CharlieKey takeaways from your favorite podcasts
← 28 October 2026Masters of Scale

Eric Schmidt scores US vs. China: Who’s winning the AI race? | Masters of Scale Summit 2026

Eric Schmidt argues the US-China AI race is a two-horse contest defined by the near-zero cost of software, energy limits, and open versus closed models, and that leaders should start using AI now.

Watch on YouTube ↗

Charlie's summary

  1. The cost of software has collapsed to almost nothing, and the US-China contest is a race to superhuman intelligence

    Eric says the biggest change in the past year is that software has gone “from very expensive to almost nothing,” so we will get far more of it. He calls the US-China competition the defining one of our age because both sides are building reasoning systems beyond human ability. He compares it to Waymos, which see better than human drivers. He calls it “the race of a of literally a millennium.”

    ▶ 2:01
  2. Chip export controls bought less time than expected: China is now about six months behind, not ten years

    Eric helped put chip restrictions on China and assumed they would hold for about 10 years. China looked one to two years behind a couple of years ago and now looks about six months behind. China got there by developing algorithms that use roughly one-tenth of the energy, at somewhat lower performance. Eric says it is “a two-horse race.”

    ▶ 3:02
  3. Slowing AI down is unrealistic because it becomes a two-body problem and China already thinks it is behind

    Some argue China is an excuse to avoid regulation. Eric asks how a slowdown would work. Each country has five to seven frontier labs. China's government can order its labs to comply, while the US would need a voluntary agreement. China thinks it is behind and is not stopping, so Eric says a slowdown is not happening. The US must find other ways to make AI safe while staying ahead.

    ▶ 4:35
  4. America's bottleneck is energy and data centers, so build your own power on empty land

    Eric says China has infinite energy, mostly renewable, with no seven-year interconnection queues. The US has the talent, controversy, and deep funding markets but lacks energy and data centers. His answer is to put data centers on empty land and build the power plants yourself. He estimates the US needs 60 to 80 gigawatts through 2030-31. One gigawatt costs about $50 billion and is roughly one nuclear plant.

    ▶ 6:10

Give your email to unlock the rest of the key points — for every episode.

Do more with Charlie

Free — sign in with your email, no password.