, both because of the engineering attention to detail that is needed to make these kind of technical arguments, but also their rationale legally, and because the, the rules are changing and in the process of changing. So I think, I think in my view of the world, I think those things are going to align on timing, and I think you’re going to wind up with a, a, a kind of you’re okay kind of feedback from the Office of Vehicle Safety Compliance. I do think this will put additional pressure for NHTSA to be perhaps either require more reporting under their SGO stating general order, which is kind of the, the process under which kind of incident reports have to be filed to the federal government. And that’s, you know, if you get into an incident that obviously causes physical harm to somebody, but also property damage of a certain kind. But I think it will also kind of push them more towards a, a more robust reporting requirement generally in, in some of the rules that are upcoming. because I think the biggest difference for NHTSA between the petition process and the self-certification process, beyond the, the, the legal differences, is the reporting requirement difference. So Zoox has to report a lot more and a lot more frequently under the petition process to NHTSA. And I think, I think for the industry, those reporting requirements are not necessarily super onerous, but they are rigorous, and they do provide a sense of transparency. Not a sense, but it is transparency, and it, it provides more data through which I think people like NHTSA and thus the general public can feel comfortable with what’s going on with these vehicles. Tesla, if it, and when it’s approved, in my opinion, self-certification, won’t have those similar reporting requirements. So I think that’s kind of– I think you’re gonna get the approval for Tesla. I do think that’s more likely not to happen, 65 to 35 is how I would weight it in my head. but I think it’s gonna force NHTSA to, to really think about kind of the additional reporting it wants from, from those that aren’t going through the petition process, that are running a level four purpose-built or even a level four traditionally built vehicle, because I think it’s gonna say, “Wait a second. Okay, w-we wanna move this forward, but now there’s an information gap, and we don’t like that.”

Grayson Brulte: I like the odds that, that you put out there. We’ll p- we’ll, we’re gonna follow those odds and, and hopefully the odds are correct and, and, and your intuition is correct, which brings us to risk two. Federal pedal-less and wheel-less rulemaking stays unfinished, prolonging certification uncertainty. It’s the federal government, so after all, you never know

Rob Grant: Yeah. And I think, I think, right, this is a, a little bit down on the, the, the risk register from the nuclear option of Te- of NHTSA saying no to Tesla. It is, NHTSA takes its time, right? Because I think it would be kind of wrapped up in a bow if these rule changes were to happen prior to any final decision on Tesla’s self-certification. If the rule gets delayed, which is certainly something that happens all too frequently with rulemaking you could wind up seeing Tesla’s audit query delayed. And while right now, right, we’ve said Tesla can go forward, there’s no real restrictions on what they can do in terms of commercialization under the audit query. It does kind of imbue risk onto Tesla in terms of how bold it wants to be in its bet that they’re sure they’re going to win the self certification argument. And so yeah, I think, I think uncertainty, not fatal to Tesla here in terms of the ultimate decision because it could be just because the rulemaking is delayed. You know, from a market perspective in terms of, you know, I mean, Tesla’s public equities and stocks and things like that, uncertainty is probably not super kind to them. I think the market would crave certainty here because then it can project with, with much more confidence. Okay, if I know, you know, what the Gigafactory is capable of, if I look at their supply chain, if I look at kind of the, the margins that they’re saving with their unboxed kind of manufacturing process. If I look at some of the bottlenecks perhaps with the chips and the batteries I can get a better sense of how I project Tesla’s rollout of Cybercabs and margins over a much longer time period.

Grayson Brulte: Which brings us to Omega’s take. Omega’s got a good, this is a meaty take, so I’m warning the audience here. It’s a meaty take. The signal here is a compliance conflict, not a safety defect. See, that’s the key there for folks. I’m gonna re- repeat this ’cause this is important. The signal here is a compliance conflict, not a safety defect. Media, read Omega, she’ll teach you. AQ26002 is a narrow FMVSS self-certification audit on the control-free Cybercab, not a coordinated federal stand down or an autonomy architecture probe. The connective signal is a pr- is a prove compliance first gate that keeps Austin driverless service operating while Tesla defends inappability with technical data on compliance. Tesla’s engineering leadership will hold the Cybercab hardware roadmap frozen and address the AQ26002 through expanded self-certification data and compliance staffing rather than manual override retrofits. Brilliant take

