Daniel Abreu Marques: Welcome to Autonomy Insiders, the show where global industry leaders in autonomous driving unpack their real-world insights.
George Kalligeros: we build robotic pit stops for self-driving cars. Effectively this is distributed depot footprint directly within the heart of the operating zone. What the vehicles need is a very dense, very widely on demand available network directly in the pockets of demand, such that You cannot just offer pit stops in the center of the city on demand. You can do micro resets. Okay? Because this is where the real economic unlock is.
Daniel Abreu Marques: Welcome to Autonomy Insiders. I'm your host, Daniel, and most conversations about autonomy are about the driver, the software, the sensors, the miles. This episode is about the other part cleaning, inspection, getting a car ready to earn money again. And my guest is building a robot that sits in a single parking space and does all of that, so the car never has to leave the neighborhood it is earning in. George Kalligeros is co-founder and CEO of Aseon Labs and Aseon came out of stealth in May and announced a ten million dollar seed in June. So George, welcome on the show and I'm so happy to have you on today.
George Kalligeros: Thanks, Daniel. Thanks so much for inviting me.
Daniel Abreu Marques: So before we go anywhere, maybe in plain terms, so what is Aseon and what does a fleet operator actually get from you?
George Kalligeros: Yeah. So at Aseon we build robotic pit stops for self-driving cars. Effectively this is distributed depot footprint directly within the heart of the operating zone. So an AV operator gets from us powered ground and fully robotic operations as a service.
Daniel Abreu Marques: And can you also describe this this pod as a physical object? So maybe if I would stand next to it in a parking lot, so how big is it? What do I see? Maybe also what moves?
George Kalligeros: Yeah, for sure. So as you described, it very much looks like a box and it occupies one typical charging bay, right? So the product in European terms is about six meters long and two point seven meters wide. And you don't really see much from the outside. I'll I'll share with you a photo that you can pull up. This is the the latest version of the product that we're building currently in Manlo Park, California. It's gonna be right in October. you can peer from the inside and you'll see a robotic system. which uses two robotic arms with six degrees of freedom that pick from a range of custom tooling to do a bunch of interesting things on the autonomous vehicle.
Daniel Abreu Marques: And then from the car's point of view, so it pulls in and then and what so roughly how long is it in there before it's then back taking rides? yeah, maybe high level for now we can get into the mechanics later, but just to get a get a picture.
George Kalligeros: Yeah. So the concept of the product is to compress as much of the depot workflows in a single box. That's why it's called a micro depot. So the car pulls in, the first thing that happens is after an inspection, the vehicle gets plugged into charge and then a sequence of events takes place that takes anywhere between twenty to thirty minutes. the main thing that drives the time inside the pod is the charging rate and how long we keep the vehicle into charge. every single other operation that we do effectively falls within that window. So yeah, this is this is the core.
Daniel Abreu Marques: So now we have a good picture. And now I want to understand what that replaces. maybe walk me through what happens today without you and your product. So a robotaxi, for example, in San Francisco needs a reset at two in the afternoon. So what does that actually cost the operator? Maybe also in time and probably also miles.
George Kalligeros: Yeah, for sure. I mean typically the primary driver that takes vehicles out of service is that the vehicle operates and it it falls below a threshold, typically state of charge, right? So the vehicle runs low on battery, maybe it's twenty, twenty-five percent battery, and then it needs to call into either a maintenance hub, a depot, or a pit stop location, which are sort of like smaller sort of rapidly deployed locations near the operating zone. San Francisco actually has good infrastructure to support Robotaxi Ops and it's a relatively compressed city. So maybe you're depending on where you're in the city, maybe within three to five miles, the car can drive to a depot and then gets charged, right? And serviced. The service typically takes about I'd say anywhere between 40 minutes to 80 minutes. Let's say the hour is the average, and then it might need to travel another 40 minutes back and forth. So typically the vehicle is out of service. let's say two hours with every sort of recharge operation, but that happens multiple times in a day.
Daniel Abreu Marques: Makes sense. And you and also I think your co-founder, you visited also AV depots while researching all of this. So what did you see there that you did not expect?
George Kalligeros: Yeah, we we did. I didn't I don't think we saw anything that we did not expect. they ran efficiently, they were well equipped, they're relatively new, right? I think what became strikingly obvious is it how difficult it is to run human operations and how difficult it is to incentivize the human operators to to effectively carry out the repeat high frequency tasks in the depot. And Yeah, this effectively kicked off a lot of the inspiration for why we are now building with robotics.
Daniel Abreu Marques: them I think you said also that roughly a third of a fleet is offline at any given time. So is that also a number operators recognize and accept or is it also more a number that surprises them maybe when you put it in front of them
George Kalligeros: Yeah, so I mean these figures were pulled from a from a public report that I think it was Harry Campbell basically doing some analysis on the Waymo fleet a year and a half ago. But the operators have a really, really good grasp of their core metrics. And that number came from what what's the percentage of fleet that could be servicing customers but are currently either traveling to the depot or being serviced or they're queuing or they're in inspection. And cannot deliver a ride at any point in time. we crunch a lot of these numbers to kind of understand how like what is the effect of the locality of the depot. If it's far from the city, what is the effect in terms of the utilization of the fleet? Because there is this antithesis where today's model of a depot which is very human-centric, it's all about high throughput and turning the fleet around and trying to compress as much of the servicing in the late hours of the day, whereas the fleet Pretty much just needs to be constantly operating, utilizing, and capturing rides
Daniel Abreu Marques: And before before Aseon you also spent four years at Tier also deploying hardware into cities at scale. So what did that period teach you about operational costs that you could not have learned from the outside, so to speak?
George Kalligeros: Operation cost. Tier was a very interesting case study in general. It was a company that transitioned from a growth phase to an operational, sort of like an economic sanity and a profitability phase really quickly. In terms of operating costs, I think the biggest learning was Tier and all the companies at the time had very much their own in-house operations, for everything from maintenance, rebalancing, charging. And I think Lime innovated on this model where they started outsourcing operations to local entrepreneurs, gig economy workers that were much more highly and better incentivized to to turn the fleet around. Right. So I think the learning there was doing a really sort of like deep and vertical ops structure it's hard to incentivize the people that are really working with the assets. Yeah. I think that that was really the key learning. In terms of cost.
