Sovereign AI is going to be the most important category. The cost of energy, at the end of the day, is not the one thing that matters in the economics of the project. Here's the last hope in Europe for us to build a frontier AI lab. Yep, you guessed it. I have Arthur Mensch, co-founder and CEO of Mistral, on the show today. It is now a $21 billion company. It is the company that is keeping Europe relevant in the race for frontier models. This is Arthur like you've never seen him before. He's unleashed. He's funny, and he says some incredible things. This was such a joy to do, and I'm so excited to welcome Arthur to the show.
It's pretty clear that there is not enough compute. I mean, the inference market behaves very much like a commodity market, which is pretty high margin today.
Ready to go.
Arthur, dude, I am so excited for this. I have interviewed some of our favorite counterparts in the US. You are our remaining hope in Europe. No pressure, but thank you so much for joining me today, dude.
Thank you, Harry, for having me today.
Now, I was doing some research, and I say some quite opinionated things, Arthur, but it seems I'm not the only one, because you the other day said that leaving frontier AI to the US and China will end up costing Europe what remains of its place in the world.
I did say that, and I truly believe that this is such an important technology that it needs to be distributed in the world. We're talking about a technology that is comparable to electricity, a technology that is falling more and more under export control. And so AI is going to be a big enabler of economic growth, and if there's no AI being produced in Europe in particular, Europe will struggle to actually produce the growth it needs to produce.
I'm going direct. To what extent does the success of Mistral depend on US overreach? You know, obviously when we have a Trump administration, we see things like Fable being banned, and it causes a lot of questions and the need for sovereignty. In a world where there's not US overreach, does that threaten sovereign providers?
I'd say this is structural. This is not really linked to any way of operating or of showing power. It's more linked to the nature of the technology itself. If you start as an enterprise or as a state to use AI in all of your core processes, it means that you want to make sure that the money you're spending on those processes actually occurs pretty close to you. If you don't, it means you have a trade balance problem, it means you have enormous flows of cash that are going from one place in the world, can be Europe, to another place in the world, which can be the US. And so there is a macroeconomic problem to not being able to produce artificial intelligence to do R&D on in Europe territory that is not linked to a certain way of saying that Europe is falling behind or a certain way of using soft power, let's say. So it's structural. I won't say that some people are not making it a little bit emotional, but when we started Mistral in 2023, we started with this conviction that this was a sovereign technology, in that as a state, wherever you are, whether in the US, in Europe, or in Asia, you need to own the technology, you need to be able to customize the systems, to run them very close to your data, you need to be able to control the cost, you need to work in true partnerships with companies that understand how to build AI systems, and sovereignty precedes 2025 elections. So it turned out to be true what we said in 2023. It turned out to become quite emotional when we start to speak Greenland. But it's mostly an economic problem. It's not really a political problem. It's just too important a technology.
When I hear you speak, I think of Lyn at Fireworks who came on the show and said actually very similar things. We need companies to have their own models to train them very close to their own data. How do you feel when you hear me say that? And is that the evolution of the company, allowing companies to own their own models and work with their own data?
And what the category we define and the reason why companies work with us is that we bring them sovereign technology. And so in sovereign technology you need three things. You need to have a source of model that is secured. So you need a company that can train the models itself. So we have that model factory. Mr. Larur is actually leading the open weight models in agency coding. It's leading in cyber defense and cyber security. So that aspect of being able to constantly produce improvements on models and to be able to tell our customers that in two years from now you'll get better models because we are constantly training them and spending billions on it is super critical to anything sovereign. So it's the reason why Mistral has been created and it's the main part where we're investing money. Now you also need to have control over your data. And so in general you don't want that data to flow back to the AI providers, and so we've made sure at the technological level that our platform to run the agents, that our platform for customizing models actually sit on our customer tenant. That's super important in many cases when we work with manufacturing companies, they don't want their IP to flow back to you as a processor.
