Groq's Revenue Woes, Waabi's Driverless Trucks, and AI Healthcare & E-commerce Agent Battles | July 30, 2025

0:13Welcome, everyone, to the Informations TI TV. My name is Akash Pasricha. It is Wednesday, July 30th, and we have got Meta and Microsoft reporting earnings tonight. We are going to get to that. I should say before we start, Mark Zuckerberg just released a clip this morning with some more color about how he is thinking about superintelligence. You can find that on our social feeds. But today on the show, we have got the founders of two of the fastest growing autonomous trucking and AI healthcare companies, two separate companies, I should say. The founders of those companies are coming on the show.

0:47We are also talking about a big scoop from our finance team about the revenue projections of a big startup taking on NVIDIA. And we are closing out the show by talking about the AI agent wars in the land of e-commerce. Now, before we start, I want to talk about two big headlines coming out of our own newsroom. Cohere, the Canadian AI company competing with Anthropic and OpenAI. They are talking to investors about raising money at a$6.3 billion valuation. That is a 25 % jump from the last valuation it got a year ago. The company expects to generate more than$200 million in annualized revenue by the end of this year.

1:27And the second funding scoop that we had is that crypto exchange Kraken is looking to raise$500 million at a$15 billion valuation. That is up from the$11 billion valuation that it fetched back in 2022. Of course, crypto has changed a lot in three years, but that scoop comes from our Dealmaker newsletter that goes out on Tuesdays and Thursdays. So be sure to check that out. A couple other headlines for you before we get started. Palo Alto Networks is buying identity management company CyberArk for$25 billion. Palo Alto Networks hasn't traditionally done as much in the software space related to logins.

2:05Okta, Microsoft, and Cisco. These are some of the companies that have. This is clearly a play from Palo Alto to start diving into that login space. So we will see how that shakes out. I mentioned Meta and Microsoft are reporting tonight. Big thing to pay attention to with Meta. I mentioned Mark Zuckerberg, his comments this morning. Whether or not he says anything about all the gobs of money that he is spending on i, that will be the thing to listen for on the call tonight. Microsoft, we said it yesterday, but it's all eyes on Azure. Azure growth accelerated last quarter to 33%. And so if it can top that, that will be something worth paying attention to.

2:42Okay, let's get to our first guest. We have written a lot at the information about how chip companies are vying to take on NVIDIA. One of the best known startups in that space is Grok, not to be confused with XAI's Grok chatbot. This is Grok with a Q. The company has been in talks to raise money at a$6 billion valuation, But we were first to report this week that the company has had to revise its revenue projections. And I want to bring on Myles Krupa, the reporter who broke that story, to tell us more about what he's found out. Myles, welcome to TITV. Thanks, Akash. So tell us, what did you learn about Grok?

3:22Well, what we learned really is that after having given out these fairly high revenue projections early in the year, they said that they were expecting a bit over$2 billion in revenue this year. That more recently when fundraising, they've been telling investors they're expecting closer to$500 million in revenue. And so that's a very big difference. And we tried to get to the bottom of it. And what the company ended up telling us is that they shifted some of their revenue expectations to 2026 because of a requirement that they deliver their chips to certain data centers in a certain region. but basically that the data centers weren't ready to accept the chips.

4:13You know, Grok recognizes revenue upon delivery. And so because of that, they have pushed back their revenue projections into next year. So they're telling you that it's kind of a capacity issue, that the data centers are what's holding them back from getting that revenue. Is that right? That's right. And while they didn't specifically point to Saudi Arabia as the reason for this, the context basically is that in February of this year, around the time they gave out the projections, they had signed a big new contract with Ramco Digital, which is part of Ramco, the Saudi oil company. You know, Saudi Arabia is undergoing a major push to open data centers and go big on AI.

5:04But, you know, there are lots of competing initiatives in Saudi Arabia right now. They're spending a lot of money on Neom, for instance, their big futuristic city in the desert. And they've also started importing chips from AMD and NVIDIA. And so there's just a lot going on right now. in Saudi Arabia. And that's sort of been the cornerstone of Grok's business. So, you know, while Grok wouldn't say that Saudi Arabia was to blame here, you know, it's possible to draw the conclusion that something's going on there. I want to come back to Saudi Arabia and sort of the demand element. But, you know, let's talk about Grok more fundamentally.

