Even though autonomous passenger vehicles have entered the mainstream in cities across the country, autonomous trucks still lag behind. But Humble Robotics thinks it has cracked the code with a new design that completely does away with the tractor-trailer model we see on the highway every day.
In this episode, Shayle speaks to Eyal Cohen, founder and CEO of Humble. The company built its electric trucks from the ground up. Fully cabless, they combine the tractor and trailer into a single platform designed to optimize energy efficiency, unit economics, and roadway safety.
Shayle and Eyal explore topics including:
- The differences between autonomous passenger and freight vehicles
- The challenge of transporting heavy payloads at high speeds
- Why Humble has shifted away from LiDar in favor of a camera-centric approach offered by visual language models (VLMs)
- The unit economics of electric and autonomous freight
- Why Humble is embracing a “hub-to-hub” model for its trucks
- The evolving regulatory landscape for autonomous trucking
Resources
- Catalyst: Volts crossover: Six big energy questions
- Latitude Media: Can the Tesla Semi finalize decarbonize trucking?
- Latitude Media: Rivian and EnergyHub are teaming up on managed charging
- The Green Blueprint: A billion-dollar play on electrified transport
Credits: Hosted by Shayle Kann. Produced and edited by Max Savage Levenson. Original music and engineering by Sean Marquand. Stephen Lacey is our executive editor.
Catalyst is brought to you by FischTank PR, an award-winning climate and energy tech, renewables, and sustainability-focused PR firm dedicated to elevating the work of both early-stage and established companies. Learn more about their PR approach and how they can support your company’s messaging by visiting fischtankpr.com.
Catalyst is brought to you by EnergyHub. EnergyHub helps utilities build next-generation virtual power plants that unlock reliable flexibility at every level of the grid. See how EnergyHub helps unlock the power of flexibility at scale, and deliver more value through cross-DER dispatch with their leading Edge DERMS platform, by visiting energyhub.com.
Tune into Critical Capital, a brand new podcast from Crux and Latitude Studios. Hosted by Crux CEO Alfred Johnson, Critical Capital explores the interlocking forces powering clean and critical infrastructure. Join us every other Tuesday for in-depth conversations at the intersection of energy, government, finance, and global markets. Listen here, or wherever you get podcasts.
Transcript
Shayle Kann: I’m Shayle Kann. I lead the early stage venture strategy at Energy Impact Partners. Welcome to Catalyst.
So Waymo’s first public ride service was in December 2018, years before the meteoric rides of LLMs that we’ve seen since then. And it’s interesting to think about what’s happened with autonomous vehicles, which were capable of driving on public roads, at least in some limited capacity even back then. Not all that limited, I should say, but limited nonetheless before the current wave of AI. Anyway, I was listening to an interview recently with the co- CEO of Waymo who described how they’ve been able to integrate the newer AI capabilities into their system and how that’s allowed them to basically supercharge their growth and to deal with a wider variety of edge conditions than they would have otherwise. Obviously, Waymo now is ubiquitous where I live in the Bay Area and increasingly is becoming so in many other cities.
So the path to autonomy for light duty vehicles seems very clear. Less so at this point for trucking. Though the market is huge, trucks hauled over 11.2 billion tons in the last year in the US alone. There have been a bunch of attempts at autonomy for trucking, and in some ways it intuitively seems like it should be an easier challenge, but in reality, turns out maybe to be a harder challenge. We’ll get back to that. Anyway, we just don’t have this equivalent of Waymo in trucking yet. So what would it look like for someone to start fresh now, fully immersed from day one in the new wave of AI, particularly in this context, having benefited from the advent of things like vision language action models, which is sort of an offshoot of LLMs aimed more at the physical world. By the way, I find autonomy interesting in its own regard, given that autonomy really goes hand in hand with electrification.
We’ve seen that already in the light duty vehicle world. I think we’re starting to see it now in heavy duty. In any event, Eyal Cohen is the founder and CEO of Humble Robotics, a company that we at EIP just announced we invested in a couple weeks ago. Eyal is a veteran of the autonomous trucking world. He worked at Apple, Uber, Spark AI, and Wabi before starting Humble. And his view is that starting with a clean sheet, both from a vehicle perspective and from a tech stack perspective, will allow Humble to dramatically accelerate adoption of trucking autonomy. So I talked to him about the broader world of autonomous trucking, where we are today and where we’re headed and what he’s building at Humble. That’s coming up after the break.