Now that's not enough in general because AI is consuming a lot of energy, and because when you move at scale you actually need a lot of GPUs that as an enterprise you have difficulty to procure. You also need to have token generation services and you need to have token factories basically. You need a combination of data that sits in your infrastructure and tokens that are produced in managed infrastructure at a lower cost. And so what we do as a company is do three things. We train the models. We deploy the platform and the data plane with our customers on their tenant, and then we host what is energy, what is intense in energy. And that's the reason why we've been building our own infrastructure. So we have these three things, and those are the three ingredients that are necessary to serve sovereign workers.
Can I ask if we start on one? You said the model is secured. I have the world's best experts on the show, Arthur. I'm very lucky. But they tell me different things. Okay, some people say, you know, the Chinese open source models, you download them, you run them locally. Hey, they're totally safe. And I'm like, wow, great. And then next week someone says, hey, but we have no idea whether there's prompt injections that you don't know. There's back doors.
Absolutely not. You cannot be sure that they're safe.
Help me out here, Arthur.
Well, at Mistral we have a, I mean one of our business is to really run the inference for open models. So we run ours, but we also run others.
Why, why, why do you run others?
Because customers want choice, and when you have infrastructure you want to provide as much choice as possible to saturate the infrastructure. So that is the reason why that brought us to actually serve more and more of open models, because in different domains you may want to pick a different model. You may want to pick smaller ones, you may want to pick bigger ones. So the more choice you have, the better usage you can make of the infrastructure, the lower your cost can be. And in a way it's a commodity market. So you want the cost to be as low as possible. So that's the reason why we sell those models. But what you say is true. Models are compression of knowledge. If you don't control the data supply chain, if you don't know what goes into the models, you cannot guarantee anything on the output. And so when we serve our models, we can give further guarantees to our customers than when we serve third party models, precisely because theoretically you can make sure that in a certain environment with a certain context the model behaves in a certain way. There's no way of excluding that. There's no way of measuring that. We can monitor that online, that's what we do, but you can't exclude that it's being used as a vector of attack. I think it is very theoretical because I don't think it's the case today. But on the other hand, this is part of the threat model. And so our customers, they are asking us to be able to provide them with all of the open weight models that are available. We're helping them do it in a secured way where we can help them monitor what's going on. But in many cases they are actually very keen to use our models because we're able to tell them that the data supply chain, everything we've trained them on, is something that we have validated before.
What percent of token traffic roughly is your own models versus other open source?
I would say around 80% comes from our own models.
Wow. Okay. That's a lot more than I thought. Okay. On the 20%, everyone tells me that, oh my god, American CIOs and CTOs would never use Chinese open source models, and actually that fear really is there. You have them as customers, you have amazing customers. Are they fearful of using Chinese open source models?
It depends how critical their industry is in general to governments. If you work with defense, for instance, we work with defense customers in Europe, they would never use Chinese technology. If you work on the finance sector, again because it's critical infrastructure, they are very reluctant to do it. If you work in retail, if you work in certain domains of manufacturing, again it depends where. The reluctance is different. Again, I think there is a common understanding that open weight models, if they're being deployed on infrastructure like ours, the only valid threat is that they go rogue and have weird behaviors in certain domains because you may have a supply chain attack on it. There's a common understanding that this is very theoretical and it's certainly not the case today. But indeed, it's part of the conversation. I say what is also part of the conversation is not this, is it a threat to deploy third party models that I don't fully understand. What is part of the conversation is am I going to be able to get updates in 6 months from now? Because what's going on is models have gotten so strong that they're able to do cyber attacks that you can use for pretty much advanced purposes, let's say. And so the consequence is that they are more and more being treated as defense systems. It's being treated de facto by the US as a defense system because at some point in May they started to ban them from outside of the US. And if you look at what China is doing today, they're also looking at ways of restricting exports beyond a certain threshold. The reason for that is that effectively this is a dual use technology. And so as any dual use technology, security of supply matters even more. And so the companies we work with realize that they need security of supply not only on the infrastructure side but also on the model IP side.
