Mistral exploring in-house chip design as infrastructure costs mount

Mistral CEO Arthur Mensch told CNBC the AI startup is exploring the design of its own chips to manage surging infrastructure and compute costs. The move mirrors OpenAI and Anthropic's hardware strategies as foundational model costs accelerate.

Custom silicon has become a lever for LLM labs to control economics and reduce reliance on NVIDIA—though designing chips to production takes 18–24 months, signaling Mistral's infrastructure commitment.