Daniel Abreu Marques: Interesting and maybe now let's let's open the box. So I want to know which parts of this are resolved and which was maybe also are still hard right now. So there's charging, there's exterior wash, interior clean, lost items, sensors inspection, data sync. So that is a lot of different jobs for one machine, so to speak. which of those is working today and which is maybe also still an open
George Kalligeros: Yeah, yeah, it's a good question. I'll preface this by saying we should define what a solved problem is because just in robotics, getting to a working demo, getting to a ninety nine percent maturity of a feature is let's say ten percent of the work, right? so we can give you a working demo, the work really is in the last one percent, which is being able to correctly and repeatedly tackle edge cases. Right. And building a reliable system in doing that. So I'll preface it by saying this. Now we've been very deliberate with selecting which features we build first. So exterior washing, sensor detailing is not something we're going to do with the first version of the product. We're also going to be very deliberate about which types of events we will handle. So the product needs a very robust sort of inspection layer that kind of decides whether we can actually service a vehicle. in on the edge in the heart of the city versus it needing to go back into the depot. So what you can expect from V1 is obviously autonomously plugging the vehicle into charge. We do the data sync and the reboot of the vehicle in our station. we inspect and clean the interior of the cabin and that includes removing items left behind. And in the case of Lost and Found, we're simply going to have a runner, a driver that takes a lost wallet or forgotten wallet back to the depot and we can effectively release the vehicle back into operation right in the center of the city. So this is this is V one.
Daniel Abreu Marques: Super interesting. And I think interior cleaning is also one that I keep coming back to. So for example, a spilled coffee is one thing. Somebody maybe being sick in a backseat is maybe a completely other thing. So what share of this interior resets can a robot actually close out and what still needs a human?
George Kalligeros: So firstly, from our analysis, the the vast majority of the cleaning operations are generally relatively routine, right? So the vehicle collects dust, basically the soiling is predictable. It's only a few occurrences where you'll have a biohazard or somebody has left something behind. So I think we can comfortably say Ninety percent of it is just routine cleaning. so this is what we're tackling with our system. we're very deliberately not trying to build such that we can cater to every single exception directly in the outskirts of the city. So vehicle comes in, if it's a biohazard, as as you say, somebody somebody's been sick in the back of a vehicle, it's gonna be extremely difficult for robotics to ever tackle this unless the robotaxis built specifically for really ease of cleaning. So these cases go back to the depot.
Daniel Abreu Marques: and power is also a constraint that a lot of people are talking about so you are dropping then a charging capable unit into the parking space in a dense city. So how do you get the energy in? And is it then also via partnership or how do you do that?
George Kalligeros: Yeah. So so charging is not included with the core products. Effectively, what our station does is it automates the pit stop operations, right? Robotically. And then for the charging side, there's two models. So we either partner with charge point operators. So these are owners and operators of networks of DC fast charging infrastructure across the center of the city. EVGO, for example, one of the largest in the US, 2.2% of their capacity is already tied into AV usage, just to give you example. The other option is to deploy in a satellite constellation and bring off-grid power in, again via a partnership. So it would be a renewable propane generator, or it would be even hydrogen, a battery energy storage system. There is a whole host of options.
Daniel Abreu Marques: Makes sense. And now also every AV platform is is more or less different. So you have the Jaguar I-Pace from Waymo, you have a Zeekr vehicle from Waymo, then you have purpose-built Zoox or cybercabs. So maybe there are also different charge ports, different sensor placements, different interiors. So how do you build then one pod that services all of them? And does that force also maybe a bit of a standardization conversation with the OEMs?
George Kalligeros: Hmm, yeah. I'll answer the second part of your question first. I I think there is a general trend, not of standardization, but of defining products that are fit for robotaxi use and fit for sort of sliding doors, easy to clean interiors, right? we're gonna see a trend. Now, anybody that requires a purpose-by built vehicle to operate is not gonna work, right? Only someone like Tesla is able to really define the vehicle in such narrow depth. the other part of our question is so how do we make a single product that does multiple vehicles at a time? So if you look at the subset of AVs in operation today and the future platforms coming online, it's a relatively small subset. So we have a three-tiered approach. the core structure is designed to accommodate from the largest vehicle, the Volkswagen ID buzz, down to the smallest, the Tesla CyberCab. And then we have a universal robotic system that's effectively able to interface. With all of these vehicles within that 3D space, right? We can cover the entire point cloud. The last 10%, which is what we do really well, is we build bespoke tools, bespoke end of ROM tooling end effectors that allow our robotic system, which is the same across all vehicles, to do the specific vehicle really well. So the vast majority of the of the CapEx and the hardware cost sits in the pods in the robotic system. And then we make bespoke tooling that gives you the ability to do a very specific vehicle as good as a human can.
Daniel Abreu Marques: Super interesting. And another question that came up in my mind is like you have one pod one parking space, and maybe there are three cars in that zone that all need a reset at the same time. So how does that get orchestrated and who decides also maybe the queue?
George Kalligeros: Yeah, it's a good question. I can give you an answer for what it looks like in the future, but initially two things. First of all, we're going to deploy over capacity. so we're going to make sure the the network's going to be not as highly utilized, so pods need to be available on demand. But on the orchestration front, every constellation of pods will be kind of a micro depot, and then we'll simply rely on the AV operator's orchestration layer or even directly the AV developer. to dispatch and send vehicles our way. We're not able to queue in the locations where we operate. So it needs to be a just in time. reserve the pod vehicle shows up. So a very basic API that exposes a pod and allows them to reserve and then everything else is on a on a data exchange. level
Daniel Abreu Marques: And what happens when the pod itself fails, maybe with a with a car inside? So a driverless vehicle stuck in a robot on a public adjacent land, so how do you handle that?
George Kalligeros: Yeah. the initial deployments are gonna be manned. So we're gonna have somebody to respond on the spot immediately. mid-long term, we're gonna have our own ground ops, right? So we're gonna have physical operations on the ground that will attend to these instances, and of course teleoperations and customer service. Now, first of all, if an odd occurrence happens, the robotic system freezes, the vehicle is safely stored passively in a box, right? Nobody can access the pod. if we simply are not able to complete an autonomous cycle of a pit stop, which we're accounting for, we think at least ten percent of occurrences with the early version of the product in pilot will do this, then we can simply charge the vehicle and release it to go back to the depot if it needs a clean in this instance. Otherwise, I mean we show up and we either reset the system or we then allow the vehicle to be released and then it goes back into the depot. Is it solved?
Daniel Abreu Marques: Right. And you also made a deliberate choice not to sell this as a as a product, so to speak. And I would would like to understand why. So you operate the pod as a network rather than selling it as a hardware to fleet. So why take on that capital burden instead of shipping the boxes, for example?