Can I ask you with that in mind, do you think your OpenAI and your Anthropic withholding their best models is good or bad for society, and is that just an inherent feature that will be prevalent within this industry?
Is it good or bad for society? I don't know. What is clear is that there is immense leverage to centralized AI providers in a world where there's no providers that is allowing the decentralized deployment, that is allowing sovereign deployment. So what we are doing is extremely important for companies. If we end up in a world where all companies need to rely on two or three providers that are centralized AI providers that give them enormous leverage, they will use the data that is being produced by the systems to make better systems and to go after the business of their customers, because at the end this is the end game. This is pretty clearly what is being priced in today. So the approach we're taking is to make sure that this world in which you have an oligopoly of AI providers that are leveraging all data being produced by enterprises to make better systems that can be intermediate with enterprises is not the world we live in. And we're very aligned in that with all of our customers that of course don't want to live in such a world. So I would say the fact that there are internal models that are being used to do certain research, for instance, is an early sign of what can happen in a world where you have massive concentration of compute and artificial intelligence. Luckily, it's not that complex to build models. It's about deploying capital. It's about having the right team that is training the right thing. And so we've been able to do it. The open weight frontier is catching up with the closed frontier. In many domains it's actually above, and so that means that I don't think we'll end up in a world where if you want to do mathematics you need to work at OpenAI. Luckily, there is enough people in the world that don't want this to occur.
You've said the word energy a couple of times, and one of the core underlying tenets of being successful is having the energy and the infrastructure to power European AI. We have a very different energy system to the US, and we have a very different energy system to China. You know, in China they'll move three million people to put some form of energy infrastructure, and in the UK we won't move a birdhouse. It's quite concerning despite the joke. How do you think about, ruminate on the energy supply challenge?
Well, I think the first thing to realize is that energy is around 10% of the bill in the total cost of ownership of compute. So we're not talking about the place where you can make most of the optimization. So the cost of energy at the end of the day is not the one thing that matters in the economics of a project. What matters is the cost of the chips and what matters is the cost of capital. This is dominating the bill of material. So it's a problem. It will become more and more of a problem as we deploy more compute in Europe, because we'll run into supply constraints. Again, it's not evenly distributed. So if you look at France, for instance, there is excessive energy, you can build capacity. If you look at Nordics, it's also the case. In the UK, it's a little less the case. Germany, it's a little less the case. But there is a realization that AI is going to drive energy consumption forward. I would say in the next two or three years the true fight, the true challenge is to actually make sure that for Mistral, make sure that we build as fast as possible to use the excess energy that is already there. And so that is what we are at work doing: raising capital, building more, building the team to build all of this capacity. There is excess energy. It's not the bottleneck today. It will become the bottleneck in 2029, 2030. But right now the question is how can we build on existing energy. Now we need to think ahead of time, starting in 2029, 2030, as multiple gigawatts of capacity are being deployed every year on AI, we do need to ramp up the production. What we are hopeful for is that in France, for instance, there is an enormous EPR program where nuclear energy is going to be produced much more. You have large efforts on the solar side as well. So I would say it's going to be a problem. It's mostly a coordination problem, because at the end if you have demand coming from AI on energy, the business case for building new nuclear plants is pretty well made.
It totally, it totally is, and I agree with you. I am a big believer in Adam Smith's invisible hand, but the lead time is very real, and so that is just the 2029, 2030 is a challenge.
Yeah, nuclear plants take longer than that to build. Solar plants can go faster. And I do believe in that invisible hand. The states have to play a role in facilitating the buildout, but also in planning ahead of time. They have a longer-term vision than enterprises in general, and so we are spending some time, I'm spending some time explaining that the energy demand is going to grow, and that it needs to be addressed somehow, it needs to be part of the business plan of any energy supplier.
