New Delhi: For years, Europe watched its best AI researchers migrate to the US. India has faced the same challenge – a steady bleeding of top engineering and computer science talent to Silicon Valley. Reversing this flow is central to the global AI race, says Arthur Mensch, co-founder and CEO of French startup Mistral AI. “Like Europe, India was bleeding talent to the US,” Mensch said in an interview with TOI. “The more talent you retain and create value locally, the better.” Founded in 2023 by Mensch, Guillaume Lample and Timothee Lacroix, Mistral AI was created to challenge the opaque, closed nature of “big AI” and make frontier models more accessible through open architectures. In just three years, the company has scaled rapidly – reaching a $400 million revenue run rate and targeting over $1 billion in revenue soon. As of last year, it was valued at $14 billion. But for Mensch, the AI race is not just commercial. It is structural – and geopolitical. With Europe heavily reliant on overseas digital infrastructure, particularly US hyperscalers, Mistral has chosen to build its own capabilities. “We don’t position ourselves as a sovereign alternative. We position ourselves as a global competitor in the AI race,” he said. “But tactically, to have capacity you fully control, you need infrastructure – the servers that run the technology.“ About 60% of Mistral’s business comes from Europe and 40% from the rest of the world. Some customers opt for its platform precisely because it can be deployed on their own infrastructure, reducing dependence on hyperscalers. Sovereign AI, he believes, is both a strategic and political necessity. As AI begins to run large parts of the global economy, govts and defence systems cannot risk external control. For businesses, over-dependence weakens negotiating leverage and continuity. India, with one of the world’s largest developer pools, has a similar inflection point. “We’ve brought many European researchers back to Europe. India has a unique opportunity to do the same. Universities here produce excellent AI and computer science talent. The focus should be on ensuring they innovate and build value here,” Mensch said. Mistral is currently building commercial partnerships in India, with a local technology centre a possible next step. While rivals such as OpenAI and Anthropic explore IPO paths, Mensch said a listing is “down the road.” Profitability and global scale would be prerequisites. Mensch argues that much of enterprise AI adoption has stumbled because companies treated GenAI as a collection of tools rather than a platform shift. Early chatbot deployments focused on small productivity gains. “That doesn’t change the bottomline,” he said. Instead, Mistral focuses on high-ROI use cases that address major sources of business friction. At the core of its strategy is open source. “If you have access to model parameters, you can deploy wherever you want – including local infrastructure,” Mensch said. Open models also allow customisation. “Running a business should not mean over-dependence on a single service provider. I’ve said it multiple times – Europe cannot become an AI colony of the US.”
