Mistral Funding Round - Our Breakdown

What today’s raise means for Europe’s tech sovereignty

Hey all,

Mistral has just raised fresh funding at a €11.7 billion valuation, cementing its status as the leading “European” LLM company (though still slightly trailing Germany’s defense AI start-up Helsing with €12 billion valuation from this summer).

Why does this matter?

Mistral, positioned as Europe’s flagship LLM contender, is reaching a critical fork in the road: Exit, new funding, or something in between, all under the shadow of Europe’s push for ‘technological sovereignty.’ The choices from here will shape not only the company but Europe’s place in the global AI market.

Last week, we wrote up predictions on how this would play out (and why!). To keep ourselves accountable, we publish our analysis here without further edits.

We remain fairly confident in our analysis.

What we got right: a new funding round and no exit yet, which increases consolidation and cooperation in the European industrial space

What we were slightly surprised by: contrary to our expectations, U.S. VCs deepened their involvement (even though seemingly only slightly), while UAE’s MGX sat this round out.

Leevi + Frederike

Mistral’s Fork in the Road

Mistral is approaching a stage in its lifecycle where an exit strategy or a new round of funding becomes critical. ‘This moment is infused with particular geopolitical sensitivities. Mistral has been widely hailed as the one potential European AI champion to protect the elusive ‘technological sovereignty’ of the European Union in the intensifying AI race narrative. This is a fork in the road in terms of the strategical choices downstream for the company, with ramifications for the EU AI industrial aspirations. While a new funding round or a rumored IPO would align with French state interests, who also have a stake in Mistral through the national development bank BPIFrance, a lucrative acquisition offer from a major tech firm such as Apple might prove difficult to resist, especially given that Mistral’s board is partially funded by influential US venture capital (VC) backers, maybe eager to exit their positions at the peak of the AI cycle.

We suspect that some tensions between these two strategies are emerging amongst its backers reflecting underlying conflict between the blitzscaling venture model and the industrial champion-model, that’s more focused on deep integration with high-value customers in specific verticals.

Mistral’s challenges: struggling to find a customer base

Mistral is facing some challenges. While precise data is hard to find, the available evidence suggests that the "DeepSeek-moment" from January proved challenging to Mistral, with it gradually starting to fall behind on benchmarks vis-a-vis American and, crucially, Chinese open-source counterparts.

Such benchmarks bear only fleeting resemblance to the actual capacities of the models and typical customers will hardly notice the differences at the top. In the heavily VC-subsidized race-to-the-bottom pricing environment of LLMs, however, such rankings have outsized importance.

As a result of the pricing environment and the structural advantages of the incumbents, we suspect that Mistral is struggling to grow its customer base. According to Openrouter data that measures the usage of LLMs through APIs, the market share of Mistral is around 2%, down from the peak of 10% last year. Mistral’s usage in absolute terms has tripled over the same time, yet the company has struggled to keep up with the race for market share.

Consumer subscriptions may not be the answer

Customer subscriptions aren’t a saving grace for Mistral either. OpenAI’s ChatGPT is the only subscription-based software that makes significant revenues from subscriptions in this market. The self-reporting on the profitability and revenue numbers could be skewed by a creative use of annualized revenue figures with limited visibility on the sustainability of its pricing (i.e., how much money OpenAI makes per subscription). Open AI’s subscriptions are likely heavily unprofitable. Regardless, fueled by investors’ willingness to keep burning cash to sustain market share, Open AI has secured a strong foothold in the consumer market that makes it hard for competitors like Mistral to compete.

https://www.similarweb.com/top-websites/

Enterprise AI may not be the answer either

As conquering the API and consumer subscription market are an uphill struggle due to the structural advantages of the incumbents, the strategy of Mistral has been to embed itself into the distribution networks of existing industrial companies in the European Union, as we predicted last autumn. The extent to which this strategy has been profitable is not public. Recent reporting speaks of the importance of several key large contracts where ‘a small number of large customers’ ‘each worth at least $100mn over three to five year contracts constitute the majority of business. We think much of the contracts are forged with the industrial giants in the EU AI Champions Initiative. The extent to which this model can grow to a venture-scale returns is debatable. The amount of unattached enterprises willing and able to enter into such contracts is limited. Likely, in order to gain traction Mistral will have pressure to keep prices affordable, with limited headroom to raise prices. Moreover, using Forward Deployed Engineers to work on site with the customers is likely expensive. Recent innovations, such as contextual memory are likely strategies to gain traction and retention amongst the users.

