As someone who sits on the board of two of Europe’s hottest startups, French large-language model (LLM) builder Mistral and Germany’s image generation startup Black Forest Labs, Andreessen Horowitz (a16z) general partner Anjney Midha has a good idea of what’s preoccupying some of the region’s brightest AI entrepreneurs.
Against all odds, it’s not regulation.
“The most urgent bottleneck to progress in Europe, or to the startup ecosystem, is not necessarily policymakers or regulators,” Midha tells Sifted. “I think it is inaction on the part of the CEOs and executives of Europe’s largest companies, who are being slow to adopt the productivity renaissance that’s exploding because of AI.”
Midha, who’s based in San Francisco, has been on the board of Mistral and Black Forest Labs for more than a year. A16z, one of the US’s most high-profile VC firms, backed Mistral’s €385m Series A in 2023 and €600m Series B in 2024, and led Black Forest Labs’s €30m seed round last year.
“Can Europe lead on tech? Absolutely,” says Midha. “These are companies and teams […] that are no less than the best scientists from any other team in any other country, they’ve raised from the world’s best capital sources and now they’re shipping products used by the world’s best enterprises.
“The TBD now is, are they going to actually accelerate adoption in the enterprise?”
European enterprises’ AI lag
Mistral produces open-source generative AI models that it says can compete with OpenAI’s technology, while being safer to deploy for enterprises dealing with sensitive data. Black Forest Labs, whose founding team helped create UK AI image startup Stability AI’s Stable Diffusion Models, has built a technology that is currently powering the image generating feature for Elon Musk’s chatbot Grok.
The challenge for companies like them, says Midha, is that Europe’s largest enterprises have “significantly lagged” behind their US counterparts when it comes to adopting AI.
In the US, CEOs of leading tech companies started experimenting with AI two years ago, he says. It’s a dynamic he experienced first-hand: from 2021 to 2023, he was vice president of product at US social media platform Discord.
Towards the end of 2022, Discord was already experimenting with Open’AI’s GPT models. “The leadership at Discord has the vision to say: ‘We don’t know what this is going to be good for, but let’s give it a shot,’” says Midha.
“And what we saw was on several different internal tasks, like trust and safety and content moderation, it was extraordinary.”
Midha says that the past two years have seen “very aggressive partnering” between large US companies and leading AI startups to experiment with the technology.
But that’s not been duplicated in Europe. “It’s a cultural issue,” says Midha. “If you’re running some of Europe’s largest enterprises, it can be comfortable to keep doing things the way you always have, right? Until it’s too late.”
Targeting mission-critical industries
For a company like Mistral, the imperative is to build a product that enterprises can seamlessly integrate, says Midha.
“They have to abstract away all the complexity of models and instead focus on solving the enterprise’s needs — usually to have faster, cheaper, more reliable, more privacy-preserving tools,” he says.
He adds that the French startup will be targeting deals in “mission-critical industries” like healthcare, finance or defence — where Mistral’s open-source approach, which it says enables better control, privacy and customisation, is likely to resonate.
Mistral’s recent partnership with German defence startup Helsing is an example of what to expect, says Midha. The French startup has also secured a deal with AMIAD, the French government agency dedicated to enhancing the country’s defence capabilities with AI, and is said to be courting officials close to both the British and German militaries.
“I think you’re going to see a host of really interesting announcements, not just from European governments,” says Midha.
“Every major country is asking themselves how to build their AI stack with a partner. I think Mistral is a leading choice for many of those governments, because it can accelerate their local AI ecosystems by taking a great base model and customising it to a culture.”
This week, Mistral released a new model dubbed Saba, which was trained on datasets from across the Middle East and South Asia, and supports Arabic as well as a number of Indian origin languages like Tamil.
Read the orginal article: https://sifted.eu/articles/a16z-anjney-midha-interview/
