Is Mistral's monster fundraise a win for European sovereignty?

0:08Hello and welcome to the Sifted podcast. I'm your host Freya Pratty, Sifted's Associate Editor. Joining me today is Daphne Le Prance-Renge. Hello. Hello. Where are you joining us from? I'm joining from Paris. Nice. And I am at Sifted HQ in the London office. So this week, a round that we've been tracking all summer and that the ecosystem has been talking about all summer was finally confirmed. French AI startup Mistral announced its 3 billion euro Series D round, giving the company a 21 billion euro valuation. Daphne, talk us through who joined the round. Sure. So I think the headline news is who led the round.

0:52There'd been a lot of speculation and reporting prior to the official confirmation by Mistral this week. So we finally have the confirmed leads. It's three leads. We have Samsung, the South Korean tech giant. we have um the european union's scale-up fund uh which is managed by eqt the swedish investment firm and then the third co-lead is psg equity which is a us uh based growth equity firm um and then in terms of who joined the round so we know the round was obviously extremely competitive uh oversubscribed it was capped at three billion by mistral and i my understanding thing is that they easily reached that cap.

1:35New investors worthy of note are US-based Advent International, who's also an investor in OpenAI. And then a lot, plenty of existing investors joined the round. ASML, who is the Dutch semiconductor equipment manufacturer who led Mistral's previous round which was almost exactly a year ago that was a 1.7 billion series c so asml returned um to back mistral in this round nvidia the chip um chip designer is also returning to back mistral and then we have plenty of vcs including quite a few us vcs a16z general catalyst headline Lightspeed and obviously some French investors as well, Eurozio, which is a VC firm, public bank BPI France.

2:30So yeah, quite a wide ranging list of investors. Yeah, it's an interesting list. Obviously, for something like the EU scale up fund, the reason they'd want to get into this is quite obvious. They want to back sovereign European solutions. But someone like Samsung, that's kind of the main name in this, out of the leads. What do we know about why they would want to get into it? So Samsung, that investment is really interesting. So one of the things to know about Samsung is that obviously they make a lot of consumer electronics, but they're also one of very few, one of a handful of producers of memory, which is a key component that sits alongside GPUs in order to carry out AI computations.

3:18And just like GPUs, memory components are also in shortage because of high AI demand. So they sit higher up than Mr. Al in the AI value chain in that they produce these components. So obviously, if you pair that with the fact that Mr. Al last year, Secureless Series C that was led by ASML, ASML which makes equipment that makes chips. You can see that Mistral is kind of securing these investments from very strategic players across the AI value chain. Obviously, for Mistral, there is a strategic interest in doing that. But for Samsung there is also in that they are investing into a player that is a high consumer of the devices that they produce themselves.

4:09So several investors have directly compared this move to what NVIDIA has been doing a lot. So obviously NVIDIA has been investing lots of money into OpenAI, into Anthropic, which are obviously going to then be consuming the chips that it sells. The move from Samsung has been compared very like to like to what NVIDIA has been doing. It's obviously a move from Samsung to push a player that will be a consumer of its own technology. And geographically, Samsung, South Korean, that's interesting. Mastral builds itself as Europe's answer to something like OpenAI and Anthropic, though they do have quite a lot of American money in the cap table.

4:49But what have people been saying about the tie between French company and South Korean investor. I think that is probably one of the most interesting aspects of that trio of co-leads. So obviously, as you said, Mistral has always pitched itself as a sovereign alternative to largely US-based providers of AI technology and AI models. This has been the pitch since 2023 and it resonated a lot at the time, but I think it's only become more relevant as geopolitical relations have become more and more destabilized, particularly between the US and Europe. And actually, when you look at the press release that came with the Series D announcement, there is a really strong focus on this idea that Mistral is, you know, this sovereign provider of AI that gives users more control, more choice, more independence from the dominant AI providers to date.

5:56So I think Mistral is really doubling down on that positioning. From that, it becomes quite hard to reconcile with having a lot of backing from US investors, for example. I think that would have if Mistral had gone to deep pocketed investors in the US it would have raised a lot of questions in Europe around whether it is as independent from that kind of US-based ecosystem as it claims to be so I think Mistral needed to find deep pockets outside of the US Europe is difficult because obviously we famously or rather that infamously don't have these deep pockets of capital. The Scalep Fund is obviously co-leading the rounds and that is a new pool of capital that has opened specifically to answer that problem.