5:44I mean, this is one of many NVIDIA challengers. Grok has gotten a lot of funding. What makes Grok different? And why are investors so bullish on the company? Yeah, there are a few different companies that have raised a lot of money trying to develop AI chips that can compete with NVIDIAs. What's interesting about Grok, this was a company founded about a decade ago. Jonathan Ross, the CEO, worked on Google's internal AI chip, the TPU, which is widely viewed as a huge success. And what they've come up with is a sort of different architecture to what NVIDIA does. And Grok is really focused on inference.

6:28So they're focused on trying to run AI models as quickly and cheaply as possible. The other thing to know about Grok is that its chips really work in systems of sort of hundreds. So you have to link together a few hundreds of them to get the performance of a few NVIDIA chips. Once you do, you know, the company claims they run a lot faster and a lot more efficiently. So, I mean, let's come back to sort of the revenue projections. This is a pretty big change in my view. I guess what I'm wondering is, is it common for companies to change their revenue figures so drastically in this space? you know i wouldn't say it's super common um clearly hardware companies have a different set of challenges than software companies do you know the marginal cost of building a new unit of software is close to zero whereas hardware has all these questions about supply chain and in the case of a company like grok clearly are there physical um buildings where you can put your hardware that are ready to accept the chips.

7:44You know, I will say Grok is currently fundraising. You know, when a company is fundraising, usually they're trying to tell investors that we beat expectations, that we're ahead of projections. So, you know, I think it's safe to say that this is not ideal for Grok. You know, another thing that this kind of reminded me of was the SPAC boom of a few years ago, you would see companies come out with these very optimistic projections to entice public investors, become public companies, and then frequently miss those revenue targets or sales targets. And so it just feels like that might also be another sort of example to point to of, you know, where we might see this kind of thing.

8:35Yeah. And look, I think it's a good point. You mentioned about this sort of being similar to the SPAC boom. Let's take a step back a bit. I mean, your beat at the information is covering AI and financing. And so when you think about that and this context in the story of that, I mean, what kind of themes are you focused on right now as a reporter? Yeah, you know, the sort of big question is basically, where's the money for AI going to come from? And I think it's clear just in the few months that I've been here that the Middle East is going to be a huge part of that. Saudi Arabia ramping up its AI efforts with a new company called Humane.

9:17They're starting a$10 billion venture fund. We know they're in talks with Grok about investing. um you know i think uh the sort of rising stature of the middle east in ai not to mention abu dhabi um that's a huge theme something i'm going to be watching closely um and then i think also um just more broadly this kind of question of whether nvidia's dominance in ai chips can continue and for how long. With companies like Grok growing fast and winning new customers, with companies like AMD trying very hard to get their chips in the hands of more AI developers, helping to start companies like Tensorwave, which is sort of the core weave of AMD.

10:11These are just some really interesting themes And I'm going to be, I think, spending a lot of time on chips over the next few months. Before I let you go, you know, we talked about how Grok is sort of, you know, framing this as a supply issue or rather a capacity issue, I should say, in terms of its ability to, you know, deliver. I just want to come back to the demand side because you mentioned the contract with Saudi Arabia. It's a$1.5 billion contract. You know, that was supposed to be conceivably a large part of that revenue projection that it had initially put forward outside of the Saudi contract.

10:49What kind of a sense are you getting in terms of what demand customers have for Grok's chips? Yeah, it's interesting. There are kind of two different prongs of the business. Basically, Grok had almost no revenue in 2023. It was not having an easy time selling its chips directly to other companies. And so it created this cloud business that lets you run open source models like Llama on Grok hardware. And so that business is growing nicely, although it's not profitable. So they have some customers like Dropbox that are using open source models on Grok hardware through that service. And then they're doing these kind of partnerships or direct sales like they're doing with Saudi Arabia.

11:43Another one is Bell Canada, which is building some data centers with Grok chips to serve Canadian customers. um you know and then um grok is also just going out and uh filling data centers with uh grok chips in anticipation of bigger demand you know they just built a big data center in finland i believe um so uh they're moving forward full steam ahead anticipating there will be great demand for their chips well great look it's it's a fascinating story to cover and i will say that you know, it's not just Grok. We've got a database of all the AI chip companies that are taking on NVIDIA. And so we will link that in the show notes because, you know, there's companies like Sampanova, like Cerebrus, we have a scoop about them.

