Moonshots EP 207: Eric Schmidt on AI’s Biggest Threats — China’s Strategy, Cyberwar and Biotech Risks

Summary

Part 2 of Eric Schmidt’s appearance on Moonshots, interviewed by Dave Blundin. Schmidt assesses the US-China AI race, arguing the US will likely win the AGI race due to hardware export controls, but China will dominate robotics and clean energy hardware — mirroring the EV playbook. He identifies the US energy deficit (92 GW needed by 2030, near-zero new nuclear starts) as America’s critical vulnerability, noting some AI training may migrate to Middle Eastern data centers. Schmidt outlines three imminent AI threat vectors: misinformation, cyber attacks, and bioengineered pathogens — with bio being the hardest to contain. He warns that open-source model distillation (at ~1% of training cost) creates proliferation risks analogous to nuclear weapons but far more compressible. For founders, he advises building around learning loops and scalable platforms, noting the barrier to starting a company is now effectively zero.

Key Segments

• [00:01-03:00] US vs China AI race — US leads AGI path, China adopts AI broadly but differently
• [03:00-07:00] China’s robotics/EV dominance, US energy crisis (92 GW gap), electricity as strategic bottleneck
• [07:00-09:00] Middle East data center deals as fallback for US AI training
• [10:00-17:00] Three AI threats (misinfo, cyber, bio); distillation and proliferation risks; open-source vs closed-source geopolitics
• [18:00-25:00] Founder advice — zero barriers to entry, learning loops, platform lock-in as the path to scale

Notable Claims

• US needs 92 GW of new electricity by 2030; effectively zero new nuclear plants being started
• China installed 172 GW of solar in one year
• Distillation/transfer learning achieves ~same results at ~1% of original training cost
• Largest US models may never be released publicly; largest Chinese models will be open-source
• Barrier to starting a company is “effectively zero” now

Guests

• Eric Schmidt — Former CEO of Google, Chair/CEO of Relativity Space
• Dave Blundin — Co-host, Moonshots regular

RDCO Mapping

• Sanity Check angle: The energy-as-bottleneck framing is strong newsletter material — the 92 GW stat vs zero nuclear starts is a concrete hook
• Vault cross-ref: Connects to AI infrastructure costs, US-China dynamics, and proliferation threads
• Data point: Distillation at 1% cost is a useful benchmark for explaining why model moats are eroding