6:53But there aren't many investors in Europe that are capable of lining up hundreds of millions of euros or a billion euro to back a company like Mistral. So to me, it seems obvious that they went out there with the idea of diversifying as much as possible their cap table And that definitely shows in the fact that we have a trio of co-leads, one US, one European, one Asian. It also definitely opens up a new market for Mistral. There haven't been specifics around what the exact deal beyond the financial investment from Samsung is between the two companies. But Mistral has said that they will be working together with Samsung to apply their AI to some of Samsung's processes.

7:39So that does open a new geography for Mr. Isle where it can hope to take perhaps more market share than if it was competing directly against U.S. models in the U.S. And I think generally it kind of goes with the times and this idea that in the global ecosystem of AI at the moment where the U.S. has obviously a very strong position, China is increasingly becoming very dominant with a host of open source AI systems. it's a good idea to leverage other geographical zones that might represent huge markets and be a way to kind of counterbalance that duopoly that is slowly emerging. I think generally, I agree, people have been very positive about it.

8:25It's not easy for a company like Mistral to find the capital that they need in a way that aligns with the constraints they've kind of put upon themselves in positioning themselves as this independent sovereign player. So generally what I've been hearing was well played and it aligns with who they say they are and it also makes sense commercially, strategically. There's obviously a question around bringing in a lot of strategic investors which always comes with strings attached. We don't know in this case what the strings attached are But typically there would be contracts or, you know, some kind of commitments to aligning with Samsung's governance or timelines or expectations.

9:13And that obviously is a constraint. And it's not the first time that Mistral raises a lot of money led by a strategic investor. Obviously, it was ASML last year. So one investor that I spoke to was saying, you know, we'll have to see how much that constrains Mistral. and they even went so far as to say potentially this could be even more constraining than some regulations. So, you know, the impact of that could be quite high, but obviously we don't know the details of the deal between Samsung and Mistral, so it will be one to keep an eye on. This is brought to you by HSBC Innovation Banking. You're not ordinary.

9:56No 9 to 5, meal deal lunches, meetings about meetings, compulsory team building living for the weekend you're made for more for the extraordinary and you need an extraordinary bank hsbc innovation banking exists exclusively to help visionary vc-backed tech and life science businesses achieve their ambitions from pre-seed to ipo and beyond we provide custom solutions co-created by actual industry experts to help you scale delivered by a dedicated team you can call whenever you need to. Find out why 4 in 10 UK unicorns have chosen us at hsbcinnovationbanking.co.uk. Interesting. What do you think that feeds into?

10:38Mastral will presumably need more money. Its capital needs will only ever go up. How long do you think it can keep up this fundraising rhythm given it's set out these constraints for itself and wants to be sovereign? Yeah, absolutely. So I think historically Mistral has been compared to Anthropik in terms of the type of company that it is, the kind of needs that it has for capital. Whether or not that comparison is totally accurate is obviously up for debate. But I think if we wanted to kind of roughly compare Mistral to Anthropik right now, we need to go back to 2024, which is when Anthropik was around the same age as Mistral, three years old.

11:21same kind of maturity um anthropic was raising similar amounts of money at a roughly similar valuation so actually the mistral is is not doing too bad in terms of if you compare her equivalent maturity um to to a rival like anthropic after that anthropic hit um the billion dollar revenue figure they raised a series e which was give me one sec so i don't say the wrong numbers having my piece here they raised a series e which was 3.5 billion dollars at a 61.5 billion valuation so mr owl has said that they are on track to hit that same billion dollar of annual revenue by the end of the year um so presumably you can expect that these will be the kind of valuations that mr will be raising at next if they manage to deliver on their promise to hit one billion of revenue.

12:19Obviously, we're starting to go very high valuations, a very fast rhythm of fundraising, because Anthropik since then has only kept raising at even higher valuations. So the question is, you know, where are they going to be seeking money next? They're probably going to be fundraising again soon. Where are they going to turn to next time? Obviously, in Europe, as I said, there's not many investors who are capable of following up with these kind of numbers and at these kind of valuations so yeah there is a question over where next what kind of geographies what kind of investors um I remember when before Mr.

12:59Al confirmed the fundraise and who was leading the fundraise and we were discussing on the podcast actually who might be leading um a few of my sources had suggested sovereign wealth funds like Temasek, for example, or Middle Eastern funds. So this is obviously all completely speculative, but we'll have to see if this is the type of investor that Mistral turns to next, obviously looking for very deep pockets. It will be really intriguing to see who gets in on the next round. The other thing I thought was interesting in the press release. I mean, it's kind of well covered now, but they had foregrounded that they are still a research company, which I know is a criticism that is made of them, that they have become more of a consultancy helping companies work out how to deploy their technology rather than a frontier lab.