The ,, Moreover, Mistrale's original differentiators i.e. compute efficiency and open source are now facing headwinds. The improvements in efficiency translate to competitive advantage only imperfectly, as competitors are willing tosubsidize the costs of large and inefficient models so that consumers do not face the full costs in pursuit of market share. Moreover, the open source models from Chinese and US companies entangled with geopolitically motivated AI model diffusion to global markets are becoming more prominent in recent months

Second, the input prices are going up on both key cost centers of an AI company. The longer ‘reasoning’ models based on chain-of-thought inference are pushing up the costs of running the model in a way that is not keeping up with the decrease of the computing price. The inflation of the salaries of leading AI engineers, we suspect, is also creating pressures towards the employees. As outrageous salary offers in the tune of 200 million dollars and 32 billion pre-product, pre-idea acquihire offers are thrown around in the industry, especially by Meta, the capacity for smaller players like Mistral to keep hold of their key staff might be limited. As colleagues at Rue Ménars start buying condos and fancy EVs in the nicer parts of Paris, the very human tendency for comparison and envy might start sparking restlessness.

Mistral’s conundrum: an Apple acquisition would provide an exit for investors

And it is here we come back to Mistral’s conundrum. If the argument above is correct, the hyperscaling venture capital returns through dominating the consumer-facing markets seems unlikely. This decreases the appetite for power-law driven traditional VC to provide more capital for the company. The projected valuation of 10 to 14 billion dollars is high, but pales in comparison to the 183 billion dollar valuation announced by Anthropic last week. Hence, the rumored Apple acquisition (or rather, an acquihire) offer would provide an exit for those investors who are not satisfied with the growth rates of the company. Due to the size of Mistral, not many others would be able to come up with the required cash.

Our verdict: a new funding round

Overall, our bets are on a new funding round. The French state is not likely to allow an acquisition of the European AI darling by an American company, and the domestic investment screening law provides a legal basis to block the acquisition by force. In addition, whether Apple truly wants to buy Mistral is an open question, with most of the anticipatory reporting referring to influential, but minority voices inside the company. Strategically, the core justification would be the hiring of the talented (and relatively affordable) engineering talent from the Parisian tech hub. More cynically,leaking of such prospects might have been strategically savvy for some actors to gain leverage in the middle of funding negotiations

Instead, we expect the funding round to be more of a strategic nature reflecting the increasing third-party reactions to the emerging bipolar fight between China and the US. State development bank BPIFrance is likely to increase its stake. The UAE sovereign tech investment fund MGX, currently investing in Mistral’s datacenter, is rumoured to take a stake in the round pointing to new geopolitical alliances spurred by the needs of the current wave of tech development. Lastly, domestic industrial giants such as ASML are rumored to take a leading 1.3 billion € stake in the 1.7 billion € round, providing patient anchor capital that shields Mistral from the voracious pressures of the venture funding model.

The outcome in this case would be increasing consolidation and cooperation in the European industrial space, gearing up to meet the moment of the putative AI race.

Addition:

Contrary to our suggestions, American VCs remained invested in the company, slightly increasing their stake. However, we do note that the UAE’s sovereign fund MGX did not participate in the end. We are still figuring out whether they remain in talks in providing funding for Mistral’s compute infrastructure through a separate financing instrument (e.g. SPV á la Meta.). Also, we did not mention that longtime backer of Mistral, king of all kings in the current AI cycle, Nvidia, still maintains their role in supporting EU’s AI aspirations.