A very warm welcome to Squawk Box Europe. I'm Steve Sedgwick with Ben Boulos and these are headlines. Mistral AI raising 3 billion euros at a valuation topping 21 billion among Europe's largest ever tech equity fundraising rounds. The CEO Arthur Mensch tells CNBC the money will power future growth.
We are very confident that the fundraising we are doing today is also accelerating and enabling further growth down the line in 2027 in particular because this is capital that we'll put to use and that we'll use for creating more compute and more capacity to serve our customers.
Oil prices hit a six-week high with Brent teetering near the $100 a barrel mark. Elevated energy prices weighing on market sentiment and raising the chances of central bank tightening. And global trade in focus, Chinese exports surge accelerating 25% in August and putting Beijing in pole position for another record annual trade surplus. Meanwhile elsewhere, Canada's counter tariffs on the US coming to effect.
Okay, French AI startup Mistral has raised 3 billion euros in a series D funding round putting the company valuation at 21 billion euros. The round was led by Samsung Electronics alongside the EU-backed scale-up Europe fund and US-based PSG Equity. Arjan, who obviously has this great relationship with the company and knows them very well, joins us now. Arash, is this the moment when Europe does something that it hasn't done before, which is managed to get companies that are small to a stage where they are big enough to start competing on a world stage, and go toe-to-toe with the big US and international tech companies? Because, all right, 21 billion, it's a fraction compared to those multi-trillion-dollar companies in the States. And yet, it feels like it's beginning to bridge the gap.
Yeah, I think there's certainly signs of that. And I'd say Mistral is very much the poster child of that effort. Mistral has been set up as this sort of competitor to OpenAI, to Anthropic, to some of the Chinese labs. It's taking a slightly different approach in the sense that it is training its own models, but what it's focusing on is building out infrastructure and trying to work with individual companies to integrate AI into their processes. So, last year, they took on a big funding round with ASML, and in that process, they're working with their ASML to integrate AI into their manufacturing processes. This is the big Dutch equipment maker. Now, Samsung is in board in this round, and they're looking to do something similar. I had a chance to catch up with Arthur Mensch, the CEO, and ask a little bit about what they plan to do with this fresh injection of capital. Let's listen in.
It's all about accelerating. The first thing that we are going to be accelerating is to scale our training compute capacity. We have been building better and better algorithms and better and better techniques, and the team is ready to scale the techniques that we've been building. So, we're excited to train bigger and faster models. It's also about building more infrastructure. We are deploying technology that is based on open-source models, and that allows to have sovereignty for our customers, and that actually takes a very significant physical footprint, in particular in Europe. And finally, we're expanding more and more our go-to-market and our operations with customers so in Asia and in the US as well as in Europe. So part of the proceeds of this fundraising will also be about building further verticalization in manufacturing as with Samsung in particular but also with financial services where we have been growing significantly in the US and defense and the public sector that is again are very concerned about AI sovereignty and for whom we have very good solutions.
I want to talk about one of your investors because fascinating company Samsung coming on board here in a bigger way. Last time you had the big investment round last year led by ASML which led to a big commercial partnership between the two of you working on various different ways to integrate your technology with what ASML was doing on the manufacturing side etc. What work are you planning to do with Samsung already underway with Samsung?
Yeah, so a lot of it is confidential but today announcement is really about the investment itself. But as you can tell from what we've been doing with ASML which has been to embed deeply into their core businesses and to making their machines work better and to designing better things. These things are the way we work with customers and so when working manufacturing companies we embed our teams we work on their deepest and dearest data and we help them build the AI system that can leverage that IP in a way where they feel confident this IP remains their own and the system remains their own. So this is of course something we're looking forward to do with companies like Samsung and the fact that they are now at the cap table sitting alongside ASML and Nvidia as representative of the semi world is I think telling about the focus we have on industry AI and on making AI useful for high-end manufacturing. So, a lot of opportunities in further partnering together there. And that investment is indeed the first step towards advancing technology in the semi space.
Arthur, you're getting a lot of commercial traction now, some big partnerships, big deals coming through. I think you previously said Mistral's annual recurring revenue will we're on track to exceed a billion dollars this year. Do any of this new funding, new partnerships mean that that ARR may maybe higher than you had previously anticipated?
Yeah, I think we are absolutely tracking towards that target and expect to be beating it if everything happens as they are trending.
So, Arthur mentioned talking a little about where that money's going, particularly into infrastructure, which is required for the company can to continue to train its models to compete with some of those big labs out there and particularly out of China because this is interesting where Mistral has basically said that it wants to continue down this path of open weight open source AI models. Mistral and Arthur Mensch believe this is key to competing globally, but it's facing very intense competition from the Chinese companies like ZAI, like Moonshot, like Alibaba, who are also releasing very capable, very strong AI models as well. In fact, Mistral even decided very recently to start selling GLM 5.2, which is a model produced by Chinese firm ZAI, which begs the question, is Mistral giving up on the frontier and can this company compete with these Chinese labs and Chinese models that are hitting the market right now? Does it mean the Chinese models will dominate Europe's AI scene? This is a question I put to Mensch. Let's listen in.