12:36Smaller companies like Dmatrix, all these companies are out there fundraising right now. And so I look forward to having you back on the show, Miles, to talk about those companies. Thank you so much for being here with us. That is Myles Krupa on our finance team here at The Information. Now, autonomous vehicles have been all the rage lately. You've got Waymo, you've got Tesla, Uber is running its own tests right now through partnerships. There is also the arena of autonomous trucking, which itself was a fast growing space that has gotten a lot of funding. Of course, that space hasn't come without hardship.

13:09But I want to bring on someone who has been in the arena for nearly a decade. Raquel Ortasun is the founder and CEO of autonomous trucking company Wabi. The company is backed by Coastal Ventures and Raquel's former employer, Uber, they have raised more than$280 million. Raquel, thank you so much for being here. Thank you for having me, Akash. So help me understand this, Wabi designs software that you then install into trucks that then makes them autonomous? Yeah, the way to think about us is that, you know, we are revolutionizing the physical world with a new generation of AI technology. And as it relates to self-driving trucks, it's our first vertical, where the OEM for us, Volvo, provides the base platform, which is the truck that is ready for autonomy.

14:01And then we install, you know, sensors, compute, as well as all the software, our proprietary software that does drive those trucks autonomously. And are these trucks on the road yet? Yes. So we've been doing commercial operations with our cell driving trucks for over two years. We are in Texas, which is kind of ground zero of cell driving tracking. Right. And what you're going to see by the end of the year is Wabi trucks with no human on board. By the end of this year. By the end of this year. And I started the company only four years ago, which is amazing to see the progress to date. And just to be clear, so when I say on the road, so these are just tests that are happening in Texas or they are literally roaming the highways right now?

14:49Yeah, so they are roaming the highways, you know, doing commercial operations with some of the biggest, you know, shippers and carriers. So really exciting times for us. And yeah, just gearing up for, you know, true commercialization, which is when you don't have, you know, any human on board. So when it's truly helpful. And what kind of companies are you selling these trucks to right now? So, what we have made public is our partnership with Uber Freight. This is a very, very big partnership with billions of miles of committed deployment of the Uber Freight network. And we really like this because Uber Freight sits in the middle of supply and demand when one third of the Fortune 500 companies can actually get access to Wabi driver technology.

15:38We have other very exciting customers. We just haven't talked publicly about them yet. I want to talk about how you are testing this technology because it strikes me that it's a big lift to put trucks on the road to test them in a safe way. You guys came out with this new product a little while ago. It's a mixed reality truck simulator is the way I understand it. Help us understand what this simulator is that you built. Yeah, yeah. So when, you know, Wabi has been, you know, has brought to market very differentiated technology within simulation-first company. And what we just announced is something we've been doing for two years, which is not only super important for safety, but it's also, you know, the coolest technology out there.

16:26And the idea is that we can actually have our cell-driving vehicles have the equivalent of augmented reality goggles so that when they are driving in the physical world, they are actually seeing a world that is not reality, but it's a mixture of reality and virtual. So that, you know, they get exposed to all sorts of safety critical scenarios and things as you are driving in the physical world. We do this in a closed course. And this is a massive unlock in terms of being able to test the safety of the system, being able to assess whether we can do things like avoid unavoidable or reduce, you know, for example, the effects of unavailable accidents, make sure that, you know, we kind of test the safety of the systems to the limit.

17:14Help me understand. I mean, like, you know, I think it's a neat idea, this virtual reality or mixed reality concept, I should say. I mean, what is the advantage of this as opposed to like just testing it on a closed course, like you said, and having it go through simulations, obstacles, you know, putting, I don't know, like a garbage can on the road or something. Yeah, yeah. So if you look at what traditionally has been doing, has been done in the industry to date, is that you basically, you know, engineer a handful of tests, typically one dozen, two dozen tests that you do, you know, where it's really like creating a movie stunt where you see teams with walkie-talkies saying, you know, oh, we're going to test, you know, a vehicle is slowing down in front of us, then go and test this, which is not scalable, it's not diverse.

