Nathan Lambert: Average OpenAI researcher has ~5,000 H100 equivalents; Chinese AI talent is equal
pstAsiatech · x · 2026-10-15
a16z highlights several key takes from Nathan Lambert's interview:
- Compute gap: An average OpenAI researcher can access 5,000 H100-equivalent GPUs at any time — a figure that, per Lambert, makes Chinese researchers' "jaws fall to the floor."
- Talent parity: Chinese labs are technically excellent and train huge numbers of high-level people; claiming China trails on talent is, he says, hard to defend.
- OpenAI's superpower: Frontier labs win by making the most ambitious bets earliest, which is why OpenAI has achieved what it has.
Overall framing: near-parity on talent, an order-of-magnitude gap in compute access.
More from AGI Musings
- AI safety researcher defends rigor of METR and Redwood work, explains why labs tout funding — austinc3301 · 2026-10-15
- AI Safety Incentives Debate: Nonprofit Worker Pushes Back on 'Inflating Problems for Funding' Critique — austinc3301 · 2026-10-15
- Schools should teach change, not skills: math community's identity crisis over AI — soumitrashukla9 · 2026-10-15
- OpenAI chief scientist: no lab has solved alignment well enough to keep max-speed scaling — austinc3301 · 2026-10-15
- AI interviews under fire: economist says rigor signals match quality in high-skill hiring — alz_zyd_ · 2026-10-15
- People with inner monologues often assume everyone cognizes the same way — repligate · 2026-10-15