Google AI infra chief Amin Vahdat: goodput per watt is the north star at $200B capex scale
samiramanabi · x · 2026-10-08
Key takeaways from an interview with Amin Vahdat, Google's SVP of AI Infrastructure overseeing DeepMind, Cloud and the TPU roadmap:
- Google alone forecasts $200B capex in 2026E, part of the biggest infrastructure build-out in history
- Goodput per watt is the north star metric: at 100,000-accelerator scale, something fails multiple times an hour, so systems must be designed around failures
- Capacity must double every six months, with chip planning done 2-5 years out
- Constraints span chips, data centers, and software-hardware co-design
Vahdat is among the most central decision-makers in AI infrastructure; the interview is valuable for understanding the engineering and economics of the compute arms race.
More from Companies & People
- Supabase CTO Ant Wilson to Speak at Osaka Meetup on Oct 18 — dshukertjr · 2026-10-08
- Meta's free 10-day virtual AI hackathon offers $150 in model API credits and a $1M prize pool — Arindam_1729 · 2026-10-08
- Open-source AI competition shifts from models to full stacks at SF Tech Week — FinanceYF5 · 2026-10-08
- Paul Graham: Amazon Banning Agents Is Its First Real Startup Opportunity Since Founding — jdjohnson · 2026-10-08
- Math group urges boycott of OpenAI over release of 700+ math research files — Polymarket · 2026-10-08
- PyCon Taiwan Keynote: Automated Discovery of Novel Optimizations and Algorithms — blelbach · 2026-10-08