For forty years the advice to founders has been the same: go asset-light, take a thin slice, scale software globally without heavy capital. But software alone struggles to reshape physical industries like transport, energy and logistics, and as AI commoditizes the bits, the durable advantage shifts back to companies willing to own the atoms too.
Keith Bradbury makes the case for the vertical integrator, using Ember, the all-electric coach operator he co-founded, as the worked example. From building its own charging hubs and custom vehicle software to an "API for atoms" that even predicts when an onboard toilet needs emptying, he shows how deep integration produces structurally better unit economics and problems no asset-light rival can fix. It demands first-principles thinking, comfort with scary long-duration assets, and a genuine commitment to long-termism and compounding gains.
Auto-generated transcript - may contain errors.
Tap a timestamp to jump the video.
Hi, everyone. Good to be here. I was trying to figure out actually a second ago whether this is the graveyard shift, whether you got put at the end, or whether this is really the wrap up keynote. So I'm gonna treat it as the wrap up keynote.
So I'm gonna be talking to you about vertical integrators, what they are, how it relates to what we're building at Ember, and why I think the future is gonna be all about them. Specifically, we're gonna be talking about what an API for ATOM is beyond just sounding like a a buzzword.
I think it is a real thing. But let's start with a pitch. There's probably founders and investors in the room, and so I wanted to start with a bit of a pitch. This is Continental Transact Inc. And highly intensive on assets, it's operating in a declining industry, its competitors are monopolies, and it's a super, super complicated business model.
And I think if there are any investors in the room, I know there's one, but he doesn't count because he's our investor. I think this looks absolutely terrible, and I think that is the current paradigm. When people see this type of stuff, they think that sounds bad.
You've gotta simplify, and you need to be taking a slice. My thesis, actually, is that this is a great business model and that we should be seeing more of it. Hopefully, by the end, you'll agree in some way. So let's start with a a bit of a history lesson.
This is a kind of technology, I guess, a market curve of what happens following technological revolutions. And, typically, what you see is that following a technological revolution, you have vertical integrators. You have people that are moving first into an industry, and there is nothing in front of them.
There is just opportunity. There's no infrastructure. There's no rails. There's nothing to build on. There's no one to copy. And so by necessity, they have to be the vertical integrators. But what you see is that over time, as the markets mature, people find ways to take thin slices.
They start unbundling the stack, and you start having this move down the curve. And, typically, there's an overcorrection. If you were here for two talks ago, there was this talk of inertia. And you have this inertia where people thin slice, thin slice. They go down.
They realize they've gutted it too much, and they start coming back up the curve. And, crucially, that right hand part of the curve where you start getting back to the sweet spot, that's what really shows the kind of maturity of a technology technology revolution.
When you start seeing companies there, you know you're getting towards the end of a paradigm. And we see this. You can look at the kind of industrial second industrial revolution post nineteen hundred, the world of atoms, and what happened there. Ford was a very famous example where their factory had raw materials going in at one side and had cars coming out the other.
They literally owned the mines. They built the ships. They built the railroads. Our investor is dying right now because he thinks we're gonna do this. And they took absolutely everything on this vertically integrated model. And they did that very successfully for sixty years.
And then what you saw was companies like Nike, who came they specifically started as a brand and marketing company. They outsourced everything else, and they were also phenomenally successful. But now you've got companies like IKEA, and I think IKEA is a particularly interesting example because they did go up and down the supply chain.
They do have the retail units. But they took this concept of knockdown production and really sent it everywhere. So they worked with furniture producers to get it made like that. They pushed that onto their customers so everyone is building at home, and that really changed the industry.
And you see the same with the computer revolution, so in the world of bits. So if you go back to the sixties and seventies and the dawn of the silicon revolution, you see the same. You see companies that were doing vertically integrated approaches.
They were doing the hardware and the software. And then you've seen this decline into thin slice platforms and other businesses. And in conjunction with this, the last twenty years has seen, well, let's go forty years. The last forty years has seen this massive rise of venture capital.