Rob Grant: It is. I mean, hey, y- you know, if you only got 32 minutes to listen to something today, that was it right there, and that last minute that you read, that is invaluable to folks. Read it, listen to it. Ping us if you have additional questions about it. We’re happy to talk further about it, but that is the, the, the money, big money takeaway was that take right there

Grayson Brulte: As Rob said, we, we get great feedback from listeners and viewers. Clients, reach out if you want a full briefing on this. This is, this is very important. Or if you’re not a client and you wanna become a client, reach out to [email protected]. We can walk you through all of this and what it truly, truly means. Which brings us to signal number two. Waymo is playing music in Nashville. They’ve got the Waymo app cooking, maybe that’s got the blues, and then Lyft’s got the honky-tonk. And you can get it any which way you want. What is the signal here that Waymo’s available on Lyft now and the Waymo app in the Music City? And, and rest in peace, Dolly Parton, you were wonderful

Waymo and Lyft Go Live in Nashville Across Both Apps

Rob Grant: Yeah, I love this, this one fleet, two app story that they have going on. So just this week, after some, some period of time since the announcement of this kind of joint venture together, Lyft via Flexdrive and Waymo went live in Nashville as the first market where Waymo’s fully autonomous vehicles are bookable, as you said, through both the Waymo app and the Lyft app. Riders requesting standard Lyft options, which include standard, priority pickup, wait and save, extra comfort. Shout out to my Lyft heritage. Big fan of Lyft. obviously worked there for many years. It’s got a fond place in my heart. so riders requesting any of these standard Lyft options can now get dynamically matched with the driverless Waymo at no extra cost if their route falls in Waymo’s service zone. They do have the ability to opt out. And so this is a big deal, right? The first kind of signal that we talked about is kind of the regulatory signal. and that is going to be important going forward. But as we’ve talked about and we focused on for many, many weeks and many conversations, this kind of infrastructure signal is critically important, right? It’s right there with regulatory as kind of one of the more critical areas to win and to understand margins and to understand rollouts and commercialization strategies. and so that’s why this makes it as signal two this week.

Grayson Brulte: It’s a huge signal. What do we know about this rollout? We know Omegas uncovered there’s two depots in Nashville. We know that FlexDrive is going to manage the fleet. This is the first market with, with Waymo where FlexDrive is going to be the, the partner. We know that there is change at Lyft. Lyft got a new CFO. But what does all this mean to the partnership?

The Three-Layer Operational Model: Waymo, Lyft, and Flexdrive

Rob Grant: Yeah. So I think, I think this is an interesting structural signal because here we have a three-layer model developing, right? So layer one, Waymo is supplying the AV driver tech and vehicles. So these are still Waymo-owned and vehicles, right? And, and, and Waymo is the s-stack pr- powering it. Lyft is supplying the consumer demand platform, and we’ve, we’ve talked about these kind of demand and, and tech kind of combinations a lot with Uber’s kind of dominating that conversation. But don’t forget about Lyft. Lyft is in this conversation as well as a demand aggregator. But the third layer here is what’s– to me really stands out, is that Lyft subsidiary, Flexdrive, is running the physical operations layer: fleet maintenance, cleaning, charging, depot logistics. And they’re doing this out of an 80,000 square foot Nashville facility, which is, which is pretty substantive, right? I mean, some of the facilities we visited are smaller than that, some are much larger, but 80,000 square feet, that’s a pretty good size facility. And so Waymo’s local fleet that Flexdrive is running the physical operations layer for, runs over 100 vehicles, as we estimated, across roughly 60 square miles following Waymo’s independent launch into Nashville back in April. And so you know, what Omega has further uncovered is that it believes Lyft is operating on a dispatch and fleet management fee while Waymo retains the vehicle ownership and core autonomous trip economics. So in the Nashville deployment, a single Waymo-owned fleet is dispatched dynamically, as we mentioned, across both Lyft and Waymo apps, with Lyft Flexdrive providing the logistics. And so no exact percentage split was disclosed in the Lyft 10-Q or 10-K filings, which framed the deal as a network expansion. But the operational mechanics, we believe, and Omega believes, points to Waymo capturing fare minus a platform fee, while R- Lyft is retaining a service layer fee tied to Flexdrive’s fleet operations.