George Kalligeros: Yeah. okay, two elements of the answer. And first of all, our product is a form of infrastructure and it's worth understanding how AV operators currently sort of like buy or access infrastructure. they don't necessarily hold all that infrastructure on their balance sheet, right? they'll spec out a depot location, they need to grow a market or open a market. they'll work with with a real estate developer, really. Pay a deposit, and then they'll commit to a long lease of that infrastructure. So they're used to accessing the infrastructure as a service, is one point. Your second point of why don't we actually ship and sell hardware is how I understand the question. It's not that we don't do this. it's worth looking worth taking this picking this apart because a contract manufacturer makes a pod. And and they sell the pod on for a margin, right? And then somebody acquires and deploys it. the company that acquires this is not Aseon with our venture capital equity funded balance sheet. it will be a project company. And we will be an equity participant in the project company. And then the project company will raise its own asset back financing and get its own cash flows, right? So The SPV will acquire the asset and deploy it. The interesting conversation becomes who should be the equity participant in that vehicle because it shouldn't solely be Aseon. Because we're building this very much like a partnership, we want the charge point operator to participate. We in fact want the AV developer or the AV operator to be very specific. We would like them to participate. And because we're building a very much a new type of infrastructure, which is sort of infrastructure as a product, it's kind of a unique opportunity to kind of pivot the model on its head and allow the AV operators to own part of the infrastructure stack. Long wided answer, but I think it answers the who sells and who makes what.
Daniel Abreu Marques: Yeah, super interesting. thanks for taking this apart. And when we think about like the concrete offerings, so how do you then charge the fleet operators then the per reset? Is it per subscription or guaranteed capacity commitment or something like that?
George Kalligeros: so the model is flexible as long as you can effectively keep the project bankable, right? So if you charge on a per use basis, then you need a minimum utilization guarantee. Not necessarily great for the AV operator, especially for a new system. the most straightforward way is you probably decouple power. You pay for power separately, probably directly with the CPO. And then you pay for the ground and the ops on a monthly basis. Okay? I mean these are these are well controllable OPEX line items. So flexible model as long as the agreement is bankable and as as long as there is equity buy in from everybody involved, is my answer.
Daniel Abreu Marques: And if we look at maybe at at one pod, so how many resets a day or in a specific time frame does it need to work economically? So so what is the payback period, so to speak, for for such a unit?
George Kalligeros: Yeah. let's look at it on a on a utilization basis, not payback. So the big difference here is because you do not employ human staff, which is kind of a high fixed cost with overhead. your marginal cost of delivering a pit stop is very low, right? It's fixed. So we've modeled that we can effectively run a very low utilization network. and still break-even. So we're talking maybe even down to five resets a day So something like 10% utilization suffices for us to operate the network. And the idea here is to move away from this super high throughput, well-run depot that needs to pull the vehicles out of service and causes loss of utilization in the fleet. What the vehicles need is a very dense, very widely on demand available network directly in the pockets of demand, such that You cannot just offer pit stops in the center of the city on demand. You can do micro resets. Okay? Because this is where the real economic unlock is. There is natural downtime for 10, 15, 20 minutes during the day of the vehicle. If you can squeeze in partial charges, partial clean cleanups, then you can keep the fleet running constantly. And that might not be a problem today, but this is crucial when the city will tender out licenses with a cap number of vehicles. And if your thesis like Tesla Cybercab thesis is it's so cheap, it runs on its own, it's free to run, so you can drive to another state to get charged, that that doesn't apply because the externality of increasing traffic is not accounted for, right? The infrastructure has to meet the vehicle, not the vehicle taken to the infrastructure just because it drives itself.
Daniel Abreu Marques: Make sense and maybe one more question related to the cash. So I mentioned in the intro that you raised 10 million. So so what is the plan for for this 10 million that you raised
George Kalligeros: Yeah, so all of the money is going to engineering. proving that the product works and proving that we can get to a reliable autonomous pit stop, at least targeting the very core of the reset. So it's all about building out the team, building out the product and supporting our first pilot deployments with the goal of being manufacturing ready with a third iteration of our product by the end of 2027.
Daniel Abreu Marques: And if we then think about a city scale network, so this was probably a very very different number. So I don't know if you can gauge a number, but what does the capital stack look beyond this round? So at what point does it stop being then like yeah the venture in the start of being like a real infrastructure debt or something like that?
George Kalligeros: Yeah. It's it's what we touched on before. So we'll soon graduate from sort of like the product development phase and we're gonna be rolling out infrastructure across multiple metros, right? Deployments of fifty million per metro, you can be doing a dozen cities at the time. And each each metro will be its own SPV that will draw asset back finance to to finance and roll out the hardware. So The financing will of course look different to the typical startup, but you know, there is a holding company, an operating company where the IP, the team, the commercial agreements on a global level sit. And then there's gonna be regional sort of asset companies that are gonna be JVs with the regional partners that are gonna make this happen. They're gonna bring the ground, they're gonna bring the power, they're gonna they're gonna consume the product and bring the operations. so it's a two tiered approach.
Daniel Abreu Marques: Okay. Super interesting. And what I understood is also that the pods are a temporary structure, so deployable in a day and also movable if if it maybe, I don't know, a site underperforms or something like that. so how much of the business is then except the robotics, maybe also real estate, permitting and stuff like that?
George Kalligeros: Yeah, a very significant component here is to not just be a robotics product company. I'll say this. Currently the hunt for permittable powered ground where you can erect an industrial depot is scarce. You kind of if you end up with a land that you can deploy and you deploy. the beauty of our product, it's movable. Because just with a conditional use authorization and a state level certification, you can pretty much deploy across significantly more parcels of land, right? So the reason it's movable and temporary has to do with the fact that we're designing it to be extremely quick to deploy. So just within within eight weeks of looking into a specific city, we can apply and we can pretty much have operations alive. So this is this is why the product itself is designed in that.
Daniel Abreu Marques: And I think also there's currently a lot of focus on this layer, to be honest. So quite a lot of interest. and I want to place you inside it. So you have, for example, players like Lyft, they're operating an 80,000 square foot facility in Nashville for Waymo. Then you have Avis that is doing Dallas, Move that does Phoenix. so why would any of those operators route vehicles then into your network instead of building the capability into depots they are already paying for, for example.
George Kalligeros: Yeah. So first of all, we're talking about additional capacity, right? So the cities that you mentioned, the metros, they're running successful pilot scale. So now it's it's all about scaling up. Now if you run five hundred vehicles, you wanna have two thousand vehicles, you need four times also the infrastructural footprint and capacity to to turn the fleet around. And the current answer is more maintenance hub and also more sort of like the satellite locations and everything else. what they would what the operators would effectively buy from us is not just additional powered ground via the partnerships we're gonna bring in, but also the ability to deploy the decentral locations with autonomous operations, which completely flips the economics, right? Because you cannot profitably run a two or three bay depot in the center of the city stuffed with humans.