I just want to take it in turn. You said about the cost of chips being important. Can I ask you what do you know now about compute buying, forecasting demand of compute chips that you didn't know?
It's quite a process. So we've learned to run that process to match supply and demand. It's a complex industrial process. Has nothing to do with software. You need to understand the latency in the supply chain. You need to understand that the supply chain comes from multiple, I mean supply chain is fairly complex because you have GPUs, but you also have CPUs, you have storage, you have electrical parts. You need to make sure that your own suppliers have no problem with their own suppliers. So that aspect is something we actually learned, but it's a fairly complex one where you need to manage resources, you need to have multiple forecasts, you need to be able to forecast your demand for multiple buckets. So enterprises, which are our main ICP, but also AI natives that can take larger pieces of GPUs but want them for a shorter period of time. You have companies like Microsoft that are often taking capacity from us that we can use to finance more capacity. So at the end, the problem of building a cloud is about matching supply and demand. It's about managing the supply chain ahead of time, giving visibility to your suppliers, and it's about matching all of this with the finances. You need project financing. You need people that are strong at understanding credit. And so it's actually quite interesting to build such a business. What we've learned is that at the end, if you're able to front capacity ahead of demand, you get something that is pretty easy to sell, and it's going to be the case for the next few years. And so we're very grateful that we anticipated the fact that we would run out of supply fairly quickly, that our customers would have supply chain problems for their own GPUs, because now that we have those GPUs, we can serve their needs.
Totally get you there. That was one, which is cost of chips. Cost of capital was another one. If you were resource unconstrained completely, what would you do today that you're not doing?
We would build much more capacity. I think that's a big bottleneck in building capacity is of course capital, and we see the demand and we know that our customers are running out of compute. As we take some of their use cases and we help them scale it, suddenly they realize they don't need eight GPUs, but they need a thousand GPUs, and that's for one use case. And so we see that demand ahead of our customers, so we have pretty high conviction, that's why we put equity at work to build capacity. And so what we are working on, what we have always been working on, is to make sure we have more capacity under management, even by building ahead of demand. And today, if we just look at what Europe needs, it's pretty clear that there is not enough compute. You may argue that there's maybe a little too much compute being built in the US, but what is clear is that there is a lot of slack to build capacity in Europe in a way where you are 100% confident that it's going to be allocated. So that's what we're working on.
Where you said, hey, there's a role to be played by the state as well. What would you like to see the state, be it France, be it the UK, do that they're not doing today?
I mean, we've been quite vocal about the kind of ways in which they can intervene in a way which is market neutral and that just would facilitate investment. The first thing is of course to facilitate permitting and to make sure that there are dedicated sites where things are accelerated, where all stakeholders, whether regional or state level, are aligned. And I would say France, Sweden are making a great job at that, for instance. So that's mostly on facilitation of aligning stakeholders, because these are physical assets, they are built within regional communities, and it makes it super important to make sure that everybody is benefiting from it. And so the help of states happens there. What they can do at the financial level, and we've said that publicly, is that state guarantees on leasing data centers, for instance, on reserving space for 10 or 15 years are super useful for companies like us that have relatively small balance sheets and that know how to build and who to build it for. So that's the role they can play by giving some guarantees, putting some guarantee on the table to build further and ahead of time. And of course, public, we see public sector is an important customer for us from a lot of different states in Europe and outside. And again, we bring them value, they bring value to the citizens, it creates offtake for companies like us, and that allows us to build further. And if you build further and faster, you build cheaper. In this day of supply constraint,
Can I ask, you said you sell to states there, and kind of going back to that element of sovereignty, how far down does sovereignty need to go? And what I mean by that is like we obviously have Cohere in Canada, does like every country need their own model? Is it blocks that need their, how do we think about like the granularity of sovereignty?