Grayson Brulte: It is a wonderful partnership today. I have to say today. It’s a wonderful partnership today because unlike Uber, that there is no friction. But the big signal that you alluded to, and you talked about, that you and I have discussed on many shows here, is Flexdrive. Lyft is positioning Flexdrive as an asset, and I say that because they’re putting it in the press releases. Risher is talking about it on X. You’re getting blog posts from Lyft talking about, they’re really positioning Flexdrive as the strategic asset, and that is an asset that Uber doesn’t have. The way this is structured is very interesting, as we said earlier. It’s available both on the Lyft app and the, and the Waymo app. Strategically, what are the signals? W- what are we reading into this? What, what’s happening, a- and how is this so much different than the initial Uber deals in Austin and Atlanta?

Rob Grant: Yeah, I, I, I think the fundamental difference here is that Lyft is positioning itself less as a rival to Waymo and more as the central operations layer underneath it, right? A, a, a Lyft framing of a kind of a hybrid network bet where the AVs and human drivers coexist rather than one replacing the other. And so it’s really that competition aspect to me that stands out, right? I mean, we’ve seen where the Uber and Waymo partnership has really had a lot of friction over economics and exclusivity and, and things of that nature. whereas here, yes, Lyft’s demand aggregation platform is a part of the deal, but it’s not the one, as you mentioned, that they focused on, right? It is on the Flexdrive operations layer. And, you know, I’ve seen a lot of press about how former Lyft drivers are the ones that are doing the work in the operations center. And that is just a much different approach than what you’re hearing Uber saying, which is, you know, even going out and, and, and advocating for rules that say, “No, you have to have a hybrid human network because you’re gonna lose all these jobs, and we have to be you know, protective of our rideshare drivers. And but we also want, you know, to have all the apps AV stack developers on our one app because they can’t do it on their own,” right? And so, like the Waymo app, you shouldn’t even open that. You should come to, you should come to Uber and book your Waymo or book your Motional or book your whatever. And so just the way they, both companies, Uber and Lyft, talk about not only the market moving forward, talk about their role in the market, structure their deals in terms of take rates and fees and platform fees and service fees, it’s all very different, right? I think it boils down to Lyft is not putting itself out there as a rival to Waymo. It is putting out there as a kind of adjacent service provider to Waymo, right? In automotive terms, I, I, you know, m- and this may be a little bit simplistic, right? W-Waymo is like the manufacturer here, and Lyft is sort of the tier one supplier. and, and, and that, that is a what do you call it? Symbiotic relationship, right? That means that they, they, they help each other and what’s good for the goose is good for the gander. I’m thinking of the shark and the remora that swims underneath it, right? I don’t know, maybe I’m watching too many nature shows on all my flights. But it is, it is a much different positioning, and I think that is the really strategic signal to take away here is that you could see a Waymo-Lyft expansion into other markets, given the way that this has rolled out and the way they talk about it and what we understand of the economics. Whereas, like, we all already know Uber and Waymo is on the splits, right? We all think that it’s probably behooves them to even break up their partnership in Austin and Atlanta, which extends through 2028, sooner. this, this seems like a much different relationship based on much different economics, based on much different partnership

Could Waymo and Flexdrive Partner in Future Markets Without Lyft?

Grayson Brulte: One of the key differences that we were able to uncover through various blog posts is that Lyft is investing millions upon millions of dollars in the infrastructure, and they’re, they’re, they’re highlighting that. And when you look at the messaging coming out of Lyft, you look at the messaging coming out of Risher, out of Lyft. I’m beginning to think, and Omega has– We haven’t asked Omega to analyze this, so we’re doing this here on the fly, is did Waymo go to Lyft and saw the value in Flexdrive, wanted to f- to try another fleet management partner, and Lyft said, “Okay, we’ll give you Flexdrive, but oh, by the way, you’ve got to make the vehicles available”? That’s something I think we need to watch because I’m starting to see signals in the market where Waymo and Flexdrive could partner in, in, in new markets without Lyft. And I think that’s something really interesting to watch, ’cause to me, what Omega uncovered, everything that you and I have analyzed, I think there’s something bigger brewing here than we truly discussed yet

