Since Open AI launched ChatGPT in 2022, pundits were all doom and gloom about Europe’s ability to compete with the US in the AI race. We know the drill: “Europe was too far behind”, “It could never been done”…
This is until few days ago, when:
Mistral AI, a Paris based startup, released its own chatbot Le Chat 🐈,
the French President Emmanuel Macron announced 100 billion euros in AI infrastructure investments 💰, and
the European Commission pledged another 200 billion to compete with Donald Trump’s Stargate 🥊
Couple this with cheap French nuclear energy and Europe is back in the race 🇪🇺🚀
Mistral AI seems to be in the center of this renewed optimism. But what is the story behind the startup and its founders? What are their competitive advantages and values?
Here is a cheat sheet with most you need to know about Mistral AI. For a deep dive you can listen to the podcast:
Mistral AI - all you need to know 🦸🏻♂️
1. Company Overview and Growth 💡
Founded: April 2023, in Paris, France.
Rapid Growth: Mistral AI experienced rapid growth, reaching a valuation of $2 billion in its first year and achieving a $6 billion valuation by mid-2024.
Team: They have a diverse team of over 200 members from 15 different nationalities, with 50% female leaders.
Funding: They have raised over €1 billion in funding from major players like Nvidia, Andreessen Horowitz, and Microsoft.
French Darling: The company has become a source of pride in France & Europe.
2. Technological Approach
Efficient AI 🧠: Mistral's approach is designed to reduce computing costs during pre-training, allowing them to achieve significant results with fewer resources. CEO Arthur Mensch stated, "With 100 times less [computing power than US rivals], we’ve been able to make models that are pretty much on the frontier.” They see this efficiency as a "forcing function for innovation."
Mixture of Experts (MoE) 🦾: Their models utilize a mixture of experts architecture, which uses multiple smaller models to improve performance and reduce computational costs, as described in the Built In article. "When an LLM is faster and smaller to run, it’s also more cost-effective."
Open Source Availability 🔓: Their open-source models are available under the Apache 2.0 license, providing flexibility for users to adapt them for their own purposes.
Multilingual Capabilities 🗣️: Many of their models are natively fluent in English, French, Spanish, German, and Italian, allowing for a more "nuanced understanding" of language and cultural contexts.
Function Calling ☎️: Many of their models are able to integrate with other platforms, adding versatility and accuracy.
3. Market Positioning and Competition
European Alternative 🇪🇺🇺🇸: Mistral is positioned as Europe's most significant player in LLMs, a direct challenger to US giants like OpenAI, Google, and Anthropic.
Competition from China 🇨🇳: They face increasing competition from Chinese companies like DeepSeek. DeepSeek is "a player that is very similar to us," according to Mensch and is "China's Mistral."
Data Privacy and Control 🕵🏻♂️: Mistral emphasizes control, privacy, and neutrality for its corporate customers, contrasting with DeepSeek, which collects more data and adheres to Chinese censorship.
Market Share 🍕: Although performing well, their current market share is significantly smaller than that of competitors, with only 5% of the market for enterprise AI according to a Menlo Ventures study.
Revenue 💰: While growing rapidly, their annualised revenue run rate is in the tens of millions of dollars, a fraction of what the largest competitors are reporting.
4. Mission and Core Principles
Democratization of AI 🏛️: Mistral AI's central mission is to "make frontier AI accessible to everyone." They aim to achieve this by developing open-source, efficient, and innovative AI models, products, and solutions.
Challenging "Big AI” 🥊: The founders, Arthur Mensch, Guillaume Lample, and Timothée Lacroix, all with experience at Google DeepMind and Meta, aimed to challenge the "opaque-box nature of 'big AI’."
5. Strategic Partnerships and Applications
Microsoft Partnership 🤝: Microsoft’s partnership was significant as it was their first such investment in an LLM other than OpenAI.
Defense Applications 🪖: Mistral has secured contracts with the French Ministry of Armed Forces, as well as partnering with Helsing to develop models for defense applications.
Stellantis Partnership: 🚘 They are collaborating with Stellantis to integrate AI in vehicle design, data analysis, sales, and production, including an AI-powered in-car assistant.
Data Centers 💾: Mistral is building a new data center in the Paris region, backed by €300 million from HPC Capital and €3 million from the Île-de-France region, as well as investments from partners. They selected France because of its low-carbon electricity infrastructure.
6. Challenges and Criticisms
Funding Shortfall 💰: There are concerns that Mistral's funding, despite being substantial for a European startup, is still inadequate compared to the war chests of Silicon Valley rivals.
Acquisition Target? 🎯: Despite Mensch's assertion that Mistral is not for sale, there is speculation from some investors that the company will eventually be acquired by a larger tech firm.
Regulation 🇪🇺: Mensch advocates for a light touch to regulations in Europe, which is notable since regulation of AI is becoming a more significant topic.
"God Complex" Criticism ⛪️: Mensch has called the obsession with AI outsmarting humans as a "very religious" fascination, stating "The whole AGI rhetoric is about creating God… I don’t believe in God. I’m a strong atheist. So I don’t believe in AGI."
Share this post