00:00And Peter, as a CTO, what does your technology do? Where are you taking us to?
00:05Yeah, so we provide the infrastructure layer for physical AI. And so the complete software stack
00:10that enables machines to move. Everything from cars to trucks to mining and construction equipment
00:15and then drones and boats and defense.
00:18And therefore, that's what's interesting, right? Is that we're at the intersection of Silicon Valley
00:22and government. We're at the intersection of, for you, defense and also commercial.
00:27How has that movement into the world of defense happened organically?
00:31Somewhat. I mean, we're passionate about the field broadly.
00:34So certainly it was a deliberate choice on our part to enter the field.
00:37But there's so much commercial technology that we've developed that really translates so well into defense.
00:43Just some examples. We do work with the Navy, deploying the same kinds of data collection technology
00:48that we use for self-driving cars on Navy ships.
00:51And we do work broadly across the DOW on simulation testing.
00:55We call that the virtual proving ground, which is also directly taken out of our commercial business.
00:59One of the themes of the Hill and Valley Forum, not necessarily just this year,
01:03over the number of years it's run, is how easy or not easy is it doing business
01:09with the military apparatus of this nation?
01:12And throughout the day, a number of founders have given examples where you see decisions made in days
01:18where previously it had been months.
01:20What's the experience been like for Applied in getting autonomous software stack integrated in various pieces of hardware
01:27that our military apparatus relies on?
01:30Things are accelerating. There's no doubt about that.
01:34I think that in our own case, we've seen similar things where things maybe have stalled for even a year
01:39or so
01:40that are really being unblocked.
01:41There is a real desire, I think, by the whole apparatus to bring dual-use tech out to the warfighter
01:47more quickly.
01:47Physical AI is hard for some people to understand.
01:52It is physical things, hardware in the real world, where your competence in software
01:58allows that bit of kit to move around a space and make autonomous decisions.
02:05What are some of the real-world deployments that you can talk about?
02:08Because I think there's great skepticism that it has moved beyond robo-taxis
02:15and actually droned technology in the context of weapons.
02:18Yeah, so certainly self-driving cars, that is the big first application.
02:23And so we have a ton of technology that's now on a bunch of companies all over the world.
02:27Can you name them?
02:29Our website, we work with 18 of the top 20 automakers.
02:33Everyone's using some part of our technology at this point.
02:36And we've been very deep in that domain actually almost for a decade now.
02:39Right.
02:39In the defense sector, though, there's a lot more focus on drones and drone technology.
02:44And so both the autonomous software that runs on drones
02:47and also thinking about the effects that you can have when you can swarm drones together.
02:52From your perspective, the ethical dilemma for companies making the technology
02:58and then the use in the Pentagon,
03:01how is that something that you've been able to bridge as a company, as a leader of the company?
03:05So we're a technology company.
03:06And so we focus on building the technology.
03:08We're not a political company.
03:11We're not a policy company.
03:12And I don't think that's actually our role to play.
03:14We want to make sure that we have the very best technology and that that technology is available.
03:17And then I think through our ordinary political process and democracy,
03:21that's how we'll actually set those policies.
03:23I always want to offer the opportunity when a company like yours come to the show
03:28to explain what I call core competences, what they're really good at.
03:32Cara and I, for quite a long time now, have been talking about Tesla being very good at autonomy,
03:38a vision-based approach where their software stack, they would say, is pretty good.
03:42NVIDIA is now wading into the physical AI space,
03:45not just with inference chips that are good at the edge,
03:49their own software and libraries and catalogs of custom models.
03:55Well, why is it that yours are better than theirs?
03:58So I wouldn't say that we're better or worse than theirs.
04:00We're acting as an infrastructure provider and we work with the entire industry.
04:04And so our core competency is we've built really great infrastructure technology for physical AI.
04:09And we're really great at working with the actual manufacturers of these systems
04:12to get that integrated and deployed.
04:13We're really not just looking at the technology development,
04:17but also actually getting this deployed.
04:19And the hard part oftentimes is actually getting it into the real world
04:22and then solving all of those difficult problems that happen
04:25when you're really making use of those systems.
04:27So it sounds as though you feel that there's a future of partnership,
04:32no matter what, with your technology.
04:34But how do you make sure...
04:35All the automakers are rubbish at software, which is a reasonable argument, right?
04:39But why are you not going...
04:42The anxiety often is that eventually LLMs will come and eat all of the other professional applications.
04:47From your perspective, what makes NVIDIA not able to recreate what you've built?
04:51Well, again, these are huge industries, right?
04:55If you think about just the automotive industry,
04:56it's like several percentage points of the global gross domestic product.
05:00And there are millions of people and employees in all of these industries.
05:04And so there's plenty of room, honestly, for many players.
05:07And I think no one company is going to dominate the entire industry.
05:10NVIDIA plays an important role with chips,
05:12but there's many other layers to the technology stack
05:14and actually getting this technology deployed.
05:16What I suppose the different domains have in common,
05:20RoboTaxi, in the physical AI space or robotics world,
05:23you know, the use of bipedal humanoid robots in the industrial setting.
05:28This country is still waiting for sort of thorough federal-level regulatory oversight,
05:33a list of rules.
05:34That's what we hear here, at least.
05:36Where do you stand on that and the need for it?
05:39I think for RoboTaxis, there is strong need for a federal policy.
05:43Right now, it is very state-by-state,
05:45and that's absolutely impeding progress, for example, in long-haul trucking,
05:50where you actually have truck routes that would go between multiple states.
05:52And if you have to plan your trucking routes
05:54based on where the laws allow driverless trucking,
05:57that really limits the market potential.
05:59I think likewise in RoboTaxis,
06:01the legislation would be easier to follow
06:04if it were a national framework rather than state-by-state.
06:06And I think we'll see the same thing in humanoids.
06:09Humanoids industry is earlier, but it's advancing very quickly.
06:13And I actually hope that we can come up with a single framework
06:15for all of these physical AI systems.
Comments