Rob Grant: I would agree, and, and like you said, we haven’t really discussed this at all. but I’m, I’m thinking along similar lines that, that there is a moment where the Lyft platform is not a part of a, a fur- a future deal, right? And Flexdrive, you know, as, as Waymo is trying out a series of different kind of fleet operators, right? They’ve got Element and Flexdrive they are kind of in a similar way that Uber is testing AV stack developers as I think potential partners for one to maybe take the lead. I think Waymo is kind of doing a little bit of a bake-off here as to which approach is best. Maybe we do have a mix of different things, but is there one of these that can really set the bar for the others or become our default kind of fleet operator? because I, I do think unity in fleet operations is, is gonna be important. I think, I think fleet operations and uniformity go hand in hand with better margins. I think having too many different operating styles in too many different cities, that, that to me, in my experience, would be a margin compression certainty

Grayson Brulte: And this, it, we’re, we’re gonna watch it ’cause as we said, there’s some, there’s a lot more here to, to analyze. And while there’s stuff to analyze, there’s also risk, which brings us to Omega’s risk. Omega’s uncovered a couple good risks here. First risk here, Lyft-Uber competition commoditizes demand aggregation constraining Lyft logistics leverage. That is a very real possibility as Uber adds more autonomy onto their network

Rob Grant: For sure. For sure. I mean, I think, look, I mean, Lyft’s leverage is a lot less than Uber’s to begin with, right? And if the demand aggregation layer becomes the layer that’s commoditized, which is kind of Uber’s worst nightmare, right? Uber will still retain more leverage than Lyft because it’s just a bigger platform. And so you know, basically as I read this risk that Omega’s presented, it’s like the Uber and Lyfts of the world lose leverage writ large, but who’s the bigger loser out of that is Lyft because it has less leverage than Uber to begin with.

Grayson Brulte: And that brings us to the next and final risk is that Waymo gets so big that they control all the leverage and reset the terms. And I think you and I both know that risk is valid.

Rob Grant: Yes, 100%. My, my guess is that, that Waymo will take any opportunity to reset the terms in its favor. So you know, and I think as we’ve understood in our, you know, many decades of experience in this space Waymo drives a hard bargain, and they drive a hard bargain with almost all of their partners. And so whether that’s manufacturer, fleet operator you know, I don’t, I don’t care. The, the person that’s probably washing the windows at the Waymo facility down in Sunnyvale, they probably drove a hard bargain with. so I think you know, as they continue to, to become a bigger beast, the bargains will get even more more of a second look, including their current ones and future ones.

Grayson Brulte: And what you and I know from all of our relationships and all the work that we do, Waymo’s the one everybody wants to dance with. That’s the one that you and I get more, more inbound outside of Tesla and Robotaxi, more inbound, more questions about Waymo than any other company outside of Tesla. So they have the ability to drive it ‘ cause the market demand is there. Not to mention they’re probably, their bond offering is probably gonna be oversubscribed, so we’re gonna have to watch that to see where that goes. Which brings us to Omega’s take. “The Lyft-Waymo signal is an operational scale through separation of stack and operations. By having Lyft via Flexdrive fund and run depots like the 80,000 square foot Nashville hub, while a single Waymo-owned fleet dispatches across both Lyft and Waymo apps on a service layer fee. This partner-absorbed CapEx plus dual app dispatch reframes the launch slip to September 2026 as the build-out is outpacing, not a structural failure.” That’s a good take there, because it was a little bit delayed.

Rob Grant: It was. It was. And I think a lot of people were trying to figure out what was going on there, right? but I think in the end, this is gonna be a good, good s- good deal for both companies. I think this seems to be off to a really good start. and as we said, the messaging on this is quite different than what we’ve seen from the Uber-Waymo partnership

Grayson Brulte: Yeah, and what Omega’s take was, goes back to infrastructure. One of the things that I think you and I are the only podcast and company that’s, that’s studied this and talked about the infrastructure for robotaxis in depth, ’cause it makes a huge difference. And to plug us, we’ve got a Nashville Waymo field report coming up soon, so tune into our, our YouTube channel for that, ’cause it’s coming. And Rob, this brings us to signal three. Xpeng. They want to build robots like they build EVs. And when I say robots, I mean humanoids. And I’ll tell you something, when Xpeng puts out a post on X or they make a statement, all suddenly they back it up concretely with a real product. What do we know about the third signal?