The models that are going to come out of Mistral very soon are very competitive with those. And so it means that as a European company, you can rely on a European provider providing European models. But then the question is rather as a European company can you use models from China? So the first answer is you in general they the Chinese labs are actually not operating and not embedding at all and not working with enterprise customers outside of China. I think that's something to keep in mind. Now in certain cases the assets that are being put by labs in China and the US and Europe can be used and on our infrastructure to serve certain customers that we have. It is not creating a very strong dependency to the AI labs in that at the end the data is staying with us. The customization is made by our engineers. And the sovereignty that our customers are requiring is there. Now the question is can you rely on long-term support for those models? Can you know that they're going to be updated over time? Are you sure that in one year from now they will still be out there and they will still be improved? And the answer to be honest is no because you can't make any kind of assumption on how models are going to be exported. At this point in time we are seeing that the volatility in this space is actually quite extreme. And so what that means for us is that yes we provide everything that is open source and that we can provide because at the end this is beneficial for everyone. Now we also need to be a trusted partner for our customers and they want us to be sure to certify that in one year from now they will get access to better models than they have access today. And the only way we can provide that guarantee is by continuing to train our models ourselves. So there's really two reasons. One is business continuity making sure that our customers are getting long-term support. The other is that we are spinning the open source flywheel. We're putting assets out there and that means that every labs in the world that have an open source strategy wants to make something even better.
And Arthur just on this open source story, I mean one of the big stories in the past couple of weeks has been Nvidia's acquisition of Hugging Face. Hugging Face has been such an important platform for major open source players around the world including yourself. Are you concerned at all about this acquisition that it could erode the neutrality of Hugging Face and you know create more lock-in to Nvidia potentially down the road?
No, we're not particularly concerned. We don't really work with Hugging Face a lot. So we actually that's the place where we host where we put our models assets. So it's a great place for sharing technology. So far it is not linked to an inference platform. We actually don't know how it's going to evolve but overall for us Hugging Face is a place where we share assets. It's a place where every lab is sharing assets. I think it's great that if they continue to operate that platform which is super useful for hosting the weights of models. But at the end of the day our customers are not going through Hugging Face to deploy the technology and so how to comment on that? I think it's yet another proof that Nvidia has a very strong interest in seeing open source models succeed and it's also the reason why we work with them on the Demotron coalition. Again, we have very strongly aligned interest with chief providers to make sure that the foundation for building AI application in the enterprise is an open source foundation. And so that's a very positive news.
That was Arthur Mensch there talking about the potential of open source but I think you know for Mistral which is very much said open source is their future. The big competition is coming from China and that's clear and that was what
I've got two dumb questions. So I'm here for all the people out there who don't understand any of this stuff. I mean. What's the difference between open source and open weight?
So open weights is probably the more accurate term. The two are being sort of interchanged.
By being more accurate. You're being way more accurate. Yeah.
That's a first in technology.
Open weight sounds a bit jargony, but open weights ultimately. And it means that effectively this model can be downloaded and used and run on someone's own infrastructure. That's pretty key. And what Arthur mentioned, Mistral are trying to position as this kind of European open weight champion saying we can be the European sort of sovereign AI serving European and global customers who want secure models that are not US, that are not Chinese, and that are European.
Here's my number two dumb question. If models are open source or open weight, how do you control your margins on them and make as much profitability as closed models?
Yeah, and that's pretty key, I think, because that was hitting to the heart of the question around why wouldn't the Chinese models just dominate then given they are so advanced here in Europe. And his answer was effectively what Mistral is doing is they're producing their own models, which they're then going to companies to work with and say, 'Hey, what do you need? We can build that for you using our models, integrate it into your business.'
Journalism kind of thing, you know, where when I was at the Press Association for instance, we would provide a piece of copy and then people can put their own bells and whistles on it and put in the Yorkshire Post or what have you.
Effectively, but I think Mistral is trying to go even deeper and almost be this sort of you know, create these custom systems based on their models. And his argument was that the Chinese labs aren't necessarily operating in Europe in that capacity. They don't have big offices here and boots on the ground to work with these kind of companies to that. And the other point he was bringing up was it's unclear about the continuity of models. If a Chinese lab releases a model today, are they going to continue to update that in a year's time? And if you have built something based on that model, are you going to be able to upgrade that as well? That was one of the questions.
To what Ben's going to be doing next, talking about Sorry, I It it it it is
You know, before I've said it.
Yeah, yeah, yeah. You're talking about AI security. But the segue is, of course, and rightly or wrongly, don't want to get involved in the politics of it, but people aren't necessarily from a security point of view going to want Chinese AI open weight open-source models in Europe and elsewhere. We've seen all this with Huawei and others over the times, and you've covered all this.
Yeah, we've seen it all, and I think that argument is continuing to resurface now when we're talking about models, and that's precisely where I think Mistral's trying to position itself and saying, 'Well, we are the sort of European sovereign company that can provide all this AI. If you don't want to go to a US lab, if you don't want to use Chinese technology, that's our positioning.' And so, it's an interesting one because their growth is also wrapped up in the geopolitics and in the sort of around AI and technology right now, and that's I think really helping the growth that they're seeing as well.
Okay. Arge, thank you. A lot there, and if people want more on Mistral's funding round and how the startup is betting on open weight AI models to help it compete with players like OpenAI and Anthropic, Arge's write-up is available for you right now at cnbc.com. But don't on your phone, because you want to keep watching the show at the same time as well.
What if you're watching the show on the phone?
You can I minimize the player window, and then look at the app on your phone.
Apple had a foldable screen with enough room on both.
Just I was about to say that. Imagine you had a foldable.
Imagine you had a foldable.
I played with a foldable the other day.