13:51And I thought the wording of the release was interesting in making sure that came across loud and clear that this was around to pursue research. Yeah, definitely. I agree that that has been a criticism. But, you know, kind of counter to that, another thing that I found really interesting in the release was they said, and I'm reading this from the release, that during the first wave of generative AI, the central question was who could build the most powerful model. And nowadays, Mr. R claims that the question that organizations are asking is how to harness the power of AI for their mission critical needs without surrendering control over the infrastructure and intelligence loop, which suggests that the focus is not so much anymore on competing on that frontier where Anthropic and OpenAI would be positioning themselves, but rather on building that trust with their users to make sure that it's deployed in a way that's controllable and safe and trustworthy.

14:48And they do, yeah, they describe themselves as this, you know, quote-unquote full-stack AI company, meaning that they obviously build the models and they're still claiming that they are building, you know, frontier level models with a very strong research team. But they're also working hand in hand with customers to deploy the models on their own premises. They're also building compute, obviously. So, you know, this idea that they are giving customers the option to have full control over how they deploy AI and most importantly, full independence from the really duopoly that exists today that is shared between OpenAI and Anthropic.

15:28That idea about working closely with companies, we got some data, it must be a month or so ago now, showing that they had really upped the number of four deployed engineers within the company. So we can put a link to that in the episode description as well, because that really proves your point. You've been working on other Mastral-related news this week. You published a scoop this morning about the company's robotics chief raising funds for his own startup. Tell us about that round while we have you. Yeah, definitely. I mean, it was me, but it was also you, Freya. And our colleague, Tom. So we heard that he is seeking 200 million euros for this new startup, which is promising to build an AI model for robotics that is trained on more comprehensive data.

16:18This is obviously interesting because data is one of the big question marks in robotics. It's unlike large language models which are trained on text that is readily available on the internet. As soon as you are starting to build AI for the physical world, data is much harder to come across. So, you know, whoever kind of solves that question has hit quite a big market. So, yeah, still very stealthy, obviously, still early stage. Things could change, but definitely one we'll be tracking and keeping an eye on. I've already had, so we published that this morning, I've already had people pinging me saying, how can I get an intro to this guy?

17:00Very buzzy. What would it mean for Mistral? He's obviously their robotics chief and they have their own robotics bet, obviously, because they have a robotics chief. But what do you think it could mean for them? Yeah, Mistral's robotics unit is actually fairly recent. So they launched it midway through last year and they've been building their models for robotics or at least intended for edge computing. So that's when the data is processed locally on a device like a robot, for example. it's unclear what the relationship exactly between Duchenne's new venture and Mistral will be, whether or not he will be leveraging technology developed at Mistral.

17:48Some of our sources have described the company as a spin out of Mistral, but at this point, it's not clear if that is technically correct or if Mistral is involved in any way in the new venture. And Mistral themselves gave us a statement, yes, that they were not aware of the company. So things are still being defined, I'd say. Certainly in the case that it was a spin out, so if that was to be confirmed, I think it would show a desire from Mistral to maybe refocus their efforts on their main offering. which you know I don't think is necessarily a bad thing at all it also means it gives more of an opportunity to the technology that is focused on robotics to raise more capital if it's a standalone company this is often what happens with spin outs so giving it a chance to become a proper company rather than kind of be buried within everything else that Mr.

18:55L is currently working on which is a lot So yeah, we'll see how things evolve. What can we expect from you next? We're currently preparing our next AI 100 report where we list 100 companies in AI in Europe that are particularly interesting, hot, buzzy. I'm writing a feature as part of this report, which will focus on physical AI, which in my opinion has kind of been one of the big buzzy topics of the year. so looking into what it is what it can do what the challenges are and most importantly why Europe has an edge or whether Europe has an edge rather and how it can you know try and win in this kind of next phase of AI so yeah welcoming all opinions and analysis please get in touch if you've got interesting thoughts about that.

19:51Where should listeners send that to? To my email, daphne at sifted.eu. Great. And if you want to read that report when it comes out, listeners, you should sign up to our daily newsletter. We'll put links to the report in there, as well as all our coverage of Mastral, The Current Round, and anything further on The Robotics Company as we go along. We'll put a link to current reporting and the newsletter where you can follow what comes next. Thank you, Daphne. Always great to have you on the pod. Thank you for having me. Listeners, if you want to read the report or follow our future reporting on Mistral, you should sign up to our daily newsletter where we'll put all that reporting along with the report Daphne's talking about.

20:37We'll drop a link to that in the episode description along with links to all the existing reporting we've talked about today. As always, please rate, review and share the podcast. This episode was produced by Aditi Mesra.