Shayle Kann: Eyal, welcome.
Eyal Cohen: Thanks for having me.
Shayle Kann: I have a bunch of questions for you. Let me start with this one. Waymo has done 200 million miles now on public roads and is driving me all over the complicated and messy streets of San Francisco. So passenger vehicle autonomy is clearly here and now across a bunch of cities, and yet we’re not there on trucking autonomy. Why is that?
Eyal Cohen: Well, that’s a great first question. So I think we need to go back a little bit to when autonomous trucking started and the history there, and it’ll help explain a little bit of that. So there’s been unmanned military defense experiments for a long time, but from an industry that’s not in defense, autonomous trucking really started in 2016. There were two companies, Starsky Robotics and Otto. And I was part of Otto. I joined pretty early. And I joined Otto at the time. I think there’s some fallacies I think maybe that we had at the time or misunderstandings about the space at the time. That’ll help explain the answer to your question. But when I joined, I had been working on passenger car autonomy with other efforts. I was at Apple for a bit. And the challenge that I saw was cities felt very hard.
At the time, given where the tech was, we were just like, “How are we going to solve San Francisco and everything that happens in San Francisco?” And the way the tech worked at the time, and still for a lot of companies works this way, is we were doing this very complicated HD mapping procedure where basically vehicles would drive and try to capture the world in 3D and try to maintain this sort of updated high definition map in 3D to locate the vehicle across to. It all felt very hard at the time. And when I learned about auto doing trucking, I was like, “Hey, this feels like a more straightforward problem, has great commercial application. Why don’t we go tackle that? ” I didn’t know too much about trucking. I mean, I like trucks. Most people are in awe of big vehicles, right? But the highways felt easier and they felt easier because when you drive on a highway, mostly what you do is go straight.
And what it turns out is that actually the passenger cars have a lot of structural advantages in developing an autonomous system that were very difficult for trucks on highway. And so even in 2016, at Otto, we did this beer run a hundred miles long with nobody in the front seat. And everybody felt like, okay, autonomous trucking’s right around the corner. What we found over time is the highways were way more difficult than we expected. They’re challenging. I mean, by the way, my premise here is that everything will get automated and passenger cars just happen to be first, but everything will get automated. It’s just that sort of the march of the technology, the progress. But the highways we thought would be easy, they turned out to be hard because of edge cases. You’re driving at a higher speed, you’re talking about a vehicle that could weigh 80,000 pounds.
In some cases, if it’s at full gross weight and you’re stopping distances long, trucks just have challenges on highway that make them more difficult
Shayle Kann: Would it be accurate to put it as like on a highway you probably have fewer edge cases than you have on like a San Francisco street, but it’s harder to manage. It’s harder to deal with those edge cases because you’re going at high speed and an 80,000 pound truck, et cetera.
Eyal Cohen: Yeah. Yeah. I think probably like different kinds of edge cases, the way I would put it. You see them less often, which is one key issue. A lot of the times you’re driving on the highways, nothing happens. And I used to actually joke that like you could just take a truck with no perception system and put it on the highway and as long as it stayed in the waiting line and stayed at 55 miles an hour, could probably complete 200 miles with no issue. It’s just that occasionally things happen on the highway that are very difficult to manage and you don’t see those events very often to be able to handle them and learn from them. So that’s one challenge. Another challenge is that the truck, the passenger car, it’s like a Waymo, when they are confused about what to do, when the robot doesn’t know what to do next, it can stop.
It could stop and assess. And it’s annoying and I’m sure you’ve seen videos of Waymo stuck, right? But it’s mostly annoying, not a hazard. So in San Francisco, when a passenger car stops, people go around and they honk, they call, annoyed, their local politicians, whatever, but mostly it’s a safe situation. You don’t have that advantage on a highway. So stopping a truck on the highway is actually extremely dangerous and could easily cause an accident, especially if there’s like a bend or a hill where you don’t see. It’s trucking, the margin for error is less. So all these things just sort of started to pile up in trucking and I think it just took longer to solve. It will get solved, but I think we, in 2016, when I started with Otto and Starsky, we thought they were easier. It was going to be an easier road. It just turned out that we had some misconceptions about the space.
Shayle Kann: And so where are we today in terms of autonomous trucking? I mean, you said Otto and Starsky were the first two. There have been a bunch of other, I guess, what I would call serious attempts since then. How far have we gotten?