Daniel Abreu Marques: one more question related to also your partnerships because you mentioned, for example, charging, you have partners and stuff like that. but the robotics inside this box, this is wholly developed from Aseon, or do you also have there some partners which provide the robotics as well?
George Kalligeros: No, this is our own IP. So this is the core of what we build and this is the core of what we retain over the long term. we're being very strategic with what makes sense to be developed in-house because frankly, the rule of hardware is it takes multiple iterations to build a stable product. And you launch a product, it breaks, you fix what's broken, you launch it, it works, but the economics don't work. Getting to a stable hardware business is tough. So What we're optimizing for here is the speed with which we can deploy the stable solution. Right? So yeah, this part of the stack remains very much in house. And this is what we've hired for.
Daniel Abreu Marques: maybe now let's zoom out a bit. talk a bit more about the layer you're also betting on. So I think a lot of people always right now draw this comparison that AV depots are like data centers or cell towers. Do you think this is a right analogy? And also maybe what does the mature version of this market look like structurally and who owns it then?
George Kalligeros: Yeah. So first of all on the analogy, yeah, it is a good analogy. Robotaxis is the biggest robotic application, physical AI, in the world. It's gonna remain that, together with humanoids. So it's a new technology that needs its own new type of bespoke infrastructure. So in this regard, for sure, we need a build out and a significant portion of the cost of running the services, more than half, ties to that infrastructural and operations footprint. Now in terms of what happens and how it how it forms and who owns the layer, you know kind of infrastructure is a little bit fuzzy. it's very much on a regional basis. You need to look at who effectively at the end of the day owns that infrastructure. I kind of pine, but it's a TBD right? I don't think it necessarily matters as much right now. It will remain, I think, very much its own layer and then the AV operators on top. You're gonna see some companies like what we do take a more product first approach to infrastructure. And I think this very much changes what's possible in terms of owning more of that vertical stack when you're deploying a product like ours.
Daniel Abreu Marques: And one one claim that you also made is that operations run around seventy percent of the cost of of the service. So that if that is right, then autonomy economics are decided off the road rather than on it. what is your experience with the AV developers? Are they recognizing that or is their attention still on developing the the best driver, so to speak?
George Kalligeros: Yeah, well we that was a borrowed number. we've done our own analysis. We believe it's around fifty-five percent. the contribution of infrastructure and human ops, thirty percent of it is the infrastructure percentage points. And I think it's a major lever, obviously, but obviously there's still fat optimization and margin a across the entire stack. and work needs to happen. across the entire of the stack to get to profitable robotaxi operations across every city, right? So Uber operates right now in fifteen thousand cities. We have live robotaxi operations in fifteen, like Waymos 14. We have Zoox Vegas, and then you have another nine China and two Middle East, right? Like 25, 26. So there is a tail end of markets that are let's say the tier two markets We'll need to innovate on the operating model, I'd say. Yeah.
Daniel Abreu Marques: another point is that cities also probably have a say in this so you're placing your robots in the parking space and I think the curb space in general is also politically contested in a way so where has that been harder than expected and also what does a city get out of saying yes to this kind of micro depot
George Kalligeros: Good question. So, full disclosure, we haven't reached that point. So, we're not operating a network in a city at the scale where there's friction with the city. and we're also deploying our product in private land, right? So the interface with the city is, of course, on the permitting front and certification with the state. so the argument that I think might become a public argument is If we're starting to now deploy in partnership with the CPO on top of charging bays that consumers like you and I that have an electric vehicle would previously use, and we now privatize that bay and give it to autonomous vehicles, and you've taken something back from the city dweller. But the counter-argument to this is people that live in the city are the people that use the robotaxis, right? And then the person that uses the robotaxi has kind of like a right. To a slice of that piece of infrastructure. So maybe they chose not to drive an electric vehicle and chose to take a robotaxi. So their robotaxi needs to use that public infrastructure as such. So I think yes, depots are more hidden and in industrial zones and kind of like far away and to do the dirty work, but there is a real negative externality of faraway depots, which is traffic. And I think traffic is gonna be a big problem. even though if we were to freeze the level of movement of people and say that needs to be static, et cetera basically remain the same. autonomous vehicles reduce traffic, but every time we gave people a better and a cheaper way to move around, they just move more. So traffic will inevitably increase and
Daniel Abreu Marques: for the last stretch, let's look forward. and I want to look at another part of the world Europe. So you are Greek. I think your co-founder is British when I remember correct. And you also built already in Europe one time and now you're building in the US. Europe in general has less robotaxi volume, has more regulatory processes. So is Aseon some how show up here in Europe and what would have to happen first that you can also show up?
George Kalligeros: Yeah, we a hundred percent will. So yes we're European, yes we'll build build here before. we're building Aseon out of the US because I mean, frankly, I think we as Europeans we take a great idea and we make it plausible. So Aseon needs to be built out of the US in the valley, this is where the robotics talent exists, and and this is where you build a global business out of. So that's my general view, unfortunately, on Europe. Now, in terms of deployments, Europe is extremely interesting because the architecture of our cities is such that it's much more dense, right? so there is of the big 40 global metros where Robotaxi are going to operating soon, you need that inner city penetration. And our model applies immediately, right? So a lot of it is Europe and then On on the additional tail end, these are cities where you'll need to be much more operationally prudent and come up with new models to operate profitably in. Then again, our model directly applies. So think Liverpool versus London kind of thing, right? So from this 200 city list, a lot of it actually lies in Europe. we'll see how the Munich launch goes, how London goes. Obviously, like across Europe we have more deployments. I think it's coming. I think it's coming. as quickly as it's developing across the rest of the West and Europe is still a major portion of the global robotaxi rollout. So I actually can't wait to deploy a European city.
Daniel Abreu Marques: Yeah, absolutely. when I heard about your product, I also immediately thought this is perfect for the dense European cities. so really good that you're you're also looking into Europe. And maybe one last question to close here. if we look twelve months from now, so so what is the one metric you judge yourselves on? So the number that tells you that this was a successful year.
George Kalligeros: This year is all about proving out that we can autonomously reset. So the metric of success is what we call uptime. So what is the ratio of successfully completed autonomous reset cycles over attempted? anything over 90% allows us to deploy and operate profitably in the first target markets. So this is it. Well once we get to that level of product maturity we have a business and then we're aggressive to yeah.
Daniel Abreu Marques: Right. George, thank you so much. This was really interesting to hear more about Aseon and also your pods and your strategy and way forward. I'm looking forward to talk maybe in some months and talk about your progress and what you achieved so far. it was really a pleasure to to meet you today and have you on the show. Thank you so much, George.
