(Bloomberg) -- France’s Mistral AI launched in early 2023 as a European rival to OpenAI, building generative artificial intelligence tools tailored for the continent. Now Mistral is going after OpenAI’s home turf. 

In May, the Parisian startup hired Marjorie Janiewicz, the former chief revenue officer of Foursquare, as its first US general manager. In an interview, she said the startup is “gaining momentum” with its US business and plans to hire more staff. According to Janiewicz, some traction is coming from businesses looking for alternatives to the AI models and services from large providers like OpenAI and Google. “They do not want to be forced into an AI stack,” she said. “They want choices.”Mistral’s performance in the US will likely be closely watched in Europe, which has historically struggled to produce globally dominant platforms. The startup, set up by former AI researchers at Google DeepMind and Meta Platforms Inc., is the continent’s hottest AI project. It has attracted more than $500 million in investment from venture capitalists, billionaires and a French public bank, and is reportedly in talks to raise another round at a valuation of $6 billion. (Mistral didn’t comment on current fundraising plans.) In its first year, Mistral released a series of AI models, which power generative AI services, and a ChatGPT-like feature called Le Chat. Arthur Mensch, Mistral’s chief executive, has said his small company can compete with OpenAI and Google. 

The startup has also drawn intense political interest in France, a nation angling to be seen as a global AI hub. Cédric O, France’s digital minister under President Emmanuel Macron, is a Mistral co-founder. Last week, Macron said he wanted Mistral to become a “big player” in tech on its own rather than selling to a US giant. 

Even so, Mistral's US expansion is getting an assist from a US giant. In February, it announced that its newest AI model would only be available as a proprietary service on Microsoft Corp.’s Azure cloud. Microsoft invested in the startup, too. In the following months, Mistral unveiled distribution deals with IBM Corp. and Snowflake Inc.

In the US, Janiewicz said Mistral will find customers through these partners as well as via direct deals with companies using its open-source models. Open source is one of Mistral’s main selling points. Its executives have framed the approach — where the underlying computer code is openly shared and tweakable — as a safer, more versatile method than closed systems offered by OpenAI and others.  

Before her stint at Foursquare, Janiewicz, who was born in France, spent several years in leadership roles at MySQL AB and MongoDB Inc., companies that provide open-source database software. She described a growing interest in the US to run Mistral’s open-source models on-premises, with data stored in local servers instead of with public cloud providers. Janiewicz cited growing demand from companies in financial services, tech and healthcare. “They’ve been waiting for an opportunity to really leverage a player that’s portable and flexible,” she said. 

Mistral is far from the only open-source AI firm. Meta Platforms Inc. has poured resources into its model Llama, which claims 170 million downloads so far. Falcon, a competing open-source model released from the United Arab Emirates, said it had more than 43 million downloads as of the first quarter. Janiewicz said Mistral’s open-source models have been downloaded several million times to date.

Mistral is also going against giants like Google and Microsoft and well-funded startups, like Anthropic, that are quickly expanding into enterprise sales. The opportunity is huge — Bloomberg Intelligence estimated that the software and cloud market in generative AI market could reach $350 billion by 2032. Still, many enterprises are only trialing these paid AI tools.

Mistral's edge could come from customers that want to avoid being locked into particular cloud companies, according to Bloomberg Intelligence analyst Mandeep Singh. He said the startup's models perform well on available benchmarks and compared its potential business model to Red Hat Inc., the open-source provider IBM acquired for $34 billion in 2019. "Their positioning seems to be good," Singh said of Mistral, adding that Meta’s rival models might struggle to gain traction because of trust issues with the company.

Janiewicz declined to share Mistral sales figures. She said it was “early days” on determining the company’s pricing strategy for its AI offerings in the US. OpenAI was reporting about $80 million a month in revenue last year, Bloomberg News reported in August.

Still, the Parisian startup is increasingly recognized on the other side of the Atlantic, according to Janiewicz. “I’ve been pleasantly surprised with the brand awareness already,” she said. “Customers are even saying ‘Le Chat’ and ‘la platform.’ It seems that the French touch is having an effect.”

©2024 Bloomberg L.P.