It's seen money going into these thin slice SaaS asset light businesses. And so that's become the kind of the predominant paradigm. If you're a founder and you wanna build a big business, everyone is telling you, go asset light, find a way that you can scale, find a way that you can scale globally, and do it in a way that doesn't require huge amounts of capital.
My my view on this is that we are reaching the end of this paradigm. We're starting to get towards that sweet spot where companies who are focused just on the bits are reaching kind of the end of the road. And so we come to the limits of bits.
And my view now is that these kind of pure software solutions are are really insufficient for solving physical problems. And if you look at what I think are the most kind of pressing challenges and problems to solve in the world, they are very physical problems.
So stuff in transportation, obviously, because that's us, stuff in energy, logistics, agriculture, manufacturing, all of those are physical problems. You can't just take a software layer, apply it to something that exists, and expect it to fundamentally reshape industries. And so that brings me on to the new thesis or my new thesis, which hopefully everyone here will adopt as their new thesis by the end of this talk.
So we're now in atoms and bits. A a kind of crucial part of the reason that I think we've arrived here is that we've reached the maturity of bits. So where we are now in terms of kind of software and what software is capable of, it's so good.
In in many ways, could kind of say, people are already talking about this, that kind of AI is gonna commoditize software. If you believe that and if you kind of extrapolate on that and it truly does get commoditized, then kind of where is the advantage if everyone can do that?
I think the advantage is gonna come from this, this combination of atoms and bits, And it's the deep integration with the with the atoms. That's how you kind of fundamentally reimagine, rebuild physical systems, and it requires a full stack model. This is one of the prerequisites, this native understanding of both hardware and software.
There's not many people who feel like they can both pick up an oscilloscope and take a multimeter and go and interact with something in the physical world and then go and write a bunch of code and understand how to write the system that can control that.
And I think that companies that are going into this new paradigm need to be super comfortable with this idea of connecting the atoms and the bits. And the API for atoms, I guess, is you can think of it as that sort of middle bit there that's enabling these two things to talk together.
But it's kind of more than just a connector. It's more than a digital twin. What what it's really doing is enabling you to both talk to, program, control, and respond in real time to your physical assets. So that's our kind of history lesson and up to the current paradigm over.
So now I wanna talk a little bit about how it's relevant to Ember, what Ember is. I guess people might be wondering, actually, what do we do? And so this is us. We are the world's first all electric coach so And so the ATMS the ATMS part of our business certainly should be pretty obvious.
I think what's less obvious is what our thesis, what our more fundamental thesis is, and that is that we can rebuild the passenger experience. We can do it on a structurally better cost stack, and that we can do it all with zero emissions.
Those are the kind of three things that kind of come together. To do that, you need this kind of you need to reset your view on the world in a way. You need to go back to this atoms plus bits paradigm. You need to take a systems level approach to coming up with that solution, and you need to become a vertical integrator.
And so that's what we did. So what does that mean, practically, I suppose? I was I hate doing elevator pitches. And I was at an investor conference last year, and I was literally walking in to a lift on my own when someone else just kind of, you know, nipped through the doors next to me.
And I saw this person had a JPMorgan badge on, and he said, you know, very casually, he just said, what do you do? And I actually burst out laughing because I was in a lift with this guy, and I said to him, this is probably a terrible answer.
What I said was, I'm not gonna have time to explain to you what we do by the time we get to the floor. If you've got thirty minutes and we can go for a coffee, I'll explain it, but otherwise, it's just not worth getting into.
And that's because it's complicated. We are a vertical integrator. We've taken on a lot of the stack. In many ways, the the integration of these things, that is the innovation in our business. That requires systems thinking, but what it has given us is total control of the customer experience and total control of the value chain, near total control of the value chain.
And I'll come on to explain kind of what that means. That's why we have fundamentally better unit economics than anyone in the existing market. So the first part of those four things is the energy side. So I've told you that we run buses, and we'll come into that.