Eyal Cohen: Yeah, many serious attempts, I would say. And by serious, I mean real technology development, real capital, real efforts. They’ve all kind of followed a similar blueprint, I would say. They take existing tractors that have been manufactured by OEM, sometimes in partnership with an OEM like Volvo, sometimes just sort of buying from a lot and doing a quick retrofit. And the attempts so far, there have been a few driverless runs, and by driverless, I mean nobody in the front seat. There’s a lot of debate about what driverless even is.
For example, I think Aurora, which is one of the largest players in this space, they do what they call driver list runs, but they sometimes have a safety observer. This is from their public writings. Somebody sort of watching the system, but they call it driverless as in maybe they don’t touch the wheel or engage in any way. So there have been driverless runs, even going back to that 2016 Otto. There’s a company called Bot auto that just I think yesterday did a commercial, what they call the first commercial driverless run. I saw that on LinkedIn. But I would say there’s not a regular driverless service for trucking on highway that exists today. And that’s 10 years after we’ve really put serious effort into this space.
Shayle Kann: And five years plus after passenger vehicles really hit the inflection point where those rides are getting taken all over the place in multiple cities now, and even on highways now, as of more recently, of course. So yeah, so it’s interesting. And I think when you take a step back, surprising, because intuitively I would have thought the same thing I guess that you intuitively thought, which is that like all things equally you would think trucking is easier. There’s just like less fewer things going on on a highway than there are certainly in the streets of a dead city like San Francisco. But yeah, it turns out that those things are harder to handle. One thing I’m curious about though is that this first wave of autonomous vehicles, both trucking and passenger vehicles, I would say, was built pre LLMs, pre the current wave of AI. And Waymo essentially made it to market before, certainly started offering commercial passenger rides before like GPT3, for example.
And so they probably weren’t leveraging like the latest and greatest from transformer world at that time, but obviously have been incorporating it since. And I’m curious what … So if you’re … The difference is between evolving a tech stack for autonomy that was built in 2016 or 2018 or 2020, and then layering on whatever you can do today versus starting fresh today, like what is the difference between those two?
Eyal Cohen: Yeah, great question. So the landscape obviously has changed a lot. And again, if I go back to that 2016 era, this was even before ML had really taken off in mass in at least autonomous vehicles. So to stay between the lane lines, we might just actually be hand coding, look at these two different pixels, is one yellow and is one not. That’s where the lane line is, very, very simple handcrafted kind of algorithms. So the whole space has evolved tremendously. And I think a lot of the companies have had to break down their stacks and re-engineer them multiple times and I’m sure Waymo, I’m sure Waymo’s have to do the same. But today, I’ll give you one example for Humble, my company, most of my time in this technology space, we’ve been pretty LiDAR heavy, like LiDAR first, I would say.
And even for trucking on highway, and there’s been attempts to develop these very long range LiDARs that could see 300, 400 meters, so you can handle the stopping distance of a truck. And the LiDAR technology for a long time was really good. And you can … It was a little challenging to make it robust and reliable and at scale, but the technology itself was very good and cameras kind of lagged behind. We weren’t doing so much with the vision side of it. Today, for me, at Humble, we flipped that and we do a lot more with vision algorithms, we’re camera first, and that’s because of just the evolution of this technology has just changed so much where now you get to all this intelligence kind of, I want to say for free, but there’s … When I take an open source, the corollary to an LOM is a VLM, like a visual language model.
And if I take an off the shelf VLM, something that understands an image and tells me what’s in it, it’s very, very good. There’s a lot of intelligence baked in there. And so I think that every company kind of has to kind of rethink their stacks as the technology evolves. What really hasn’t changed that much though is some of the validation that you do, some of the safety engineering that you do, the evolution of the hardware technology, because that’s gotten a lot more robust. So when you start a new effort now, you’re kind of using a pretty different technical approach for the brain, the AI, but you’re using a lot of very well understood techniques for the rest of it, for how you develop a safe system. That’s evolved tremendously in 10 years, et cetera.
Shayle Kann: For those who aren’t already familiar, what are these vision language models, these VLMs that you referred to? What is it and what does it actually enable?
