Humanoid robots are moving beyond prototypes, and Apptronik says its next generation will be the real thing.
In this episode, Apptronik CEO Jeff Cardenas shares how years of testing Apollo 2 have led to Apollo 3, the company’s first true product designed for scale. He explains why today’s humanoid robots are still largely prototypes, what it takes to build reliable AI-powered robots, and why Apptronik is investing in massive “Robot Parks” where robots build skills, generate data, and prepare for large-scale deployment.
Check out our conversation here, and don’t forget to subscribe to my YouTube channel:
We also dive into Apptronik’s collaboration with Google DeepMind, why the company is building both wheeled and bipedal robots, and how advances in AI are changing what’s possible for general-purpose robotics.
Finally, Jeff explains why robotics has become the new global “space race,” why vertical integration and domestic manufacturing matter, and how Apptronik is building the supply chain needed to manufacture millions of robots.
In this conversation:
- Why Apollo 3 is Apptronik’s first real product
- Why today’s humanoid robots are still mostly prototypes
- Robot Parks and the future of robot training
- Apollo 2 lessons learned
- Wheels vs. legs: which robots win where?
- The Google DeepMind partnership and Gemini
- Scaling AI-powered humanoid robots
- Robotics, manufacturing, and the global technology race
- Why actuators and vertical integration matter
Transcript: Apptronik CEO on Apollo 3, Robot Park, and Google DeepMind
Jeff Cardenas: Apollo 3 will be a product. It will be a mature, early product, but still a product. I would argue the industry today is essentially all prototypes.
I think the last time we chatted you had raised a lot of money, and now you’re spending a lot of money.
Jeff Cardenas: Yeah. That’s a good summary.
John Koetsier: Tell me what you’re doing. Apollo 2, Apollo 3—not announced, but kind of implied. Robot Park, and not just one, but Robot Parks all over the world. What’s going on?
Jeff Cardenas: Yeah, so we’re scaling up now.
The Robot Park is really the embodiment of a dream we’ve had for a long time. When we first started, I would go get beers with my co-founder, Nick, and dream up this idea of robots building robots on the Colorado River. I called it Willy Wonka’s Robot Factory. That was kind of the dream.
And we said we would have a—
John Koetsier: I want a golden ticket.
Jeff Cardenas: Yeah, yeah. You’ve got one. Come on and come see it.
Just like we have a factory building robots, we also have a data factory as well. It just so happens that when you see a lot of robots doing different activities in a confined environment, it’s a pretty cool place to come visit as well.
So there’s a dream long term about making it more accessible to the public to come see it. We’ve had Apollo 2 since February of last year.
John Koetsier: Mm-hmm.
Jeff Cardenas: I don’t know that I would wait another 18 months to launch a robot again because I’ve gotten a lot of questions about Apollo 2 over the last year.
We wanted to make sure that we actually had it scaled up enough and really performing well before we started to show it off. That’s what you’ll see over the coming months and the rest of this year: the results of the early work that we’ve done on Apollo 2. Because we’ve had it since February of last year, that was really the beginning of our work with Google DeepMind.
We’ve learned a ton about what it takes to build a state-of-the-art AI-powered humanoid robot, how to make it reliable, and how to make it easy to service, support, scale, and build many of them. All of those lessons are now going into the next version of Apollo, which will really be the product.
You can think of Apollo 2 as really a prototype for scaled pilots and data collection, and the next version of Apollo will really be the product that will scale to the market.
John Koetsier: You’re doing both a bipedal and wheeled version, and I think that makes a ton of sense. A lot of the people that I’ve chatted with have said, “Hey, our industrial customers or logistics customers are saying we want wheels.
“We want that stability. We want that longevity, battery life,” and a bunch of other things like that. Is that what’s going on there? If so, what’s the bipedal version for?
Jeff Cardenas: Yeah, I still think that the bipedal humanoid has the highest ceiling in terms of what it can do, so it has the broadest workspace that it can perform inside of. But there’s a challenge with bipeds in that they can fall over.