But we also build, own, and operate charging a charging network, specifically designed for large high utilization vehicles. So the operating model is that the vehicles are coming in and charging twenty four seven. They're kind of coming in, charging quickly, and going out. Our hubs are capable of more than megawatt charging.
So for those of you who maybe have some involvement with EVs, you'll kind of know what these numbers mean. So our hubs themselves are capable of more than megawatt charging. We're incorporating battery storage solutions on-site to optimize our energy from the grid. We're also building on-site generation, so we have both solar and wind in different locations so we can optimize how we're getting electricity into the buses.
And often when the wind is blowing, as it does in Scotland a lot, it's literally coming from that wind turbine directly into a substation and directly to our vehicles. It's a kind of feel good thing, but fundamentally, it's going into our cost advantage.
We go deep into kind of the design of these sites. So what you see here is a kind of aerial shot of a site, which was under development up in Aberdeen. It's now complete. But because we're going super deep into things on electrical design, the choice of hardware, what's going into the ground, how deep we dig those trenches in the ground, do we deep dig them fifty centimeters or forty centimeters?
That has a cost difference. We do all of that ourselves. We're optimizing all of these parts of the charging hub, and that's why now we can build one of these charging hubs at half the cost of anyone else. Again, that's coming into kind of how you get to these structurally better unit economics is by going deep into these different areas.
And then we've got the fleet, so the second part of this stack. On the fleet, we've gone quite deep into vehicles. So we don't manufacture our own vehicles, so we do that with a third party supplier in China. They're the ones with the massive factory and the entire supply chain for it.
But we have gone into kind of quite deep custom spec with them to get our own hardware set up, so which is unique to us, which we have exclusivity on. And, also, we've got custom software. So the actual software that is running on the vehicles is custom to us.
It gives us data on every single thing that you can possibly imagine that comes from the vehicle and the batteries twenty four hours a day. I don't know how many rows our database has there, but it will be probably I don't know if it's even in the billions now, but some massive number.
The result of going deep into those vehicles is that our EVs are the highest mileage EVs in the world, as far as we know. If anyone in the room knows something different, please tell me afterwards. But we are accruing miles faster than anyone else.
We're doing about two hundred and fifty thousand kilometers per vehicle per year. We have the fastest charging vehicles. So our current vehicles so I said our charging hubs are capable of more than a megawatt. Our vehicles are capable of six hundred kilowatts, so they're a little bit lagging behind.
But six hundred kilowatts is pretty fast. We have no downtime disadvantage compared to diesel, for example, because we can charge them so quickly. As a result, we now have the biggest electric coach fleet in Europe, and that's only increasing. Then there's a network.
So, you know, kind of continuing through the chain of complexity. We design, operate, and optimize our own network. So we run twenty four seven, three six five. If you wanna get a bus on Christmas Day, we can get you a bus on Christmas Day.
But we don't just operate end to end. So this graphic is showing actually real trips between, a bunch of our our stops near here. The way that we do, our trips is that we go city to city, but we call it many towns and villages along the way.
And we do that using demand responsive routes, and I'll come on to explain how we do that. Our approach really is to try and replicate the flexibility that you have as a car user. And I think that's an important kind of product, like, mentality that we've got.
We're trying to replicate that because that's where all of the miles are. And the result is that people actually choose to use us. Not instead of the existing coach companies, They choose to use us instead of cars or trains. And that's a really big one, and I'll kind of talk about that in a minute as well.
And then the final part, this is the platform. So the platform is what we can refer to as our API for ATOM's. It's the kind of connection between the physical and digital world. It's got everything. So from battery data to driver, rotor design, through to vehicle allocation, through to all the scheduling and optimization, that's in there.
And it doesn't just give us an understanding. It actually gives us control of the stack, and that's one of the reasons that we can do these demand responsive routes. So the way our demand responsive routes work are that we have a bus that's, let's say, driving from Dundee to Edinburgh.
There's somebody due to get on in ten minutes. The driver knows that. He's driving down, the motorway. And then that passenger decides, actually, I don't I'm not gonna travel. I'm gonna cancel my ticket. So they cancel their ticket. Immediately, that goes notifies the driver.
