Mistral CEO: Why OpenAI & Anthropic Will Come for Your Business | Are Chinese Models Truly Safe?
Mistral co-founder and CEO Arthur Mensch tells Harry Stebbings why he thinks sovereign AI will be the most important category, covering Chinese model safety, energy and compute limits, and how Mistral sells to enterprises.
The gist
-
Mensch says relying on two or three centralized AI providers gives them leverage to use customer data and eventually go after their customers' businesses.
-
Mensch says hidden backdoors in Chinese open-weight models can't be ruled out, but calls the risk theoretical today.
-
He says energy is about 10% of compute cost, behind chips and capital, but expects it to become the bottleneck by 2029 or 2030.
-
Mistral started with forward-deployed engineers inside customers to learn their real problems, and now also scales through system integrators like Accenture and TCS.
-
Mensch credits Mistral's growth to focusing on enterprises in four verticals, manufacturing, finance, defense and public sector, and avoiding the crowded consumer market.
Also mentioned
4:01
Fireworks AI
· "When I hear you speak, I think of Lin at Fireworks AI who came on the show and said actually very similar..."
11:18
Anthropic
· "Do you think your OpenAIs and your Anthropics withholding their best models is good or bad for society?"
17:49
Microsoft
· "You have companies like Microsoft that are offtaking capacity from us that we can use to finance more..."
◎
Related episodes
Episodes covering similar topics
Wall Street Soars, Midterms Shake Up AI Race, Prediction Markets Win | Diet TBPN
geopolitical competition
Modal CEO: Has Open-Source LLM Inference Peaked? | What Happens to the Cost of Compute
open-source models
More from 20VC with Harry Stebbings