When you start to think about really scaling these systems into the tens of thousands, millions, or billions, depending on which entrepreneur is talking about it, there are real challenges that we need to solve and make sure we get right in order for bipeds to be successful.
We’re really focused on how to solve safety and how to be smart about the ways we solve safety for bipedal humanoid robots.
There are many applications that don’t require legs, and many things where wheels will simply be superior. If you’re transporting goods over a long distance, humans like to use wheels for many of those applications.
We can also get longer battery life. What we believe is that we want to build general-purpose robots, and the view for Apollo was that by having both wheels and legs, we get the best of both worlds. We can scale data collection on either platform. We can get big fleets of wheeled systems out to pave the way for the bipedal systems as they mature and come online.
We’re actually piloting both today, with both wheels and legs, and there are different applications for each. I think in manufacturing and logistics you’re going to see a lot of wheels, especially early on. As you start to look toward future markets, I think that’s when the bipeds will really shine.
John Koetsier: Future markets—that’s interesting. Homes obviously often have stairs and other non-traditional spaces, maybe even offices. I’m not totally sure about that. In some cases there’s usually an elevator.
That’s interesting to see.
The data is critical, obviously. There’s no one way to attack the humanoid space. The data is critical. That’s why you’re building out the Robot Park, the data park. Ninety thousand square feet, I believe you mentioned, and there are others around the globe that are planned or coming online.
How many robots will you have in there learning, training, and doing things?
Jeff Cardenas: We haven’t released the specific numbers of the Apollo 2s, but lots of robots.
We’ve learned a lot of lessons about scaling up even just a prototype like Apollo 2. All of those lessons are the backbone of what we’re doing with the next version, Apollo 3.
John Koetsier: Cool. Interesting.
It’s interesting that you had Apollo 2, but you didn’t really talk about it—just kind of a quiet website update, and boom, there was Apollo 2, and a few people noticed.
It’s also interesting that you’re honest. You’re not going to deploy 10,000 Apollo 2s in a manufacturing facility. Pretty much every humanoid, I think I’m correct in saying, is essentially a prototype. Even if it’s shipping, even if it’s doing work today, none of them are what you would consider done—something that replaces a human or performs 80% or 110% of what a human might do.
Apollo 3—what do you expect it to be able to do?
Jeff Cardenas: Apollo 3 will be a product. It will be a mature, early product, but still a product.
I would argue the industry today is essentially all prototypes. I do think there are a few early products that we’re now seeing enter the scene, but that’s after years of iteration on these platforms.
Apollo 3 will just be much better. It’s much more scalable and much more focused on BOM cost. We’ve done a lot of work on the end effectors and the sensor stack on the robot. How we solve safety is going to be unique on Apollo 3, so things like safe perception are really important to us.
It’ll be a pretty significant step forward in terms of what we’re doing—an order of magnitude more resources and team than any of the robots we’ve ever built in the past. We’re taking everything we’ve learned and putting it into building the product that we’ve always wanted to build.
John Koetsier: Cool. Talk about Google DeepMind a little bit.
We’ve already talked about the data a little bit. You have a partnership there. You’ve been working on it for some time. They’re helping you, you’re helping them, and you’re feeding them data.
Will some of that potentially go to competitors if Google DeepMind gets more and more insight from the data and gets better and better?
Jeff Cardenas: We’ve had a great partnership with Google DeepMind. They’ve been both an investor and a collaborator with us for approaching two years now. It’s hard to believe that much time has passed.
We’ve learned a lot together on both sides. The goal of the partnership and the collaboration was to push forward the state of the art.
What does an AI-powered humanoid look like, and how do we advance the state of the art on both the hardware and software side? How do we think about things like safety and doing this responsibly to get it out into the world in a big way?
We’ve learned lots of lessons. Google has been honest with us from the very beginning about their ambitions to build the Android for robotics. That means their model will go to many different partners.
