00:00A lot to discuss here. One of the big themes at the World Economic Forum this week has been
00:05geoeconomic fragmentation. I think geopolitical concerns are really front of mind for many of
00:11the attendees here. Do you as a European company actually stand to benefit from perhaps enterprises
00:18looking to move away from the U.S. for geopolitical reasons? Well, the technology we've built
00:24brings security, brings controllability, brings portability across multiple hardware platforms
00:30for AI technology. So the enterprises that are working with us, one of the reasons that brings
00:36them to us is that ability to decouple themselves from providers that oftentimes in Europe are U.S.
00:43providers. So I'd say, I don't know if it does, it certainly is a growing topic, top of mind in the
00:49CEO's mind of enterprises that their dependency to single providers when it comes to digital
00:56services might become a problem and is not a great answer to the volatility that we observe in the
01:01world. So it's true in Europe, but it's also true outside of Europe. It's true in the U.S. with
01:06certain accounts. It's true in Canada. It's true in the Southeast Asian region where we operate a lot
01:13as well. So we do see that the technology we've built and the differentiation of it, which is AI that
01:18you can customize and deploy wherever you want, is something that does resonate with the geopolitical
01:23events that we see.
01:24Is that your proposition to enterprises when you are up against some of those U.S. names,
01:31the likes of Google, Anthropic, OpenAI? How do you compete with those U.S. titans?
01:35It drives 25% of our business, I would say, that concerns around owning the technology and being able to fully
01:42deploy it and owning the power on and power off button. The 75% are more driven by our ability to
01:49customize the technology and to go all the way to the outcome. So we have four deployment teams that
01:54actually brings, creates agents, creates business applications based on our coding tools, based on
02:00our models. And because it's cheaper, because it's more customizable, we can modify the model parameters
02:06themselves. It's actually driving growth in areas that are more vertical, requires more
02:13deeper engagement than some of our competitors.
02:16Do you have any defense ambitions given the whole discussions around Greenland right now?
02:20We do serve the defense industry indirectly through defense primes, a lot of them in Europe,
02:27and directly from Ministry of Defense in Europe and in Southeast Asia as well.
02:32Do you think it's an area that's going to continue to grow because of the focus on it?
02:35Well, you know, AI is key in defense. Sovereignty in AI is about economic sovereignty,
02:42but it's also about strategic sovereignty. We unfortunately need to have AI-driven defense
02:48systems because there are AI-driven attack systems used by our adversaries. And we also need to grow
02:54our deterrence in that domain. So AI is critical when you're deploying on autonomous systems like drones.
03:01It's also critical in command and control systems. And so we actually do both with partners and with
03:08ministries. Yeah. Well, staying with the geopolitics, one of my colleagues spoke to the Anthropics CEO
03:14earlier this week. And on China, he thought that the idea of sending advanced ships there,
03:21and this is a quote, is crazy, and said it's a bit like selling nuclear weapons to North Korea.
03:27What's your view?
03:27Well, I don't think this is true. It turns out that China does very well without NVIDIA chips.
03:39We've seen China rise on the open source front, and we have been at Mistral, basically the spearhead
03:46of open source in the West, outside of China. And really, our competitors there, although open source
03:51is not really competing because you basically give things for free, are Chinese companies. And what we
03:56see, and that's probably stressing a little bit some of the CEOs in the US, is that the open source
04:01technology is actually getting to the point where it's as good or even stronger than the
04:06closed source technology, except that it also comes with a lot of much more control, much more customization
04:12and independence from single providers. So that may explain some of these lessons that we're
04:18listening from other CEOs. It is not true that it's going to be a big enabler and that they are able
04:25to source elsewhere than from NVIDIA. Is China behind, though, the West?
04:29China is not behind the West. I think this is a fairytale. In AI, they are very much at parity,
04:36and the year ahead is going to be extremely interesting in that respect. We care about Europe
04:42maintaining its position, Europe maintaining its ability to train models, because we don't think that
04:48we should rely on open source Chinese models in very critical applications. So we need to be
04:54able to create our own models and to sell them to enterprises. So it's going to be very interesting,
05:00but it's a fairytale that China is behind. Let's talk about your own business. Do you have
05:04a revenue goal for this year? Yeah, we do. We should cross a billion by the end of the year.