Eyal Cohen: Yeah, so VLM is like a LLM corollary is the way I think about it. And it’s taking an image as an input and providing some interpretation of what that image is happening in that image. And it’s not something that really I had used in my past. This is very new to me also when we started Humble and our head of autonomy, Drew Gray, has been really, really in depth in it in the VLM space. But what I’ve seen coming out of it, it’s like what I think a lot of people experience now, like if you go to ChatGPT and you upload an image and you say, “Hey, what’s going on in this image?” It has a pretty good understanding of what’s going on in that image. And sometimes you will find intelligence baked in there or some … I guess intelligence is like a hotly debated concept here right now, but bear with me, like some understanding of the image that goes beyond there’s just a dog in here.
It’s like, there’s a dog that may have come from that door on the left or this construction cone is sitting on a pickup truck, not on the highway, and therefore it’s probably not an important construction element because it’s just being carried by a truck, just to give an example of an edge case. And so the VLM brings some of that understanding right from the jump, and that’s really cool and pretty novel. I mean, when I started 10 years ago, nothing like this existed, and so you had to sort of like code in all the intelligence, like really try to understand what the features were. We would take imagery or LIDAR data, we would send it to get labeled, people would put boxes, I’m sure you’ve seen images of boxes around it that say like dog or human or whatever, right?
And we would say, “If you see a human, you probably want to do this kind of thing.” I mean, it’s more complicated than that, but we didn’t get this intelligence layer for free. And there’s open source technology out there that is very, very good. You’ll find free VLMs or open source VLMs that have a lot of intelligence just available, which is like also kind of a wild place to be. So yeah, quite an evolution.
Shayle Kann: My understanding of this sort of like sensor debate in autonomy world, at least for passenger vehicles, right? There’s the like LiDAR, radar, fusion side, which seems to be Waymo does a little bit of everything. Then there’s like Tesla, which is camera first and camera, basically maybe potentially a little bit of radar. But as I think about it, so I could see how these VLMs as an extension or corollary to LLMs allow you to do a lot more with video and just like get further faster on camera data. But my understanding was always that the limitations on camera for autonomy are more around like it can get obscured in certain conditions and things like that. And that’s where you want your LiDAR or whatever, which doesn’t have the same set of issues. Is your view like you can go camera only as a result of the VLMs, or is it you can lean more heavily on camera, but you still want this sensor fusion approach?
Eyal Cohen: I think especially for trucks and given what I was talking about earlier with this 80,000 pound vehicle that you might be moving on the road, you want it to be as safe as it can possibly be. And so, in my mind, it’s not a dogmatic debate about camera or a LiDAR. It’s what’s the best technology for the moment that makes it the safest? And from my perspective, for a vehicle, a truck, for example, you put on the road, you would want it to have camera, LiDAR and radar to do a level four, level four being driverless, to do a level four truck today. And the reason I say that is because you want to be able to see the world in multiple ways, but there’s always a sensor doing the most heavy lifting. It’s not sort of this co-equal democracy between the three sensors. It’s usually like some sort of priority is put on some algorithms depending on what you’re doing.
LiDAR doesn’t see traffic lights very well, or it doesn’t see the red green, so you’d use camera primarily for that. But LiDAR can see better at night. Radar can see through weather in some cases. And so if you want to make a safe product and trucking in particular, you want to really take advantage of all those today. A human, and when you think about the end state, a human is effectively two cameras, right? Two eyes, two cameras, and they’re able to navigate the world fairly well. And so you could make the argument that over time trucks would go to a vision only system, but I think that the debate, sometimes you see around like Tesla and Waymo and others, it feels a little dogmatic. I think most of us on the trucking side have taken just the approach that, hey, our margin for error is very low here.
It’s a very heavy vehicle. We have to be extremely safe. So we just put as much equipment as we can to sort of try to guarantee that safety and get very creative on the algorithm side and how to take all that data from the three different modalities and put them together.
Shayle Kann: So I think of Humble as having at least two things that are kind of new and different in this space. One is just, to some extent, it’s like timing. You’re starting fresh now, and so you get to build from the ground up using VLMs and these things that didn’t exist before, and benefiting from all the learnings that we’ve seen on passenger autonomy and so on. The other is the form factor of the vehicle. So you mentioned that most of the attempts in autonomous trucking have been attaching a bunch of sensors to an existing truck, which is also, by the way, how it’s been mostly historically in passenger autonomy, right?
That’s what Waymo’s doing on these Jaguar vehicles and these other ones. Tesla maybe is a little different with the robotaxi, but those aren’t really out on the market yet. So that is similar. You are taking a clean sheet approach and building a vehicle that doesn’t even have a cab.
