France’s Mistral AI Enters With $113 Million Seed Round At $260 Million Valuation To Compete With OpenAI

Image credit: John Seton Callahan/Getty Images

AI is well on its way to racing in earnest. He raised $113 million in seed funding to help the startup, just four weeks old, compete with OpenAI in building, training and applying language models and generative AI at scale.

Paris-based Mistral AI, co-founded by alumni of Google’s DeepMind and Meta, focuses on open source solutions, targets enterprises, and has CEO Arthur Mensch as one of the largest companies in the field. I plan to create what I believe to be a challenge for I can use it. “The company plans to release its first model of text-based generative AI in 2024.

Lightspeed Venture Partners led the round, along with France’s Xavier Niel, JCDecaux Holding, Rodolphe Saadé and Motier Ventures, Germany’s La Famiglia and Headline, Italy’s Exor Ventures, Belgium’s Sofina, UK’s First Minute Capital and LocalGlobe. I am joining. .

Mistral AI notes that French investment bank Bpifrance and former Google CEO Eric Schmidt are also shareholders. A source close to the company confirms that the €105 million ($113 million at current rate) funding equates Mistral AI to €240 million ($260 million). . Notably, this is the same number rumored in the French press about a month ago when people started chatting about the company.

Mensch (middle) and co-founders Timothy Lacroix (CTO – left) and Guillaume Lampre (Chief Scientific Officer – right), all in their early 30s, have known each other since school, and they all know each other. were studying all over the world. field of artificial intelligence.

Mensch worked at DeepMind in Paris, while Lacroix and Lample worked at Meta’s Paris AI hub. And Mensch said it wasn’t until last year that they started talking about AI development directions.

“Last year we saw technology really starting to accelerate,” he said in an interview today. This probably refers to the leap that his OpenAI has made with his GPT model, which was supposed to come as a shock to many people. The world of AI and the broader technology.

But even though OpenAI’s name includes the word “open,” it felt like something completely different. Mensch, Lacroix and Rumple felt that their own approach was almost becoming the norm and saw an opportunity to do things differently. “Open source is a core part of our DNA,” said Mensch.

It’s very premature to say what Mistral is doing or plans to do, it’s only been a month or so, but Mensch says the plan is to Building models using only publicly available data to avoid legal problems faced by some. training data, he said. Users can also provide their own datasets. The model and dataset will also be open sourced.

And while some believe open source has created a messy situation (and a minefield) in areas such as application security, “we believe that the benefits of using open source overcome the potential for abuse.” There are,” he added. “Open source can prove tactical in security, and we believe it will be the same here.”

It’s also too early to know how well future products will resonate with the market. But what’s interesting is that this startup is specifically focused on enterprise customers rather than consumers, and what those customers need to do and how they can do it. The idea is that there is a gap in the market to help understand

“At the moment, we have evidence that AI can help in some cases,” Mensch said. “However, there are still too many workers who are required to be creative in various fields. [with AI], and we need to figure this out for them. We want to give them easy-to-use tools to create their own products. “

Funding such a young company with no customers, let alone a product, may seem like a giant leap. Especially given that some high-profile lawsuits have already lost a penny. Neeva, another of his startups with Google pedigree, has abandoned its consumer AI search business and sold some of its technology to his Snowflake. Recently, Stability AI is also in the limelight in a good way.

But Antoine Moileux, who led the investment in Lightspeed (and, to point out, Lightspeed also backed Stability AI) believes it’s a leap worth taking. said.

“For us at Lightspped, we weren’t working in isolation,” he said in an interview. “We have had the opportunity to see several strong technical teams looking to raise funds to build LLMs across India, the US and Europe. I didn’t think about the secondary effects of LLM, even though it was technically strong.” In other words, I didn’t think about where and how to apply it.

He compares the AI ​​landscape not to applications but to larger infrastructure businesses such as cloud computing and database businesses. “In the cloud he’s going to have five or six players getting a big share, so I think there will be value here as well.

“So we figured out who we can trust the most here. Then we met with the founders of Mistral, a very talented team. There are probably only 70 to 100 people in the world who have it.” For example, Lample led the development of LLaMA, a large-scale language model for Meta, and co-founded Mistral AI before retiring. led the LLM team at Meta.

On a different note, the fact that AI startups are getting VCs to open their checkbooks is worth noting. There’s no doubt that Google, Apple, Metas, Amazon, Microsoft, and others around the world will be active in this space, but the fact that startups are being founded and finding motivated talent means it’s not all over yet. is shown.

Mistral is one of the names for the north wind that blows down over Europe, and when it does, it’s usually a sign of good weather in the days to come. Startups may be hoping for the same effect here.

People in San Francisco often refer to the “uncanny valley”, but ultimately it will take credible efforts from around the world to build and take some control over how AI will develop. Mensch said he believes there is a need for This is a Mistral play to give France a chance.

The company’s plan is to use the money to build what he calls a “world-class team” and create “the best open source model,” and it definitely needs some money to do that. will be, he said.