George Kalligeros: Thank you, Daniel. I'd love to be back. Thank you.
George Kalligeros: we build robotic pit stops for self-driving cars. Effectively this is distributed depot footprint directly within the heart of the operating zone. What the vehicles need is a very dense, very widely on demand available network directly in the pockets of demand, such that You cannot just offer pit stops in the center of the city on demand. You can do micro resets. Okay? Because this is where the real economic unlock is.
Daniel Abreu Marques: Welcome to Autonomy Insiders. I'm your host, Daniel, and most conversations about autonomy are about the driver, the software, the sensors, the miles. This episode is about the other part cleaning, inspection, getting a car ready to earn money again. And my guest is building a robot that sits in a single parking space and does all of that, so the car never has to leave the neighborhood it is earning in. George Kalligeros is co-founder and CEO of Aseon Labs and Aseon came out of stealth in May and announced a ten million dollar seed in June. So George, welcome on the show and I'm so happy to have you on today.
George Kalligeros: Thanks, Daniel. Thanks so much for inviting me.
Daniel Abreu Marques: So before we go anywhere, maybe in plain terms, so what is Aseon and what does a fleet operator actually get from you?
George Kalligeros: Yeah. So at Aseon we build robotic pit stops for self-driving cars. Effectively this is distributed depot footprint directly within the heart of the operating zone. So an AV operator gets from us powered ground and fully robotic operations as a service.
Daniel Abreu Marques: And can you also describe this this pod as a physical object? So maybe if I would stand next to it in a parking lot, so how big is it? What do I see? Maybe also what moves?
George Kalligeros: Yeah, for sure. So as you described, it very much looks like a box and it occupies one typical charging bay, right? So the product in European terms is about six meters long and two point seven meters wide. And you don't really see much from the outside. I'll I'll share with you a photo that you can pull up. This is the the latest version of the product that we're building currently in Manlo Park, California. It's gonna be right in October. you can peer from the inside and you'll see a robotic system. which uses two robotic arms with six degrees of freedom that pick from a range of custom tooling to do a bunch of interesting things on the autonomous vehicle.
Daniel Abreu Marques: And then from the car's point of view, so it pulls in and then and what so roughly how long is it in there before it's then back taking rides? yeah, maybe high level for now we can get into the mechanics later, but just to get a get a picture.
George Kalligeros: Yeah. So the concept of the product is to compress as much of the depot workflows in a single box. That's why it's called a micro depot. So the car pulls in, the first thing that happens is after an inspection, the vehicle gets plugged into charge and then a sequence of events takes place that takes anywhere between twenty to thirty minutes. the main thing that drives the time inside the pod is the charging rate and how long we keep the vehicle into charge. every single other operation that we do effectively falls within that window. So yeah, this is this is the core.
Daniel Abreu Marques: So now we have a good picture. And now I want to understand what that replaces. maybe walk me through what happens today without you and your product. So a robotaxi, for example, in San Francisco needs a reset at two in the afternoon. So what does that actually cost the operator? Maybe also in time and probably also miles.
George Kalligeros: Yeah, for sure. I mean typically the primary driver that takes vehicles out of service is that the vehicle operates and it it falls below a threshold, typically state of charge, right? So the vehicle runs low on battery, maybe it's twenty, twenty-five percent battery, and then it needs to call into either a maintenance hub, a depot, or a pit stop location, which are sort of like smaller sort of rapidly deployed locations near the operating zone. San Francisco actually has good infrastructure to support Robotaxi Ops and it's a relatively compressed city. So maybe you're depending on where you're in the city, maybe within three to five miles, the car can drive to a depot and then gets charged, right? And serviced. The service typically takes about I'd say anywhere between 40 minutes to 80 minutes. Let's say the hour is the average, and then it might need to travel another 40 minutes back and forth. So typically the vehicle is out of service. let's say two hours with every sort of recharge operation, but that happens multiple times in a day.
Daniel Abreu Marques: Makes sense. And you and also I think your co-founder, you visited also AV depots while researching all of this. So what did you see there that you did not expect?
George Kalligeros: Yeah, we we did. I didn't I don't think we saw anything that we did not expect. they ran efficiently, they were well equipped, they're relatively new, right? I think what became strikingly obvious is it how difficult it is to run human operations and how difficult it is to incentivize the human operators to to effectively carry out the repeat high frequency tasks in the depot. And Yeah, this effectively kicked off a lot of the inspiration for why we are now building with robotics.
Daniel Abreu Marques: them I think you said also that roughly a third of a fleet is offline at any given time. So is that also a number operators recognize and accept or is it also more a number that surprises them maybe when you put it in front of them
George Kalligeros: Yeah, so I mean these figures were pulled from a from a public report that I think it was Harry Campbell basically doing some analysis on the Waymo fleet a year and a half ago. But the operators have a really, really good grasp of their core metrics. And that number came from what what's the percentage of fleet that could be servicing customers but are currently either traveling to the depot or being serviced or they're queuing or they're in inspection. And cannot deliver a ride at any point in time. we crunch a lot of these numbers to kind of understand how like what is the effect of the locality of the depot. If it's far from the city, what is the effect in terms of the utilization of the fleet? Because there is this antithesis where today's model of a depot which is very human-centric, it's all about high throughput and turning the fleet around and trying to compress as much of the servicing in the late hours of the day, whereas the fleet Pretty much just needs to be constantly operating, utilizing, and capturing rides
Daniel Abreu Marques: And before before Aseon you also spent four years at Tier also deploying hardware into cities at scale. So what did that period teach you about operational costs that you could not have learned from the outside, so to speak?
George Kalligeros: Operation cost. Tier was a very interesting case study in general. It was a company that transitioned from a growth phase to an operational, sort of like an economic sanity and a profitability phase really quickly. In terms of operating costs, I think the biggest learning was Tier and all the companies at the time had very much their own in-house operations, for everything from maintenance, rebalancing, charging. And I think Lime innovated on this model where they started outsourcing operations to local entrepreneurs, gig economy workers that were much more highly and better incentivized to to turn the fleet around. Right. So I think the learning there was doing a really sort of like deep and vertical ops structure it's hard to incentivize the people that are really working with the assets. Yeah. I think that that was really the key learning. In terms of cost.