But we have this deep collaboration with Google, where we’ve been in the trenches for several years building and advancing Gemini. We like to think we have a special relationship, and we’ll see how that evolves in the coming months and years ahead.
John Koetsier: If what they do in robotics is anything like what they’ve managed to achieve in smartphones, that’s a pretty exciting opportunity because almost everybody is trying to solve the whole widget themselves.
So many people are trying to build all the hardware, all the software, and all the AI layers. I can’t argue that it’s a bad strategy to have a super-large, super-successful, super-smart, and super-motivated global giant helping you do that and taking some of the product development burden off your hands.
Jeff Cardenas: Our view is that we’ve always been very ambitious about what we want to achieve. We think we have an opportunity to build one of the biggest companies in the world.
But we’ve been in the game long enough that we’ve been humbled by the challenge many times, so we’re pragmatic about how to get there.
That means we want to work with the best minds in the world, and we want to collaborate to advance the state of the art. We’ve found that in Google.
They have some of the best researchers in the world, and we’re also mission-aligned in ensuring that this is a positive thing for humanity. We’ve been collaborating very well together for a couple of years now, and I’ll be excited to show where all that work has led.
John Koetsier: Excellent.
Maybe one last thing, and this isn’t specifically about your announcements today, but we do see increased emphasis on geopolitics in and around robotics and humanoids.
One aspect of that is potentially vertical integration and sourcing. I saw that one of the elements of your announcement today was actuators.
You’re using your own actuators, building your own actuators, and you’ve patented them in multiple ways. You’re pretty proud of what you’ve been able to accomplish with your actuators, which is a huge part of the robot. That’s, what, half the cost of a robot, or somewhere around there? Half the component cost at least.
You mentioned BOM costs earlier as being critical for Apollo 3.
Are you vertically integrating in other ways as well, and are you able to control, maybe even nationally source, your required components?
Jeff Cardenas: I think you’re going to hear more and more conversations about this.
My view is that modern robotics is the space race of our time.
The best minds across the globe are all going to be competing and working on this to push the field forward because whoever wins this will be a big part of the future, both for national competitiveness and national security. All of the major powers are going to be heavily invested in this.
Here in the U.S., we need to make sure that we’re playing an important role. We need things like a national robotics strategy, and we need to build up the supply chains that are going to be necessary to deliver millions of robots in the future.
This is something I’ve been focused on since graduate school. My view was that no one makes robots in the U.S. Certainly that’s going to be a problem.
It’s been funny watching the conversation evolve because now everyone’s talking about actuators.
My co-founder, Nick, earned his PhD in advanced actuation for these kinds of robots. At Apptronik, we’ve built nearly eighty iterations of electric actuators—all sorts of different variants and technologies: rotary, linear, liquid-cooled, quasi-direct drive, series elastic, tendon-driven.
You name it, we’ve probably tried it over the years. We’ve built a variety of different robots from those technologies, really learning from first principles what’s required because we viewed actuation as so important.
For the people who are worried that we don’t make actuators in the U.S., I think we have one of the best actuation teams in the world right here in Austin, Texas. I think my co-founder, Nick, is probably one of the best actuation experts in the world as well—certainly one of the most experienced.
As part of this space race, the supply chain is going to be really important, and we think Texas is going to play a critical role in helping build the supply chain for North America.
You’ll see more announcements later this year about us really putting our money where our mouth is and making concerted efforts to build more of that supply chain here domestically and locally so that we can deliver on the promise of robots here in the U.S.
So yes, vertical integration is very important, and I think vertical integration is something that we’ll earn more of over time. As the company matures, we can be smarter about which battles to pick and ultimately control more of our destiny as we move forward.
John Koetsier: Jeff, thanks so much for this time.
Jeff Cardenas: Awesome. Thank you.
If you’d like, I can also apply a second editorial pass that preserves every word but makes the transcript read even more naturally by smoothing disfluencies (while still not removing any substantive content).