The 9-month-old AI startup challenging silicon Valley’s giants

The 9-Month-Old AI Startup Challenging Silicon Valley’s Giants

Last year, Arthur Mensch, at 30, worked at a Google unit in Paris when artificial intelligence (AI) was gaining attention beyond science fiction. Today, his startup, Mistral AI, is just nine months old and valued at over $2 billion. The rapid progress mirrors the excitement and concerns surrounding advanced AI systems.

Mensch, along with his engineering-school friends, challenges the belief that U.S. tech giants will dominate the AI race. Despite being smaller, Mistral AI aims to outmaneuver Silicon Valley by focusing on efficiency and open-source solutions.

“I’ve always regretted that there was no Big Tech in Europe. I think this is our chance to become one,” says Mensch, now 31.

Mistral AI, named after a strong wind in France, questions the massive spending by industry giants on AI systems. Mensch, with an academic background, has dedicated his career to making AI and machine-learning systems more efficient.

The startup recently unveiled Mistral Large, a new AI model comparable to GPT-4 and Gemini Ultra, but costing less than $22 million to train. This cost efficiency has attracted interest from corporate clients and investors, including Microsoft, which is adding Mistral’s model to its Azure cloud service.

Mistral's approach involves giving away many AI systems as open-source software, aiming to be the most capital-efficient company in the AI world. Despite its smaller size, Mistral AI collaborates with companies like Salesforce and Nvidia, demonstrating its potential to challenge industry norms.

Mensch, a blend of academic and entrepreneur, has always sought efficiency. From scaling up software for analyzing brain images during his doctorate to working on large language models at Google, he aims to avoid the bureaucracy of big tech and build transparent technology.

Mistral AI's journey involves lobbying French policymakers against regulations that could hinder innovation. While they released initial AI systems as open source, their most advanced models, including Mistral Large, remain proprietary, striking a balance between a business model and open-source values.

As Mensch puts it, “We want to invent new things, new architectures, and we still want to have something to sell extra to our customers.” Mistral AI's unique approach and cost-effective models challenge the status quo, proving that innovation can come from unexpected places.