Daniel Abreu Marques: Interesting and maybe now let's let's open the box. So I want to know which parts of this are resolved and which was maybe also are still hard right now. So there's charging, there's exterior wash, interior clean, lost items, sensors inspection, data sync. So that is a lot of different jobs for one machine, so to speak. which of those is working today and which is maybe also still an open
George Kalligeros: Yeah, yeah, it's a good question. I'll preface this by saying we should define what a solved problem is because just in robotics, getting to a working demo, getting to a ninety nine percent maturity of a feature is let's say ten percent of the work, right? so we can give you a working demo, the work really is in the last one percent, which is being able to correctly and repeatedly tackle edge cases. Right. And building a reliable system in doing that. So I'll preface it by saying this. Now we've been very deliberate with selecting which features we build first. So exterior washing, sensor detailing is not something we're going to do with the first version of the product. We're also going to be very deliberate about which types of events we will handle. So the product needs a very robust sort of inspection layer that kind of decides whether we can actually service a vehicle. in on the edge in the heart of the city versus it needing to go back into the depot. So what you can expect from V1 is obviously autonomously plugging the vehicle into charge. We do the data sync and the reboot of the vehicle in our station. we inspect and clean the interior of the cabin and that includes removing items left behind. And in the case of Lost and Found, we're simply going to have a runner, a driver that takes a lost wallet or forgotten wallet back to the depot and we can effectively release the vehicle back into operation right in the center of the city. So this is this is V one.
Daniel Abreu Marques: Super interesting. And I think interior cleaning is also one that I keep coming back to. So for example, a spilled coffee is one thing. Somebody maybe being sick in a backseat is maybe a completely other thing. So what share of this interior resets can a robot actually close out and what still needs a human?
George Kalligeros: So firstly, from our analysis, the the vast majority of the cleaning operations are generally relatively routine, right? So the vehicle collects dust, basically the soiling is predictable. It's only a few occurrences where you'll have a biohazard or somebody has left something behind. So I think we can comfortably say Ninety percent of it is just routine cleaning. so this is what we're tackling with our system. we're very deliberately not trying to build such that we can cater to every single exception directly in the outskirts of the city. So vehicle comes in, if it's a biohazard, as as you say, somebody somebody's been sick in the back of a vehicle, it's gonna be extremely difficult for robotics to ever tackle this unless the robotaxis built specifically for really ease of cleaning. So these cases go back to the depot.
Daniel Abreu Marques: and power is also a constraint that a lot of people are talking about so you are dropping then a charging capable unit into the parking space in a dense city. So how do you get the energy in? And is it then also via partnership or how do you do that?
George Kalligeros: Yeah. So so charging is not included with the core products. Effectively, what our station does is it automates the pit stop operations, right? Robotically. And then for the charging side, there's two models. So we either partner with charge point operators. So these are owners and operators of networks of DC fast charging infrastructure across the center of the city. EVGO, for example, one of the largest in the US, 2.2% of their capacity is already tied into AV usage, just to give you example. The other option is to deploy in a satellite constellation and bring off-grid power in, again via a partnership. So it would be a renewable propane generator, or it would be even hydrogen, a battery energy storage system. There is a whole host of options.
Daniel Abreu Marques: Makes sense. And now also every AV platform is is more or less different. So you have the Jaguar I-Pace from Waymo, you have a Zeekr vehicle from Waymo, then you have purpose-built Zoox or cybercabs. So maybe there are also different charge ports, different sensor placements, different interiors. So how do you build then one pod that services all of them? And does that force also maybe a bit of a standardization conversation with the OEMs?
George Kalligeros: Hmm, yeah. I'll answer the second part of your question first. I I think there is a general trend, not of standardization, but of defining products that are fit for robotaxi use and fit for sort of sliding doors, easy to clean interiors, right? we're gonna see a trend. Now, anybody that requires a purpose-by built vehicle to operate is not gonna work, right? Only someone like Tesla is able to really define the vehicle in such narrow depth. the other part of our question is so how do we make a single product that does multiple vehicles at a time? So if you look at the subset of AVs in operation today and the future platforms coming online, it's a relatively small subset. So we have a three-tiered approach. the core structure is designed to accommodate from the largest vehicle, the Volkswagen ID buzz, down to the smallest, the Tesla CyberCab. And then we have a universal robotic system that's effectively able to interface. With all of these vehicles within that 3D space, right? We can cover the entire point cloud. The last 10%, which is what we do really well, is we build bespoke tools, bespoke end of ROM tooling end effectors that allow our robotic system, which is the same across all vehicles, to do the specific vehicle really well. So the vast majority of the of the CapEx and the hardware cost sits in the pods in the robotic system. And then we make bespoke tooling that gives you the ability to do a very specific vehicle as good as a human can.
Daniel Abreu Marques: Super interesting. And another question that came up in my mind is like you have one pod one parking space, and maybe there are three cars in that zone that all need a reset at the same time. So how does that get orchestrated and who decides also maybe the queue?
George Kalligeros: Yeah, it's a good question. I can give you an answer for what it looks like in the future, but initially two things. First of all, we're going to deploy over capacity. so we're going to make sure the the network's going to be not as highly utilized, so pods need to be available on demand. But on the orchestration front, every constellation of pods will be kind of a micro depot, and then we'll simply rely on the AV operator's orchestration layer or even directly the AV developer. to dispatch and send vehicles our way. We're not able to queue in the locations where we operate. So it needs to be a just in time. reserve the pod vehicle shows up. So a very basic API that exposes a pod and allows them to reserve and then everything else is on a on a data exchange. level
Daniel Abreu Marques: And what happens when the pod itself fails, maybe with a with a car inside? So a driverless vehicle stuck in a robot on a public adjacent land, so how do you handle that?
George Kalligeros: Yeah. the initial deployments are gonna be manned. So we're gonna have somebody to respond on the spot immediately. mid-long term, we're gonna have our own ground ops, right? So we're gonna have physical operations on the ground that will attend to these instances, and of course teleoperations and customer service. Now, first of all, if an odd occurrence happens, the robotic system freezes, the vehicle is safely stored passively in a box, right? Nobody can access the pod. if we simply are not able to complete an autonomous cycle of a pit stop, which we're accounting for, we think at least ten percent of occurrences with the early version of the product in pilot will do this, then we can simply charge the vehicle and release it to go back to the depot if it needs a clean in this instance. Otherwise, I mean we show up and we either reset the system or we then allow the vehicle to be released and then it goes back into the depot. Is it solved?
Daniel Abreu Marques: Right. And you also made a deliberate choice not to sell this as a as a product, so to speak. And I would would like to understand why. So you operate the pod as a network rather than selling it as a hardware to fleet. So why take on that capital burden instead of shipping the boxes, for example?