18:07You cannot really push the safety limits because they are real vehicles or real pedestrians, so consequences of if the system cannot handle it could be tremendous, right? But this is what everybody does in the industry today. That's the equivalent of the safety test for self-driving. So as you can imagine, you know, this is absolutely not sufficient. And then what people do is like, let's go to public roads and let's, you know, crank miles to see whether you can handle certain situations. But obviously, you know, the red events, the safety critical, you will need, you know, hundreds of millions, billions to expose, you know, to even see those events once, right?

18:49So you don't have a good way of testing. with this augmented reality with no consequences, we can test all those things in the physical world. So it's a massive unlock. So it's cheaper. It's much cheaper. It's cheaper. It's much more efficient. For the first time, you can test things that was impossible before. And again, with no consequences in terms of the safety of anybody around. And that is a massive unlock on the industry. And definitely since we have showcased this, uh it has created you know a lot of uh you know noise and excitement i would say uh you know across across every competitor right oh i want to take a step back here you know you came from uber where you basically helped lead the company's autonomous driving efforts um of course we know that those self-driving efforts were sold to a company called aurora innovation um you know they are also doing autonomous trucking we'll we'll come back to uber's role in a second i find it interesting that they're your customer and, you know, they also, you know, they sort of sold their stuff to Aurora.

19:54But let's focus on you versus Aurora. I mean, what makes Wabi different than what Aurora is doing right now? Yeah, yeah. So, you know, the main difference is not on the business model that is similar, it's really on the technology, right? And four years ago, I started Wabi because I saw a tremendous opportunity in the industry to build a disruptor. And the idea was that, you know, to have very different technology rooted on really addressing the key pain points of the industry. And this was, you know, two, I would say two theses. One was that, you know, we need to really unleash the power of AI and build an end-to-end system that is verifiable and that is able to reason.

20:38This was before, you know, Tesla or the such were doing any end-to-end approaches. And the second was that there is always going to be a data problem, so we need to build a simulator that is the same as reality, the matrix for cell driving. But, I mean, I just want to come up, I mean, why is Wabi's technology better than Aurora's? I mean, Aurora is a traditional approach that is a hand-in-genier system that is very unscalable, that has really difficulties in terms of how are you going to expand that geographically in terms of every bit of progress is harder and harder and harder. And if you look at, And it's very capital intensive, right?

21:25Look at the amount of capital that has gone into it and look at where you are in terms of milestones, et cetera. If you take Wabi's approach, the advantage is both on the development, which is much faster, as well as much more capital efficient, while we are, what, one-twentieth of their capital needs, right? But more importantly, our product is better, much better. And this is because we don't need terminals, thanks to our differentiated approach. We can go to the end customer from day one. That's massive in terms of the economics and the adding an extra complexity that you need with drainage.

22:02We can expand geographically much faster. We can expand to different OEMs, four factors, really, really easy, potentially other verticals. And since we are much more capital efficient, we can actually have much better pricing and the product is safer. and we can assess the safety of the system much better with our technology through simulation and this makes reality testing. So not only is a development advantage, but also the product is really something that others much better, the customer needs, and can be scaled much faster as well. So it's a very interesting time now in terms of how the two companies and our trajectories are crossing.

22:40Well, and I was going to say, that's the trucking space. But, you know, if we look at sort of the autonomous passenger vehicles space, right, you mentioned we've got Waymo, we've got Tesla, Uber is now doing its own test as well. I mean, let's keep trucking aside for a minute. I would love to get your take on sort of, you know, the passenger vehicle, autonomous vehicle space. How do you think about that market right now and the different players that are there? I mean, what is your sense about, you know, who you think is going to succeed? Tesla has made a huge bet on it. I mean, who do you think wins there?

23:15Yeah, so I will say that, you know, I expect not just one winner, but a small handful of winners. I think it's been great to see, you know, the deployments on the scaling of Waymo. I think, you know, before there was a question about will customers want this product versus not, that is not anymore that question. And I think that you see as well, you know, right selling market also betting, you know, in definitely big in sell driving because that's the future, right? I think, you know, as you think of who can enter this market and be a real contender, I will boil down to you can't just try to do the same as Waymo and then try to catch up.

24:03I think that ship has sailed. I think that what you need to do really is enter with a next generation of technology where capital efficiency is scale. And there is where you can actually really, you know, be a contender for that, you know, massive market. This, you know, from my perspective, it's as big as, you know, the self-driving tracking market as well. So that's sort of like what I think will happen. We will see over the next couple of years, but it's very interesting to see both robot taxis as well as trucks. Yeah. Well, like you said, it's a very interesting time in both of these spaces.