I guess in the long arc of history, of course, that’s how it’s going to end up, right? When we don’t need a driver, we shouldn’t have a space and a driver for a vehicle, but apart from that, give me the thinking that led you to building a new vehicle.
Eyal Cohen: Yeah. So right at Humble, we have this cabless autonomous electric class A truck, right? And the thinking that got me here was, it’s a couple of things that’s kind of interesting. First I was like, okay, if you were to imagine that long arc of future, like you’re saying, what does that vehicle look like? What is the simplest possible vehicle to move freight? And it’s like a box with wheels, right? Basically, it’s a platform concept where either a container is being loaded onto that platform or maybe it’s just a box of the box truck moving. And that would be, in theory, the lowest possible cost of moving freight, right? It can’t possibly get lower than that, I don’t think. Maybe there’s some new mechanism to do it, at least for on-road trucking. So that was the first though that got us here. And the second was like, okay, is this possible today?
Can we do this with the technology and where it is today? And that’s where we started exploring it with Humble and the answer became like, yeah, the technology is there. Basically, there’s enough here where we can take that long arc of history and move it in a little bit, move it forward. And then along the way, because you’re doing a clean sheet design, you could think about all these problems that you’ve run into into the space. So I’ll give a couple examples of what you can just rethink with this kind of vehicle. So one way to think about our vehicle is that it combines a tractor and a trailer together into the single platform, right? And a lot of challenges with doing this, by the way, so –
Shayle Kann: Right, because part of the market is like the tractor and the trailer are owned by different entities as it stands today, right? Some, it’s not the entire market. Sometimes it’s the same player, but they get separated sometimes and sent off in different directions.
Eyal Cohen: They get separated and sometimes for very good reasons. Trailers are relatively inexpensive, tractors are relatively expensive. So there’s a lot of kind of interesting thinking around trying to combine this concept, but just from the technology side, for example, if you’ve made the tractor smart, like the other autonomous truck players and you’ve left the trailer conventional or like a dumb trailer, quote unquote, right? You can’t, for example, put sensors on the back of the trailer. Like you have a smart tractor, but you’re just taking trailers from everywhere. So you can’t see behind you. You can’t see directly behind you. If you can’t see directly behind you, you cannot handle, for example, an accident where somebody just drives right into you. And I think there’s maybe some clever ways to do that, but that’s tricky. You cannot back into a dock because that requires some amount of understanding of what’s going on behind you.
One interesting, this is just kind of like an inside baseball thing, but one interesting challenge we’ve had in the industry is that trucks today, if they’re pulled over, they’re required by law to deploy warning triangles and those warning triangles go behind the vehicle. If you only have made the tractor smart and the trailer has not been touched and you don’t have a driver there, how do you deploy those warning triangles? And so companies like Aurora and I think Waymo when they were working on trucking, they proposed using lights that are high up on the vehicle to indicate that the vehicle stalled as a replacement for the triangles, but there’s advantages and disadvantages to that, right? And so it’s being, just being able to access the rear of the vehicle and have this full vision from a technology perspective just allows us to do more. And when I was thinking about what is the real future here for freight, like again, sort of now long arc, but what you would want is all freight to be fully automated, completely hands off, nobody touching it.
It goes to a warehouse, it loads up, it loads goods from a warehouse into that vehicle and drives its destination and unloads, right? And completely hands off, no human. I think that’s where we want to get to, but to do that, you have to kind of treat the whole vehicle as smart, not just the tractor. And so that’s how we got … So it’s kind of this interesting combination for us of the technology is ready, the simplest concept that we can come up with for moving freight is a catalyst vehicle, and there’s actually a real structural technology advantage to doing it in the long term. And so that’s where Humble is today.
Shayle Kann: You mentioned your vehicle is electric autonomous. There’s this very interesting and very appealing direct correlation between autonomy and electrification that you see broadly. I presume you sort of knew from day one this was going to be an electric truck, but electric truck, I mean, setting aside autonomy even, it’s got its own set of open questions on, of course, things like range and weight of the battery and charging infrastructure and charge time and so on. So how do you think about the electric component of your vehicle?