I think the granularity that is allowed is that you need a fairly large market to build relevant models. And today Mistral is able to build relevant models because we operate not at national level. We're not a French company, but we serve Europe, we serve the US, we serve Asia, and what we need is alternatives, but there will not be a single sovereign AI provider and model trainer at every state level. This is not going to happen.
So what percent of your customers are European?
If we look at our revenue base, it's around 60%. And the rest comes from the US and comes from APAC.
And they use you to do post-training harnesses for their own models?
Yeah, European ones, they use us for multiple things, for multiple reasons, but we help them build applications on either things that are related to the efficiency of their businesses or related to accelerating the R&D. We work with manufacturing companies, with financial services, we work with defense companies, and we have pretty strong verticalization approach to all of these three domains, and the fourth being public sector. The way we work with manufacturing companies is in general where we are with their R&D teams and we help them build better products. So we help ASML build better machines, operate better machines. We help Airbus build better planes, and we bring reasoning models, but we also bring the physics models, the models that have a quantitative understanding of the world. In financial services, we build models and harnesses that understand markets. We also bring models that understand quantitative time series to help do prediction. So we come with a set of models that allows to quickly build applications that bring value to our customers. So that's how we serve them. And I guess that back to your point on sovereignty, all of these companies have a sovereign strategy. They need deployment of the technology in a way where they have business continuity guarantees, affordability constraints, etc.
We see OpenAI, Anthropic, DeepSeek move into application layer and also move into chip layer with Jalapeno, with Anthropic's rumored, DeepSeek are doing their own. Will we see a Mistral chip? How do you think about the importance of verticalization and owning every layer?
No, it's not planned. We have gone down the stack, or I would say upstream if you look at it from an industrial perspective. We have gone from training models to serving the models we train on infrastructure that we own. So that's already a pretty significant move. That's a move that took around 18 months to set up and that is now pretty successful because we can serve the GPUs, we can serve tokens. We have tokens that are purely Europe generated in that it
escapes any form of US extraterritorial law. It matters a lot for our customers. So that took us some effort. We are doubling down on this effort to scale that further. And I think that's going to keep us busy for a while. So we are pretty vertically integrated in that we have cloud services infrastructure and we go all the way to building the business applications for our customers.
You mentioned the application layer in AI. You bring the platform and you build custom applications for customers. That's how it works. That is what AI allows you to do. And so this is through an FDE army.
Yeah. It's FDE but building on top of common primitives. So workflow management, agent management, management of sandboxes, observability of the deployment of different business applications. So at the end we are building cloud services that are adapted to run and build with AI agents.
How do you think about doing it through FDEs versus partnerships, which we've traditionally had?
It has been super helpful to have our FDEs front and center at the beginning of our journey, because it allowed us to understand exactly where enterprises were struggling, how they started with individual productivity. It was not enough. Chatbots were not delivering any kind of value creation for them. And so we went and spent some time with the managers, the business unit owners and understood the kind of core processes they needed to automate with artificial intelligence. We did it in a way where we built business applications, custom applications managing state, applications where agents are operating the process end to end, and it was super helpful for us to understand the primitives to build. So we build durable process execution, for instance, we understood how to orchestrate agents in a way that was very effective. So having this FDE sit with customers directly has been super useful.
Now what we have done since last year is to start propagating the tools to some of our partners, like Axonure, like TCS, to system integrators that are great at scaling things, and that once you give them the right tools, once you give them the right training, they can help scale because they have just more people. But having people actually sitting hands on keyboard from Mistral with our customers is super useful for our customers, and it's in high demand. It's super useful for us. It's not something we can scale indefinitely, because we can't have tens of thousands of FDEs working for us. But it's something that now we are scaling through partners and the two motions coexist. So we are in many accounts we have FDEs that are working directly with customers, they're operating the projects, they're making sure that the delivery is occurring, and we work hands-in with system integrators to do the work, to then do the maintenance and these kind of things.
