This week, French AI Mistral raised Europe’s largest-ever tech equity round at a €21bn-plus valuation. In an exclusive interview, early investor Paul Murphy told Pathfounders, the bigger story is not chasing OpenAI or Anthropic on frontier benchmarks, but what enterprises actually need, and acting as a “western counterpart” to China’s booming openweight models.
The round was led by Samsung Electronics, alongside EQT’s Scaleup Europe Fund and PSG Equity, and takes Mistral’s total funding since 2023 to around €6bn.
Paul Murphy, formerly of Lightspeed and the first investor to join Mistral’s board, says the company has long been misunderstood as simply Europe’s answer to OpenAI.
Murphy met Mistral co-founder Arthur Mensch in 2023 and says that Lightspeed decided within hours to back the team. “We knew from their backgrounds that they were a special team,” he said. “We did the whole thing in two days.”
The more important comparison today, Murphy argues, may not be OpenAI or Anthropic, but China’s rapidly growing ecosystem of open-weight models. Mistral is “uniquely positioned to be the kind of Western counterpart” to that movement, he said.
That also explains why Mistral is not obsessed with topping every frontier-model benchmark.
For large companies, Murphy argues, raw model performance is only part of the equation. CIOs and CTOs care about “hallucinations, latency [and] token pricing,” but above all whether AI solves a real business problem. “You actually don’t need to be at the frontier” for many enterprise applications, he said.
Mistral will continue developing frontier models, but it is increasingly building the stack around them, such as compute, data centres, deployment infrastructure and engineering services.
This strategy is designed to meet the demand amongst governments and companies for AI sovereignty. Some governments and large enterprises want to run models in private clouds, on-premise or under tighter control over where their data sits.
Murphy says Mistral has already had an advantage here because its models are relatively efficient to deploy. By contrast, some open-weight Chinese models can require huge amounts of memory and be complicated to run.
The company is targeting 1GW of European compute capacity by 2030. Murphy believes financing will not necessarily be the constraint: debt markets can fund much of the infrastructure, while the harder bottlenecks could be securing energy, finding suitable sites, and building them quickly enough.
Samsung’s arrival is also significant beyond the money.
Murphy sees the investment as evidence that Mistral is breaking out of the “European champion” box. “This is much more than a European story,” he said. “This is clearly Samsung saying Mistral is going to be an incredibly strategic company globally for AI.”
Mistral is also increasingly agnostic about whose models customers use. It has begun hosting third-party open-weight models, including Chinese systems.
Murphy says that is consistent with the company’s original philosophy. In other words, if another model works better for a particular task, there is no reason not to use it. He expects pre-trained models themselves to become increasingly commoditised, with more value shifting to fine-tuning, reasoning, and applications built on top.
Rather than trying to outspend OpenAI and Anthropic in a permanent race for the biggest and best frontier model, Mistral is betting that enterprises and governments will pay for control over their models, infrastructure and data. And, obviously, for an alternative to being dependent on a handful of US AI providers.
Murphy argues that Mistral’s position is deliberate, rather than evidence that it cannot compete at the frontier.
“If they had wanted to raise another €10bn” to chase the bleeding edge, he said, “the capabilities [were] there. It’s just that that’s not the business strategy.”
Our exclusive interview is available now on YouTube and will be shortly on Spotify and Apple iTunes.