Now the driver will see the navigation has changed. They're no longer driving to that bus stop because there's no passengers to get on or off. That means that the end to end journey time is gonna be reduced for the people who are on board.
They're not driving to a stop that has no passengers. And for us as a business, we're not wasting time driving to empty stops. It was actually one of my real pet peeves as a bus passenger for many years going to empty stops, deviating.
You get there. There's nobody there. You're thinking, okay. Great. What was the point of that? Kind of we have solved that. And all of this was built in house. That can feel slow. As a company that is doing vertical integration, that can feel slow.
But it is crucial to have this connection between the atoms and the bits. And, fundamentally, what we've built in Ember OS is different to anything that exists in the market. There is nothing like this, as far as we know. Again, I can always be wrong about this stuff, so just come and tell me.
So what does exist in the market? Like, what is the status quo? This. This is Flixbus. They are what I would call the darling of the of the coach industry in Europe and globally, actually. They've raised over a billion. They're worth, I don't know how much, something probably closer to five to ten billion.
This is what you read if you go to their review pages. I really have not cherry picked this. You know? Everyone is welcome just to go and read all of their reviews. It's not that they don't have any good reviews, but they have a lot of this kind of thing.
And I think it's rubbish. I don't think this is how you change markets. I don't think this is how you kind of sell the benefit of shared transport into society. I don't think this is what is gonna get people to leave their cars at home and, you know, do have some positive contribution to the climate crisis.
And to me, this is a fundamental limitation of an asset light business model, where they have tried to be a marketing company and take this asset light approach to scaling. And so that's where I think the ATOM plus bits is completely required to solve problems like this.
So I wanna spend a little bit of time talking shit, basically. So how do toilets work in an atoms plus bits world? So if you wanna solve this problem, the first thing you need to do is really understand the problem. And when I say really understand, I mean I don't mean, like, reading the spec of a water pump in a PDF.
I mean actually going out and getting the pump and connecting it up and seeing when it works and when it doesn't work. You need to understand how toilets are actually being used. You need to know, you know, sounds a bit gross, but you need to know, are people doing number ones or number twos?
Are people from some stops using them more or less? Are some passenger types using them more or less? Is it by time of day? Like, what is the thing that is causing people to use your toilets? You need to make sure you have that data to get some insights.
And then you have to imagine what does the system level solution look like. And for us, it was kind of an extra challenge because the first coaches that we we ever bought from China, toilets were not an option. So it was like just a you couldn't get it on an electric coach.
You not get a toilet. And so we decided, okay. Let's do toilets ourselves. We're gonna just get the coaches, and then we're gonna do the toilets in the UK. And to do that, we went really deep. We went we started looking at all of the available sensors.
What can you measure? How can you measure it? And then once we understood what the sensors could do, we wanted to make sure we had that data in the can feed of the vehicle. It's one thing to have the data. It's another thing to be able to do something with that data.
It's kind of not good enough just to have the data. So we did that. We worked with the manufacturer to integrate the data into the can feed of the vehicle, which got sent to our system twenty four seven every second of every day.
We know how full that toilet tank is. From there, we then combined it with all of the other data that we have in Ember OS. From there, we can then start to do interesting things and just really solve this problem of toilets being out of order or broken.
And so what we can now do is we can predict an action when a toilet actually does need to be emptied, so when the tank is getting full. And so the system now can automatically send an alert. It can know where the vehicle is.
It can know how much poop is in the tank, and it can send an alert to someone back at the hub to say, hey. This vehicle is coming back, and it will arrive with you at four twenty eight. Between four twenty eight and four thirty eight, you need to empty the toilet, and it will assign a task to the person that it knows is on shift back at our hub.
Once that person has got the vehicle, they will empty it. They will mark the task as complete, And we know that it's been emptied, and that vehicle is back out on the road, that toilet is operational. That's kind of how you solve it at a more systems level, and that means that our toilets are always operational.