24:49You've got a very interesting approach. The mixed reality approach is, I got to be honest, I never would have thought that we would have seen autonomous trucks in mixed reality simulation. It is not an intersection I thought we would have seen, but thank you very much for coming on and explaining it to us. And when more of those trucks are on the road by the end of this year, we're going to have you on and we're going to do an interview with you in the truck while the truck is driving. I know you said it's driverless, but you're going to be FaceTiming us from there and we'll do another interview from there.

25:22How about that? Yeah, I think, Akash, you should come and also experience mixed reality on our driverless trucks so that it will blow your mind. You will see it's amazing to see this life. But happy to showcase our technology life. Great. Well, thank you, Raquel, for coming on. That is Raquel Ortasun, the founder of Wabi, the autonomous trucking company. Now, for our next guest, AI for health... Let me pause. AI for healthcare is a booming industry, I should say. We have seen a ton of companies raising money in this space, not the least of which was Ambience Healthcare. The company raised$243 million this week at a$1.25 billion valuation.

26:06It was led by Andresen Horowitz and by Oak HCFT. I am told that that stands for Healthcare and FinTech. That's a good acronym. I want to bring on Ambience Healthcare's co-founder and chief scientist Nikhil Baduma to talk about why he thinks his company will win this race. Nikhil, welcome to the show. Thanks for being here on TITV. Thank you for having me. So you guys have this neat platform. I mean, the AI listens to the appointment. It takes all the notes and it produces the chart for the doctor. The doctor can ask questions. So all this money that you raised, I mean, is it to make those features better or are there new products that you're working on?

26:47Yeah. So the platform today, the goal is to really address the holistic clinical workflow for a physician. So that starts before the clinician even walks in the room. It actually looks through the electronic medical record, summarizes the past notes, the labs, the fax documents into a single dashboard that they need to catch up on the context of the patient. And like you said, during the visit, it's listening in the background. It's automatically generating the documentation for the clinician and also a summary for the patient and their family so they understand the care that's happening. And then afterwards, it turns out that healthcare is full of these administrative workflows downstream in revenue cycle, coding and billing prior auth, which is how health systems get paid.

27:30And our technology automates and simplifies a lot of those administrative burdens as well. And so the goal of the platform is to solve for the clinical burden holistically. So clinicians are liberated from the keyboard and they can really just focus on taking care of their patients. I think for us, this funding round is for a couple of things. The first is a big part of how we've been able to do what we do is investing incredibly deeply into the foundation model layer. We work very closely with folks like OpenAI and other foundation model makers to create the most capable models in healthcare and medicine.

28:01And so this is going to be an opportunity for us to double down and continue to make sure we have the most capable clinical reasoning models in the market. The second component is we're just seeing an incredible amount of growth. And as you can imagine, healthcare systems need a capable partner on the other side to help them implement and actually realize the full value of AI across their organizations. And so this funding is going to help us make sure that we are able to staff up our care transformation team to help organizations realize the full value of the AI technology. How many doctors are using your product right now?

28:32It's rapidly growing. We work with about 40 plus different healthcare institutions, including folks like Cleveland Clinic, UCSF, Houston Methodist. I'll give you an example. Cleveland Clinic, 80 % of all their clinicians use ambience every single day in clinic. They use ambience for 70 % of their visits. And so... So it's on in the background for 80 % of the physicians at Cleveland Clinic. That's right. It's transcribing. It's taking the notes right now as it is. That's right. And you walk, if you're in San Francisco, you walk into UCSF, chances are your clinician is using ambience in the background as well.

29:06And I mean, you know, did these... So after the... I'm just imagining how I use AI. Presumably, these doctors have to go in and edit these notes after, right? I mean, there must be some hallucinations involved or errors or stuff like that. I think the clinician is indeed responsible for the fidelity and integrity of everything that ultimately ends up in the medical record, that's for sure. But the level of precision and accuracy that we've been able to build into these models across the 100-plus different subspecialties that are at an academic medical center like like a UCSF or a Cleveland Clinic, we're getting to a place where we're taking the time that clinicians are spending after hours charting from two, three hours a day, pulling them down to maybe 20, 30 minutes in total.