Eyal Cohen: Yeah. And I think it starts with that vision that I talked about of like having fully hands off freight. If you were going to have fully hands off freight, you would want it to be electric to charge, to handle the charging in an automated way. It’s really hard to do that with diesel. How are you going to get a diesel pump into a … I mean, you could maybe do it with robots, but it could be challenging. So part of the electric story for me is just like, how do you get to this fully automated freight vision?
So electric is good from that perspective. There are challenges charging … Forget the autonomy side, right? Electric trucks in the US have had a challenging rollout. The electric trucks themselves have been fairly expensive, in some cases, four or $500,000. A truck today, a tractor today is somewhere between 150 and 250, depending on what you’re buying. So tractors, a very expensive electric truck, there’s charging infrastructure that needs to exist. Electric trucks are kind of having a moment again now because of the volatility in the world and some of the challenges that are going on. But in general, it’s been a tough rollout. But what we’ve seen in countries like China is that electric trucks are really taking off because the infrastructure developed and once it’s there and the costs are where they need to be, it makes a lot of sense. It’s harder on the long haul.
And I think that’s something that as an industry we have to reckon with a bit. You have to put a very large battery to do long haul. It is very heavy. You need a lot of power to charge, you need a lot of charging infrastructure. But for use cases that are more local, short haul, drayage, drainage is moving, freight out of ports, electric makes a lot of sense. And so my career started in electrification. I worked … Bay Area was very big for electrification for a long time. That kind of moved out a little bit, but I loved working on motors and batteries and the technology there. So I always, in the back of my mind, wanted to get back to it. And this was an opportunity for me because of that vision, that long-term vision, because it makes a lot of sense for the short haul moves that we’re aiming to do with Humble.
And like I said, electric trucks are having a moment. It is the technology that some countries have adopted in mass. It just needs to be applied in the right way. And there’s always going to be challenges in tech rollout. Charging is certainly one of them, but they’re solvable. They’re all solvable.
Shayle Kann: Do you think you need to solve them? There’s like a chicken or egg challenge often in things like this where it’s like you need that charging infrastructure to be there in order for your customers to buy and utilize your truck. You don’t necessarily want to be an EV charging company, I suspect, but somebody’s got to do it. And that there has to be enough of a demand signal from your customers such that somebody will build the charging infrastructure. And if it’s third party owned charging infrastructure, make money on it. It’s got to be high enough utilization and so on. I think we’re starting to see this happen at ports and things like that to some extent where you’ve seen a little bit more electrification already, but is your view kind of like a build it and they will come sort of thing. If we build the vehicle, our customers will want the vehicle, charging will show up. Or do you need to be more proactive?
Eyal Cohen: Yeah, I think it’s a little bit of both. So there has been quietly an industry built up on electric trucking. There are a few startups and companies out there that have been working on depots and solving electric trucking. And I think there’s been some shift to more short haul. Tesla Semi obviously I think just rolled out of production, which is really great for the industry and that’ll induce some charging efforts. So it has been happening. It’s just been happening relatively slowly, I would say, compared to what the industry expected. But I feel like it’s inevitable. And as Humble grows, my conversations with customers around this is we will help you solve it. It might not be that Humble itself is developing technology around charging. We have our hands full with autonomy and a clean sheet design. There’s a lot there, right? But we will help the customers solve it because we know everybody in the industry.
We’ve been working with them for a long time and there’s a myriad of solutions, right? You could do charging at the ports and the warehouses themselves. You can have depots. There’s public infrastructure, private infrastructure. But I think what we’ll see is, again, especially with what’s going on in the world right now, I think we’ll see a steady improvement in that. And eventually you’ll have real uptake on electric trucking because from a cost structure perspective, once that charging is in place and you’ve sort of changed your operations and navigated that a little bit, there’s a lot of advantage to it.
Shayle Kann: What do you think about the rollout of autonomy … Back to autonomous, sorry, not just electric, but what’s the rollout of autonomous trucking going to look like? I think people now have an intuition for what it looks like on the passenger side. It’s sort of like a, okay, company X, usually Waymo, maybe it’s going to be Tesla soon, maybe it’s going to be Uber at some point says, okay, we’re now offering publicly available rides in city X and it’s ring fenced. You can only go within these boundaries. And they do that for a while and then they expand the boundaries and then they go to a new city and so on. So now we have a sense of like, here’s what’s coming in terms of passenger autonomy. I imagine it can’t quite work the same in trucking. So what’s that going to look like? Paint me that picture.
