Mistral raises $640M to take on OpenAI, Anthropic in gen AI race

A year after entering the generative AI race with Europe’s largest-ever seed round of $113 million, Mistral AI is raising more cash — lots more, like, almost 6X that amount.

The Paris-based startup today announced it has secured $640 million series B funding in a combination of debt and equity, taking its valuation to nearly $6 billion.

The investment has been led by General Catalyst, with contributions from several new and existing investors, including Lightspeed, Andreessen Horowitz, Bpifrance as well as the investment arms of IBM, Samsung, Salesforce, Cisco and Nvidia. Notably, the development marks a 23-fold increase in Mistral’s valuation, which was just $240 million in June 2023.

“I am delighted to see new and existing investors renew their confidence in our business and provide new support for its expansion. This new round puts us in a unique position to push the frontier of AI and bring state-of-the-art technology to everyone’s hands. It guarantees the company’s continued independence, which remains fully under the founders’ control,” Arthur Mensch, co-founder and CEO of Mistral, said in a statement provided to VentureBeat.

The company noted that it plans to use the money to build more compute capacity as well as increase the team size across departments to scale commercialization efforts internationally and better take on rivals in the global AI race, including OpenAI and Anthropic, the former of which just inked a deal to be the first third-party LLM provider to come bundled with Apple software in the new iOS 18.

Mistral’s rise as the open AI powerhouse

Since its launch last year, Mistral has made quite an image with a major focus on open-source models and the unorthodox approach of dropping torrent links to download and test them. 

The company first Mistral 7B and Mixtral 8x7B “mixture of experts” models and then followed it up with Mixtral 8x22B and Codestral, a 22B parameter, open-weight generative AI model that specializes in coding tasks, right from generation to completion.

It also launched Mistral Large, a proprietary model that sits just behind GPT-4 and outperforms Claude 2, Gemini Pro and GPT 3.5 with native proficiency across five languages and a context window of 32K tokens.

Notably, one remains primarily accessible via the company’s API and Microsoft Azure, rather than an open release but can be tested via the company’s recently introduced chat app.

Overall, Mistal claims it has seen over 27 million downloads from public repositories, with the open models being used across complex use cases, particularly in the financial industry, the tech scene and the public sector.

As the performance of the open models made waves in Europe and beyond, Mistral accelerated its commercialization effort with major industry partnerships. Over the last six months or so, the company has teamed up with several tech giants and upstarts, including Microsoft, Snowflake, Databricks, Capgemini, IBM and Perplexity. Many of them, including Microsoft, also made investments in the company.

The approach has not only expanded the startup’s reach but also cemented its position as a vendor-trusted player in the AI category dominated by OpenAI, Anthropic and Google. Now, with the latest round, which includes $500 million equity and $140 million debt, the company is expected to accelerate this effort and adopt more distribution channels to accelerate the use of its models in different markets, including the United States.

According to Bloomberg Intelligence, the capital-intensive generative AI market, which was just $40 billion in size in 2022, is exploding and poised to hit $1.3 trillion over the next ten years with a CAGR of 42%.