Xpeng Commissions Humanoid Production Facility and Iron Unit Walks Off the Line

Rob Grant: Yeah. First of all, you know, people go look at the, the, the X tweet. I’m, I’m forgetting the current term for it right now. that Xpeng put out on this because they have a great video, and it’s very dramatic with this kind of humanoid walking off the line. and I give them kudos. It very much looked like kind of a scene out of a Stanley Kubrick film, kind of dim lighting and very, very dramatic. I was like, “Wow, they’re, they’re really kind of setting a, a, a, a, a, a, a, like a mini film up here.” It’s great. But for those who may, may be wondering what I’m babbling on about on– just earlier this week, Xpeng commissioned its dedicated humanoid robot production facility in Guangzhou, and the first production line, Iron Unit, walked off the line under its own power. This is the video they showed. And the transition point, this marks the transition point from R&D prototype to line manufactured hardware. And so, I mean, what this wasn’t, this wasn’t a product reveal, right? The Iron Unit of theirs has been something that they promoted for a bit. But the, the real signal and, and why we’re mentioning it is what Xpeng was putting forward, right? What the, the kind of harbinger that it wanted people to walk away with is, “Hey, you know we can build our automobiles in a really solid fashion at really high numbers with great quality and you know, kind of great facility behind it.” We’re doing that for our humanoids now. So yeah, you might, you know, to, to all those out there, you might have, you know, a couple of, of things that you could put together and maybe put 30 or 40, you know, units together in a month. We’re bringing auto-grade production to our humanoids, so, you know, watch out. Like, we can kind of overwhelm you with what we’re able to do here. and I think that was a really interesting signal. It’s, it’s a– It, to me, it starts to mark a shift from as we’re seeing similar to autonomous vehicles and what we saw in, in drones, which is it’s, it’s no longer satisfactory to have a prototype, right? You have to show both the kind of technical capabilities but you also have to show the manufacturing capabilities, right? It is a, a, an infrastructure play again, right? Do you have the capability to mass produce your humanoid? ‘Cause if you don’t, we’re gonna overwhelm you because we know how to do it, ’cause we’ve done it with cars, and we’re showing you right now that we can do this with robotics as well. So that was the, the kind of how of, of this that matters more than the what, right? Like, the what is great. Okay, we produced a robot that walked off a production line. But the how is, yeah, we can do this a thousand times over again because we know how to do this when it comes to vehicles, and we’re using the s-same production processes to do it for our humanoids

Grayson Brulte: It sounds eerily similar to what Tesla’s doing in the Fremont factory, getting rid of the X and getting rid of the S, so no more sexy ’cause the S and the X are missing, and the middle letter is the E. You could fill it in and you know what that is. We’re a children’s we are a PG show here, so you, I’m not gonna fill that out and spell it, but you know what that means. So, so they’re, so Xpeng is showing, okay, automotive grade, and then they’re saying, “Well, wait a second. No, no, no, no, not just that, 80% automated for the build.” They’re, they’re flexing their automotive manufacturing skills. That’s Impressive

Rob Grant: Yeah, 100%. Yes, very much, right? It’s betting that its edge is industrialization capability, not just the robot specs. and the fact that we got robots building robots man, if I could talk to my grandmother today, she’d j- she’d just laugh. She would have no idea what I’m talking about. But, I mean, this is a real kind of you know, big signal that people should take into account because I think others that are coming are gonna be judged by this. Okay, you know, a humanoid manufacturer or humanoid developer, of which there are seemingly endless of them now, that’s great. But how are you gonna manufacture it, and how many can you manufacture, and what’s the process you’re going to do with it? Do you have to spend capital to do that? XPeng’s like, “I’ve already got a line. I can go back and forth between my automotive line and my humanoid line. 80% of that’s already automated. I can use humanoids to fill in some of the rest of that 20%, and I can just start cranking these things out faster than you can even sign a, a letter of intent to provide some kind of pilot scale operation.” So I think it’s a, it’s a really kind of. The reason they made such a big deal out of it, it, it is, I think, a transition point because it’s, it’s asking or demanding of others, like, you need to be measured along these lines because this is where we’re gonna separate the winners from losers. It’s, it’s not only technical capabilities, it’s industrial capacity and manufacturing capacity

Grayson Brulte: Xpeng clearly has grand ambitions. What are the technical details? What do we know about the iron humanoid? Have they released any details on that?