Eyal Cohen: Yeah. So I’ll put Humble aside for a second and just look at the industry as a whole. And you have a number of very large players, very well capitalized, some public companies, all kind of going after the on- road, long haul autonomous trucking. And they will continue to work on that and you will see autonomous trucks on the highway and I think you’ll see them very soon and they will be driverless. All the sort of pieces have started to come into place like hardware and supply chain and regulatory efforts. That took a while to sort of make sure state and federal regulators understood kind of the technology and how to deploy it. So those large players, I think they will start deploying … I mean, traditionally they’ve been calling it a hub to hub model, right where it’s like you can imagine at either end of a highway segment, a destination for autonomous truck to go and drop off freight.
Sometimes you’ll see it go right to customers. I don’t think you’ll see generalized solutions where there’s just like autonomous trucks zagging everywhere for a while. But I think what you’ll see is some specific segments, mostly in Texas, where you’ll see occasional autonomous trucks and some scale applied there. That’s for the long haul side, at least in the space that Humble plays in for Class A trucking heavy haul and the short haul, there really hasn’t been a whole lot of effort in that space yet. So we’re one of the first companies to tackle that. And so we’ll see what our rollout looks like over time and how that goes. We’re fairly young, we’re less than a year, but we’re moving very fast on that front. But I think you’ll see long haul autonomous trucks moving. I think the question will shift to, can we get a truck safely across the road in most cases to, can we handle the operational concerns?
Can we handle the unit economics? That’s a challenge. If you take a very expensive truck and then you make it significantly more expensive and then you’re saying we’ve saved on some labor costs, you have to make that argument work really well.
Shayle Kann: I was going to ask you that question on unit economics. I mean, I guess one way to get at it is like, in a normal truck move, this is going to vary substantially, I’m sure, in terms of long haul versus short haul and other things, but like what portion of the fully loaded cost of delivery is the driver? How much savings do you get economically removing the driver?
Eyal Cohen: Yeah, so don’t quote me, but we use ATRI, it’s an industry guide for trucking costs and trucking costs in ‘26 or ‘25 will have to be about $2.30, 40 cents a mile, at least for the long haul side. And I think the driver wages are about a dollar, a dollar ton of that per mile, so fairly significant.
Shayle Kann: 30, 40%, something like that, is the driver. So that’s maybe if you’re trying to save money and go autonomous, that’s like the headroom you have to play with to make the vehicle more expensive. If you have to make the vehicle more expensive because you’re adding sensors and things like that.
Eyal Cohen: Correct. Yes, exactly. So it’s like you have to offset that labor reduction or you have some room to play with there, but you’ll have some remote assistance. The vehicle will be a lot more expensive because of the sensors. There’s no getting around that. A lot of the current OEM efforts to … Basically the way it works a lot in industry now is that an OEM will make a tractor and will partner with an autonomous provider to put the sensors on it. And they try to do this at the factory. That’s the goal. But often they’re using like fairly expensive tractors, like the higher end models because they have more room in the tractor for equipment. They call those sleepers where like a driver might sleep and they replace some of the bed with a computer, for example. So those vehicles are expensive. They become very expensive until they’re at scale.
So there’s a challenge for sure in the unit economics. And I think like you even see that with Waymo’s today, right? They have to get the unit economics down low enough to make the labor argument worth it. So I think that’ll take time. That’ll take time. And part of the story for Humble and the reason that I started thinking about the lowest possible cost of moving freight was the unit economics. I was like, well, if you remove a cab, you’ve taken out some significant costs from the vehicle. You’ve also taken out significant weight from the vehicle. And that allows us to kind of rethink the unit economics a little bit.
Shayle Kann: Yeah. It strikes me that at least for a period of time, Waymo is getting away with inferior unit economics. They definitely have inferior unit economics. And I can tell you from personal experience, as I’m sure you have too, like I ride Waymo’s around even though they’re more expensive than Uber’s right now because it’s cool and it’s novel and-
Eyal Cohen: Novel. Yeah.
Shayle Kann: Yeah. I doubt that same dynamic exists to the same extent in trucking. It’s like a classic thing of a B2C versus a B2B market. 100%. You have to save money, I assume in trucking in order for it to be adopted at any meaningful scale, maybe people will pilot something, but I don’t know why else they would do anything at scale unless they can save money.