So that, I think, is kind of an example of one actually, one example of many, many, many thousands of things like this that exist in the world. And I think it's all of these small things that are just pure inefficiency. They're kind of pure sludge in a way.
In our case, literally sludge. The result is this. People have traveled so far a hundred and fifty million zero emission kilometers with Ember. And I'm kind of really proud of this. Even though it's a small number, I'm really proud of it because it didn't exist.
Before we launched, this just did not exist. It still doesn't exist with anyone else. But when you fix real problems, like when you go into this detail of things like toilets there are other examples, by the way. I just picked this one as an example for the tour.
There's lots of other ones. You see the results. So it took us four and a half years to reach a hundred million zero emission kilometers. This year, we're on track to do more than a hundred million. And that is the result of kind of this long term, compounding approach to fixing lots of small problems that add up and that you then start to see the benefit of in the long term.
And that has only come from our vertically integrated approach. So you can't do what we did in an asset light model because you just don't own the assets. You have someone else, you're saying to them, oh, we got a problem with our reviews.
Everyone says the toilets are broken. Go and fix it. And that person doesn't really care about going and fixing it. It's the vertical integration. So on the inside as a company, we're now a hundred and seventy people. I know who was here for Rebecca's talk a minute ago, but we were listening to that thinking, oh my god.
We're kind of definitely a village. But we've also been able to raise a bunch of debt and equity to help the business scale. And then the pitch to investors, as I mentioned earlier, has been difficult. But the investors that got it were quick to grasp this opportunity in atoms plus bits.
And I think they were quick to understand that kind of climate climate as an opportunity or climate as a thesis needs to be out in the real world. You need to be actually doing something in hardware, and you need to be building on the software and bring this together to actually grow a really big valuable business.
So last year, we raised our series a. We're now well capitalized. We're now really focused on growing the market, and the market to us is all of that. And what this little square here is is the entirety of the bus and coach market for the trip types that we're targeting.
Everything else is basically car or train or other form of personal transport. And the kind of key insight here is that our approach is building this kind of product. We think of it as a product adjacency. So we know that you're going from a to b anyway.
We're not trying to convince you to do something else. We can just go two talks ago for the people who are here. We're not trying to convince anyone to change any habits. We're saying, go from a to b, but we're gonna do it for you better, more comfortable, more cheaply, and we're gonna do it with zero emissions.
And so that's why we think that is our true market opportunity. That's where we can become a really, really big business. So it sounds it sounds good. All of this sounds very good. So why doesn't everyone do this? Why are we the only business, basically, that's taken this approach?
And fundamentally, it's really hard. It is genuinely hard to take on all of this complexity. And I think there's a couple of key characteristics that kind of founders who are going on this journey need to have. The first is a kind of first principles approach to everything.
Everything that you're facing in your business, every problem that you're trying to understand, you need to go back to first principles. And that really is due to the fact we're back at the top of that curve. So remember the vertical integrator curve, We're at the top of it. No one has gone before us here.
We're not looking out to the market and saying, oh, yeah. That's interesting. They've done that. Maybe we could copy them. We're kinda trying to figure this stuff out. That's not easy. And I think what it requires in addition to the first principles approach is having what we call deep generalism.
So as a skill set, my cofounder and I, I guess, we refer to ourselves as deep generalists, and I'd actually call that our specialism. So when someone says, what are you really, really good at? I'll say, being a deep generalist. And it kinda sounds like maybe a little bit counterintuitive, but that depth and breadth is what allows us to tackle all of these different problems.
So there's some other reasons, though, that not everyone is doing it. So hardware is hard. Everyone says that. That's the mantra. Assets are scary. If you're investing in real world assets that have ten to twenty year durations, that is scary. Beyond just being expensive, you need to be able to raise lots of debt and structure your business in a certain way to make it work.
But it's scary because you're putting something in the ground that is giving you lock in. If you screwed up that decision, either you bought the wrong stuff or your approach was just wrong or you misread the market or you got your timing wrong, you can't just pivot.
