Perplexity CEO tells CNBC one metric will determine who wins the AI race
Perplexity CEO just reframed the entire AI race. It's not about model size anymore—it's about value per watt per user.
Why it matters
As compute costs and power constraints become existential bottlenecks for AI labs, Srinivas is articulating a thesis that favors efficiency-first builders over scale-at-all-costs incumbents. This metric matters for how founders and investors evaluate competitive moats.
The key facts
8 to knowPerplexity CEO metric: 'most taken value per watt per user'
Implies efficiency optimization over raw model size
Signals power/cost constraints as primary competitive constraint
Published June 2026 (future date—verify publication timing)
Perplexity CEO Aravind Srinivas on CNBC
Winning metric: 'most taken value per watt per user'
Implies efficiency and unit economics trump raw compute scale
Published June 3, 2026 — forward-looking commentary on AI competitive dynamics
Go to the source
CNBC Technologycnbc.com
Publisher excerpt: Perplexity CEO Aravind Srinivas said that whichever company can provide the "most taken value per watt per user" will be the AI winners in the future.