George Kalligeros: Yeah. okay, two elements of the answer. And first of all, our product is a form of infrastructure and it's worth understanding how AV operators currently sort of like buy or access infrastructure. they don't necessarily hold all that infrastructure on their balance sheet, right? they'll spec out a depot location, they need to grow a market or open a market. they'll work with with a real estate developer, really. Pay a deposit, and then they'll commit to a long lease of that infrastructure. So they're used to accessing the infrastructure as a service, is one point. Your second point of why don't we actually ship and sell hardware is how I understand the question. It's not that we don't do this. it's worth looking worth taking this picking this apart because a contract manufacturer makes a pod. And and they sell the pod on for a margin, right? And then somebody acquires and deploys it. the company that acquires this is not Aseon with our venture capital equity funded balance sheet. it will be a project company. And we will be an equity participant in the project company. And then the project company will raise its own asset back financing and get its own cash flows, right? So The SPV will acquire the asset and deploy it. The interesting conversation becomes who should be the equity participant in that vehicle because it shouldn't solely be Aseon. Because we're building this very much like a partnership, we want the charge point operator to participate. We in fact want the AV developer or the AV operator to be very specific. We would like them to participate. And because we're building a very much a new type of infrastructure, which is sort of infrastructure as a product, it's kind of a unique opportunity to kind of pivot the model on its head and allow the AV operators to own part of the infrastructure stack. Long wided answer, but I think it answers the who sells and who makes what.
Daniel Abreu Marques: Yeah, super interesting. thanks for taking this apart. And when we think about like the concrete offerings, so how do you then charge the fleet operators then the per reset? Is it per subscription or guaranteed capacity commitment or something like that?
George Kalligeros: so the model is flexible as long as you can effectively keep the project bankable, right? So if you charge on a per use basis, then you need a minimum utilization guarantee. Not necessarily great for the AV operator, especially for a new system. the most straightforward way is you probably decouple power. You pay for power separately, probably directly with the CPO. And then you pay for the ground and the ops on a monthly basis. Okay? I mean these are these are well controllable OPEX line items. So flexible model as long as the agreement is bankable and as as long as there is equity buy in from everybody involved, is my answer.
Daniel Abreu Marques: And if we look at maybe at at one pod, so how many resets a day or in a specific time frame does it need to work economically? So so what is the payback period, so to speak, for for such a unit?
George Kalligeros: Yeah. let's look at it on a on a utilization basis, not payback. So the big difference here is because you do not employ human staff, which is kind of a high fixed cost with overhead. your marginal cost of delivering a pit stop is very low, right? It's fixed. So we've modeled that we can effectively run a very low utilization network. and still break-even. So we're talking maybe even down to five resets a day So something like 10% utilization suffices for us to operate the network. And the idea here is to move away from this super high throughput, well-run depot that needs to pull the vehicles out of service and causes loss of utilization in the fleet. What the vehicles need is a very dense, very widely on demand available network directly in the pockets of demand, such that You cannot just offer pit stops in the center of the city on demand. You can do micro resets. Okay? Because this is where the real economic unlock is. There is natural downtime for 10, 15, 20 minutes during the day of the vehicle. If you can squeeze in partial charges, partial clean cleanups, then you can keep the fleet running constantly. And that might not be a problem today, but this is crucial when the city will tender out licenses with a cap number of vehicles. And if your thesis like Tesla Cybercab thesis is it's so cheap, it runs on its own, it's free to run, so you can drive to another state to get charged, that that doesn't apply because the externality of increasing traffic is not accounted for, right? The infrastructure has to meet the vehicle, not the vehicle taken to the infrastructure just because it drives itself.
Daniel Abreu Marques: Make sense and maybe one more question related to the cash. So I mentioned in the intro that you raised 10 million. So so what is the plan for for this 10 million that you raised
George Kalligeros: Yeah, so all of the money is going to engineering. proving that the product works and proving that we can get to a reliable autonomous pit stop, at least targeting the very core of the reset. So it's all about building out the team, building out the product and supporting our first pilot deployments with the goal of being manufacturing ready with a third iteration of our product by the end of 2027.
Daniel Abreu Marques: And if we then think about a city scale network, so this was probably a very very different number. So I don't know if you can gauge a number, but what does the capital stack look beyond this round? So at what point does it stop being then like yeah the venture in the start of being like a real infrastructure debt or something like that?
George Kalligeros: Yeah. It's it's what we touched on before. So we'll soon graduate from sort of like the product development phase and we're gonna be rolling out infrastructure across multiple metros, right? Deployments of fifty million per metro, you can be doing a dozen cities at the time. And each each metro will be its own SPV that will draw asset back finance to to finance and roll out the hardware. So The financing will of course look different to the typical startup, but you know, there is a holding company, an operating company where the IP, the team, the commercial agreements on a global level sit. And then there's gonna be regional sort of asset companies that are gonna be JVs with the regional partners that are gonna make this happen. They're gonna bring the ground, they're gonna bring the power, they're gonna they're gonna consume the product and bring the operations. so it's a two tiered approach.
Daniel Abreu Marques: Okay. Super interesting. And what I understood is also that the pods are a temporary structure, so deployable in a day and also movable if if it maybe, I don't know, a site underperforms or something like that. so how much of the business is then except the robotics, maybe also real estate, permitting and stuff like that?
George Kalligeros: Yeah, a very significant component here is to not just be a robotics product company. I'll say this. Currently the hunt for permittable powered ground where you can erect an industrial depot is scarce. You kind of if you end up with a land that you can deploy and you deploy. the beauty of our product, it's movable. Because just with a conditional use authorization and a state level certification, you can pretty much deploy across significantly more parcels of land, right? So the reason it's movable and temporary has to do with the fact that we're designing it to be extremely quick to deploy. So just within within eight weeks of looking into a specific city, we can apply and we can pretty much have operations alive. So this is this is why the product itself is designed in that.
Daniel Abreu Marques: And I think also there's currently a lot of focus on this layer, to be honest. So quite a lot of interest. and I want to place you inside it. So you have, for example, players like Lyft, they're operating an 80,000 square foot facility in Nashville for Waymo. Then you have Avis that is doing Dallas, Move that does Phoenix. so why would any of those operators route vehicles then into your network instead of building the capability into depots they are already paying for, for example.
George Kalligeros: Yeah. So first of all, we're talking about additional capacity, right? So the cities that you mentioned, the metros, they're running successful pilot scale. So now it's it's all about scaling up. Now if you run five hundred vehicles, you wanna have two thousand vehicles, you need four times also the infrastructural footprint and capacity to to turn the fleet around. And the current answer is more maintenance hub and also more sort of like the satellite locations and everything else. what they would what the operators would effectively buy from us is not just additional powered ground via the partnerships we're gonna bring in, but also the ability to deploy the decentral locations with autonomous operations, which completely flips the economics, right? Because you cannot profitably run a two or three bay depot in the center of the city stuffed with humans.