In two years' time, we can't just be like, hey. Actually, let's just go in a different direction. We spend, like, tens of millions basically building all of this stuff. If it doesn't work I guess if it doesn't work, it doesn't work. But it's gonna work, but that can be a scary thing.
You've just gotta have that belief that it is gonna work. And not everyone, I guess, wants to make that decision. The second is that if you are an existing bits or atoms only company, then what you're trying to do if you think of this as a three d surface map, those bits and atoms companies are at different parts of this surface map.
They're at kind of somewhere near a local maxima, and what they're doing incrementally is trying to get a little bit better at what they do. They're trying to climb the mountain or climb the hill that they can see that's just around them. But what they're doing is optimizing on an old paradigm.
That doesn't get you to the top of the global maxima. So we're looking at this. We're looking at that peak, like the actual peak, and we're saying that is where we need to go. That means starting from scratch. And if you're existing, let's say, Atoms company, and you're then seeing this global maxima, even if you can see it, you're not thinking, oh my god.
We gotta go all the way down there. You're not gonna convince shareholders to basically go on this journey with you and fundamentally rebuild your business. Beyond just being hard to do that, it's just not likely to happen. And that is the other reason for a vertical integrator now.
Do not sell your software or what you're building to incumbents. You've got to destroy them. That is the opportunity. You're coming first. You're taking this fundamentally new approach. Don't just give it away. Don't just give them something that is gonna help them get a little bit closer.
It won't even get them the whole way, and that's the other thing. What you can do is just eat their markets. And then finally, you need to really live this view of long termism. Long termism is hard. It's not prevalent just generally in society.
You need the discipline to know what to focus on. And I think that that can be something that applies everywhere. But in the world of software and hardware, when you're thinking about atoms and bits constantly, it's even more difficult to know what to focus on because there is so much to do.
There is, like, you know, every direction you look, there is something that you can do. And the scaling journey as a vertical integrator is slow and it's lumpy. In all honesty, that doesn't fit well with the VC fund cycle. If they're on seven to ten year fund cycles and you're thinking it's gonna take us two decades of compound gains to get to something super valuable, it's not a super good fit.
So, fortunately, I think our investors are bought into our vision, I can kind of say this on stage without getting a phone call afterwards. Because we have the patience and they have the patience to kind of see it through. One key part of this long termism is knowing that you're gonna solve real problems and focusing on real problems over growth.
So it's not about we talk about this a lot internally. It's not about how much we grow this month or this quarter or this year. I don't even care, actually, what happens to our numbers this month. What we're focused on is going back to the root cause.
What is the problem? How do we solve for the products that we're trying to build? And that will get us there in five years or ten years. And this is my favorite example of long termism. So probably a bunch of people know what this is, but it's from the Berkshire Hathaway shareholder letters twenty twenty four edition, and it shows the impact of compounding and of long termism.
What looks like a relatively small difference in annual returns is like mind blowingly different. I I think this is not it doesn't come very easily when you just look at that. If I said, oh, what's the difference between ten and twenty percent per year growth for sixty years?
You probably guess there's a big difference, but it's more than a big difference. It's absolutely gigantic. And that's what we're aiming for. We're aiming to be long termist. We're aiming to use our entire platform and the vertical integration to build all of these compound gains and fundamentally build a better product for passengers.
And so that's us. What we've done today is some of what I've presented. We have a huge, huge distance to go because of this long termism thing. I hope our approach and our ambition is clear. I hope that some of you now believe in this atoms plus bits paradigm in some way at least.
Kinda desperately looking for some validation. No? And if you think this is interesting, then our head office, we're in Edinburgh. We're not far from here. We're on Charlotte Square. We are hiring for a bunch of interesting roles. If you're interested, if you wanna come and speak to us, then look.
We've got some other embers in the room. So if you look for these socks, we're all wearing our ember socks, so you can find us. I look forward to speaking to people. So thanks very much. Thank you, Keith. That was fantastic.