Rob Grant: Yeah, Omega was super helpful for me here, was able to track down some of these details. so the iron itself has 76 degrees of freedom body wide, 21 per hand, three proprietary Turing AI chips delivering up to 2,250 TOPS running XPeng’s physical IAI foundation model on device. So that means there’s no teleoperation, very low latency, and no reliance on external connectivity. So the technical specs are quite impressive as well

Grayson Brulte: So you have the technical specs, which th- they’re building something, and then you have it going into mass production by the end of 2026 and, and ramping up 2027 and 2028. That comes on the backdrop of this unit, and Omega uncovered this. The XPeng’s robotics unit raised $900 million on August 24th at a 6.3 billion post-money valuation, the largest single-round private raise in a Chinese embodied AI sector to date. So not only are they, they’re, they’re channeling Stanley Kubrick, God, who is an absolute genius. They’re saying, “Look, we’ve got the money, we’ve got the valuation. Here we come.” To me, they’re sending a message to the global stage

Rob Grant: 100%, right? And I think, I think that’s what I really liked about it. It was, it was quite bold what they did. a- and like I said, it’s, it’s bold in a way that’s not about a product reveal, right? Which, which is why this is the signal again, because it, it reveals that how we’re going to measure success in this AI autonomous world for humanoids has now matured. It has matured beyond just can you build something that’s impressive and, and that, that can be useful to people but can you do it at scale?

Grayson Brulte: Can you do it at scale? XPeng clearly has the ability to do it at scale, and, and Omega has highlighted this, that XPeng is looking for humanoids to be the next growth lever for the company. And if you look at the amount of capital they raised at the va- at the post-money valuation, the writing’s on the wall where XPeng wants to go and what they truly believe, which brings us to Omega’s risk. Now, Omega covered a good risk here. I wanna read this one to you. “Vehicle margin pressure and Q2 net loss of RMB 1.34 billion constrained development capacity and capital spend for XPeng on its robotics unit.” That’s true. But even though you raise the outside money, you still need the mothership to keep things humming along

Rob Grant: For sure, right? And, you know, when you have two big growth areas within your firm, like Xpeng does in terms of what it sees from its automotive unit and from its robotics unit when things– if they get tough, then tough decisions have to be made, which means, you know, perhaps we’re gonna focus on one big growth curve at a time as opposed to kinda simultaneously pushing both. And so that’s what that risk is basically, you know, saying in a much more sophisticated way, which is, you know, if losses continue to mount for the mothership, the mothership might have to make a difficult decision in terms of which area of manufacturing and which area of production It favors

Grayson Brulte: It, favors it, which brings us to Omega’s take, ’cause Omega did the same thing that you just said. Omega goes on for her take to say, “The signal in this Xpeng release is a manufacturing readiness milestone, not a product reveal. The company is proving it can build humanoids the way it builds cars. Xpeng is betting that mass production quality control, not novelty, is what separates a lab demo from a real second business line.” That is the key there. I’m gonna read this again. “Separates a lab demo from a real business line.” That is the difference there. They’re saying, “Okay, this is not a lab. This is a real automotive line.” I think Omega did a really good job of interpreting all the, the public documentation that Xpeng has put out there to the market

Rob Grant: 100%. I love the fact that it’s not a novelty, right? That’s what it’s saying. This is the mass production quality control, not a novelty. And I think that is the manufacturing readiness milestone. Again, it puts it together in such a beautiful way. you know, it’s like a beautiful mind. I, I, I can kind of see it coming together at times

Grayson Brulte: Russell Crowe was brilliant in A Beautiful Mind. What a, what, what a great movie about a world-famous mathematician. Well, Rob’s in Kalamazoo this week. We are ramping up field reporting. We got a lot more field reports coming, and even when we’re in the field and we’re doing field reporting, we’re gonna bring you autonomy signals each and every week. Rob will, and I will be here breaking down the signals in the market, and if you can’t wait, send a note to [email protected], alpha@autnmy, because we’re here to answer your questions and engage with you, and there’s big things coming from us. And a huge, huge thank you to KPMG for being a wonderful, wonderful sponsor and making this show possible. I won’t give you the PBS speech. All I’ll do is say thank you to KPMG. The future is bright. The future is autonomous. The future is Cybercab. Rob, another great show in the books, and we’re always learning something new

Rob Grant: That’s why I love what we do. I love learning

Grayson Brulte: And to the audience, never stop being a lifelong learner. It pays dividends