Eyal Cohen: Yeah. I think that’s a real challenge. There’s a real challenge in what you’re saying and it’s very true. It’s like in this B2B market, when you’re dealing with freight, there’s a safety component, there’s a reliability component, like is my freight coming on time, but ultimately it’s the cost of moving freight and it has to be advantage to a shipper that’s moving freight to want to use an autonomous service in some way. I’ll give you an example of a challenge that the industry faces on the long haul side. Originally, a lot of the ideas were basically make a rail, like I said. So you can imagine at the highway, there’s like a depot right at the highway and you would aggregate freight at this depot and an autonomous truck would take it hundreds of miles, maybe a thousand miles, like a long distance, and drop off the freight at the other end of that segment at another depot.
And the thinking is like, yeah, we will do these sort of short moves to the depots and then the autonomous truck with its cost advantage structure would move the freight the long haul and then you do these short moves again. Well, the challenge in that is today you don’t have those depots and somebody has to build those depots and those depots are friction. Who is moving the freight to that depot and why do they want to move it to that depot as opposed to just what they do today, which is just take it from their warehouse to the other end. And so you can see that like some of the … I think the economics here are tough and in this industry in particular, the economics matter a lot. So I think Waymo has managed the cool factor and people are willing to pay more.
And I think there is an experience component to it, right? I think people sometimes say they prefer Waymo’s for the experience. I think Waymo did an excellent job with the interior and the experience, right? But so I think they have a little bit of an advantage there. I don’t see that as much in freight. It’s hard to … What’s a premium service in freight? I mean, one example, I mean, to give the counter argument, you could say, okay, this truck could go day and night, right? There’s no restriction on how often that vehicle can go. Maybe there’s like some argument there, but in general, like, yeah, you have to get the cost structure down. Part of the reason I started humble in a way, I was just really thinking about the customers, what they need, what their challenges are, and how to just get the cost of freight to be as low and meaningful as possible.
Shayle Kann: I guess final question for you, what’s the regulatory landscape like? Are we allowed to run driverless vehicles on … And I guess, is it tied to, are you on a city street or in a drayage situation versus a highway and are those regulated differently or is trucking regulated as one category?
Eyal Cohen: Yeah, that’s a great question. So the regulatory landscape is interesting and fun. I actually kind of really enjoy this part of the industry and I love talking to the regulators about it because I kind of feel like you want … We’re all breaking new ground, the regulators and the industry, and it’s good to do that hand in hand and have good dialogue on it. The situation today is there’s … We work with NHTSA, that’s a major regulator for the trucking industry. We work with the FMCSA. It’s another regulator for the trucking industry on the federal level. There’s state laws that are state by state. A lot of the testing for autonomous trucks has happened in Texas because the regulatory landscape there has been favorable to autonomous trucks. In California, a week ago, you would not have been legally permitted to do a driverless truck and today you are.
So there was some legislation or some new rulemaking there that allows for it with certain conditions. So it’s an evolving landscape. For us developing this cabless vehicle, it’s got a sort of additional regulatory challenge in that you’re changing what the truck looks like, right? There’s no steering wheel, there’s no windshield, right? Windshield is required by law. So how do you navigate that? So the way that it works is that we talk to the regulators about what we’re trying to do and we come up with a good plan for how do we do this? How do we test it to be safe? Our vehicle looks both like a tractor and a trailer. In some ways, you could think of it as like a smart trailer that just drives around by itself, right? So if it’s a smart trailer, how do you regulate a smart trailer versus a smart truck? There’s a lot of questions like that that we just work with the regulators on. But I would say that the regulators in general in this space have been very good to work with and everybody kind of sees where the technology is going and they just want to make sure that it’s done in the safest way possible and thoughtfully, safely.
And there’s also this other component that regulators are thinking about, which is what’s happening around the world. We see actually a lot of driverless efforts deploying in places like China, right? So are we being competitive in that regard while also being safe? But yeah, it’s an interesting question about the regulatory side and actually a really fun one to kind of figure out hand in hand with them.
Shayle Kann: All right, Eyal, thank you so much for doing this. Super interesting.
Eyal Cohen: Thanks for having me.
Shayle Kann: Eyal Cohen is the founder and CEO of Humble Robotics. This show is a production of Latitude Media. You can head over to latitudemedia.com for links to today’s topics. This episode is produced by Max Savage Levenson, mixing and theme song by Sean Marquand. Anne Bailey edits the video version of the show. Stephen Lacey is our executive editor. I’m Shayle Kann, and this is Catalyst.