05:09Okay. And in terms of capex spending, what are you factoring in?
05:12We are about to do around a billion in capex spending this year on our Endeavor Mistral compute,
05:21which is what we've done last year is to start and build our clusters and to fill some of the
05:26data sensors we're renting with these clusters. The idea being that because we have the software
05:31platform, because we know how to build the models and the applications on top, combining them with the
05:36GPUs is a way to get to a vertically integrated stack AI cloud services that are competing with
05:42some of the other fully integrated providers. Yeah. So you're penciling in one trillion euros,
05:48you say one trillion euros of capex spending this year. I think OpenAI have a number of something
05:52like one and a half trillion dollars. I mean, these are massive, massive figures, right?
05:56I said one billion. One billion, sorry. One trillion is a little bit, yeah, it's harder,
06:00I would say, further for the GDP of France. Yeah. But these numbers are really quite big,
06:08right? And aggregate for the industry as a whole. And I just wonder whether you also share some
06:14concerns around there being too much froth and just too much money being thrown into the ecosystem
06:21and what might be actually quite frothy valuations. Well, I mean, we see that there's a lot of value
06:28to be created in enterprises. We see that it takes time. The viscosity of adoption is high,
06:34because it's going to change entirely how the enterprises are working. So we should expect that
06:37it's going to take a few years, because you first need to solve the technology and then you need to
06:41solve the change. And that doesn't happen overnight. Now, the question around, are we investing too much?
06:48Are we investing not enough? I guess that's the question to answer. We've chosen to be a little
06:54more conservative than some, I would say. What we find is that investing billions in training compute,
07:00so in R&D, is a little risky when you see that the open source model family is actually going up and
07:08up. So if you assume that the model layer is free, which it eventually will be, then how do you make
07:13sure that you're spending enough to maintain a good R&D and to innovate, but not too much so that
07:18you're not duplicating efforts against others that are going to give it for free. So that's the
07:23the world we're navigating. I think the opportunity of AI is in trillions. It's in multiple digits of
07:28the global GDP over the next 10 years. But then the question is, how do we invest in a way where we
07:34see profitability down the road and we do see profitability down the road? Okay. Talk to me about
07:39how the partnership with ASML has going. What has it allowed you to do? Well, it has mostly allowed
07:44ASML to do things because this is a commercial partnership where we're equipping their operations
07:50and their research. And so we've done multiple things, but one of the things that we started to
07:55do on the research side is that we're equipping their scanners for full detection, and that's
07:59increasing down the line the throughput of production of their own customers. So we're improving the
08:06machines, and that's the most exciting thing. What is the most exciting for Mistral is how do we
08:11accelerate technological progress and reduce development cycles in heavy manufacturing companies,
08:16in that this is something that is a bit under explored, and that's going to drive immense value
08:20because accelerating progress compounds, and we can expect a lot of improvement down the road this year.
08:24And you also have a partnership with HSBC. Are you looking for similar types of partnerships with
08:29other financial institutions? Absolutely. We have multiple large partnerships with banks, so HSBC is one.
08:36We have a few in the US. We also have BNP Paribas, which is a very big customer. With banks,
08:41we do multiple things. Some are really productivity oriented, and that's what we do with HSBC. We're
08:45equipping their 200,000 employees with individual productivity and automations. Some of them are
08:52more linked to their core business, so on the investment side, on the trading side, how do you maintain,
08:57how do you get signal and process it to make better decisions? Okay. And finally, I've got to ask you about
09:02the business. Any plans to fundraise? Any plans to list an IPO? That's the question you asked me last
09:08year. Well, it's one year later. Exactly. I mean, we raised in September, so we have a pretty large runway
09:16before us. We are an independent company. The path of independence leads at some point to listing, but
09:24that's very, very far away down the road. Okay. Well, I mean, 2026 looks like it's going to be a big year
09:29also for AI acquisitions. Are you on the hunt for any targets? We're definitely on, I wouldn't say on
09:36the hunt, but in the process of doing a few and looking at a few opportunities.
09:42Yes.
Comments