Daniel Abreu Marques: one more question related to also your partnerships because you mentioned, for example, charging, you have partners and stuff like that. but the robotics inside this box, this is wholly developed from Aseon, or do you also have there some partners which provide the robotics as well?
George Kalligeros: No, this is our own IP. So this is the core of what we build and this is the core of what we retain over the long term. we're being very strategic with what makes sense to be developed in-house because frankly, the rule of hardware is it takes multiple iterations to build a stable product. And you launch a product, it breaks, you fix what's broken, you launch it, it works, but the economics don't work. Getting to a stable hardware business is tough. So What we're optimizing for here is the speed with which we can deploy the stable solution. Right? So yeah, this part of the stack remains very much in house. And this is what we've hired for.
Daniel Abreu Marques: maybe now let's zoom out a bit. talk a bit more about the layer you're also betting on. So I think a lot of people always right now draw this comparison that AV depots are like data centers or cell towers. Do you think this is a right analogy? And also maybe what does the mature version of this market look like structurally and who owns it then?
George Kalligeros: Yeah. So first of all on the analogy, yeah, it is a good analogy. Robotaxis is the biggest robotic application, physical AI, in the world. It's gonna remain that, together with humanoids. So it's a new technology that needs its own new type of bespoke infrastructure. So in this regard, for sure, we need a build out and a significant portion of the cost of running the services, more than half, ties to that infrastructural and operations footprint. Now in terms of what happens and how it how it forms and who owns the layer, you know kind of infrastructure is a little bit fuzzy. it's very much on a regional basis. You need to look at who effectively at the end of the day owns that infrastructure. I kind of pine, but it's a TBD right? I don't think it necessarily matters as much right now. It will remain, I think, very much its own layer and then the AV operators on top. You're gonna see some companies like what we do take a more product first approach to infrastructure. And I think this very much changes what's possible in terms of owning more of that vertical stack when you're deploying a product like ours.
Daniel Abreu Marques: And one one claim that you also made is that operations run around seventy percent of the cost of of the service. So that if that is right, then autonomy economics are decided off the road rather than on it. what is your experience with the AV developers? Are they recognizing that or is their attention still on developing the the best driver, so to speak?
George Kalligeros: Yeah, well we that was a borrowed number. we've done our own analysis. We believe it's around fifty-five percent. the contribution of infrastructure and human ops, thirty percent of it is the infrastructure percentage points. And I think it's a major lever, obviously, but obviously there's still fat optimization and margin a across the entire stack. and work needs to happen. across the entire of the stack to get to profitable robotaxi operations across every city, right? So Uber operates right now in fifteen thousand cities. We have live robotaxi operations in fifteen, like Waymos 14. We have Zoox Vegas, and then you have another nine China and two Middle East, right? Like 25, 26. So there is a tail end of markets that are let's say the tier two markets We'll need to innovate on the operating model, I'd say. Yeah.
Daniel Abreu Marques: another point is that cities also probably have a say in this so you're placing your robots in the parking space and I think the curb space in general is also politically contested in a way so where has that been harder than expected and also what does a city get out of saying yes to this kind of micro depot
George Kalligeros: Good question. So, full disclosure, we haven't reached that point. So, we're not operating a network in a city at the scale where there's friction with the city. and we're also deploying our product in private land, right? So the interface with the city is, of course, on the permitting front and certification with the state. so the argument that I think might become a public argument is If we're starting to now deploy in partnership with the CPO on top of charging bays that consumers like you and I that have an electric vehicle would previously use, and we now privatize that bay and give it to autonomous vehicles, and you've taken something back from the city dweller. But the counter-argument to this is people that live in the city are the people that use the robotaxis, right? And then the person that uses the robotaxi has kind of like a right. To a slice of that piece of infrastructure. So maybe they chose not to drive an electric vehicle and chose to take a robotaxi. So their robotaxi needs to use that public infrastructure as such. So I think yes, depots are more hidden and in industrial zones and kind of like far away and to do the dirty work, but there is a real negative externality of faraway depots, which is traffic. And I think traffic is gonna be a big problem. even though if we were to freeze the level of movement of people and say that needs to be static, et cetera basically remain the same. autonomous vehicles reduce traffic, but every time we gave people a better and a cheaper way to move around, they just move more. So traffic will inevitably increase and
Daniel Abreu Marques: for the last stretch, let's look forward. and I want to look at another part of the world Europe. So you are Greek. I think your co-founder is British when I remember correct. And you also built already in Europe one time and now you're building in the US. Europe in general has less robotaxi volume, has more regulatory processes. So is Aseon some how show up here in Europe and what would have to happen first that you can also show up?
George Kalligeros: Yeah, we a hundred percent will. So yes we're European, yes we'll build build here before. we're building Aseon out of the US because I mean, frankly, I think we as Europeans we take a great idea and we make it plausible. So Aseon needs to be built out of the US in the valley, this is where the robotics talent exists, and and this is where you build a global business out of. So that's my general view, unfortunately, on Europe. Now, in terms of deployments, Europe is extremely interesting because the architecture of our cities is such that it's much more dense, right? so there is of the big 40 global metros where Robotaxi are going to operating soon, you need that inner city penetration. And our model applies immediately, right? So a lot of it is Europe and then On on the additional tail end, these are cities where you'll need to be much more operationally prudent and come up with new models to operate profitably in. Then again, our model directly applies. So think Liverpool versus London kind of thing, right? So from this 200 city list, a lot of it actually lies in Europe. we'll see how the Munich launch goes, how London goes. Obviously, like across Europe we have more deployments. I think it's coming. I think it's coming. as quickly as it's developing across the rest of the West and Europe is still a major portion of the global robotaxi rollout. So I actually can't wait to deploy a European city.
Daniel Abreu Marques: Yeah, absolutely. when I heard about your product, I also immediately thought this is perfect for the dense European cities. so really good that you're you're also looking into Europe. And maybe one last question to close here. if we look twelve months from now, so so what is the one metric you judge yourselves on? So the number that tells you that this was a successful year.
George Kalligeros: This year is all about proving out that we can autonomously reset. So the metric of success is what we call uptime. So what is the ratio of successfully completed autonomous reset cycles over attempted? anything over 90% allows us to deploy and operate profitably in the first target markets. So this is it. Well once we get to that level of product maturity we have a business and then we're aggressive to yeah.
Daniel Abreu Marques: Right. George, thank you so much. This was really interesting to hear more about Aseon and also your pods and your strategy and way forward. I'm looking forward to talk maybe in some months and talk about your progress and what you achieved so far. it was really a pleasure to to meet you today and have you on the show. Thank you so much, George.
George Kalligeros: Thank you, Daniel. I'd love to be back. Thank you.
