
Eric Schmidt: Singularity's Arrival, the 92-Gigawatt Problem, and Recursive Self-Improvement Timelines | 241
Moonshots with Peter Diamandis
- 43m
- Transcript
Summary
Written by a model from the transcript; not yet reviewed.
Eric Schmidt joins Peter Diamandis to assess where AI stands and what comes next. Schmidt argues we are 10 to 15 percent into AI's impacts, describes the 'San Francisco consensus' that recursive self-improvement will yield superintelligence within two to three years, and identifies a 92-gigawatt power shortage by 2030 as the binding constraint—roughly 60 nuclear plants' worth of capacity. He walks through the economics: one gigawatt maps to about $50 billion in hardware, and data centers now run at 400 megawatts with chips drawing two kilowatts each.
Schmidt also recounts the $600 million DeepMind acquisition that paid for itself through data-center air-conditioning efficiency, tells the story of watching AlphaGo's win probability climb past 50 percent in a Korean hotel room, and warns that China is winning the low-end robotics hardware race. He proposes universities immediately teach prompt engineering to all freshmen, flags the suicide risk to teenagers from LLMs as a line that cannot be crossed, and suggests a modest Chernobyl-scale tragedy may be needed to focus political attention on AI safety.
- The 92-gigawatt US power shortage by 2030 equals roughly 60 nuclear plants, and each gigawatt corresponds to about $50 billion in hardware and data-center buildout. 18:11
- Schmidt proposes universities immediately design a first-semester prompt engineering course for all freshmen, arguing the platform will be the expression medium for every student's work. 11:47
- The 'San Francisco consensus' holds that recursive self-improvement will produce a superintelligence moment within two to three years, driven by agents that need only electricity. 6:05
- China is winning the low-end robotics hardware race, and Schmidt calls the US loss in electric vehicles 'an error' that must not be repeated. 31:02
- The $600 million DeepMind acquisition paid for itself entirely through more efficient data-center air conditioning before it became the AI changing the world. 22:23
Highlights
- 5:01
A Million AI Research Agentsstrategy
Schmidt describes a thought experiment: a company with 1,000 AI researchers turns on AI research agents and is limited only by electricity, potentially running a million of them. The slope of progress goes from linear to exponential.
- 8:32
The Programmer Who Sleeps Through Inventionstory
A young startup programmer writes a spec and evaluation function at 7 p.m., goes to dinner and bed, and wakes at 4 a.m. to find the AI has built what would have taken Schmidt six months and 10 programmers at Google.
- 13:56
The Underage Suicide Probleminsight
Schmidt states flatly that it is not okay for 13-year-olds to be committing suicide because of an LLM and that it needs to be addressed right now, calling it a line that cannot be crossed.
- 18:11
Schmidt testifies to Congress that America faces a 92-gigawatt electricity shortage by 2030, equivalent to about 60 nuclear plants, while the country is building essentially zero or one. He identifies this as the real resource constraint, not capital or talent.
- 22:10
DeepMind's $600 Million Bargainstory
Schmidt recounts how the $600 million DeepMind acquisition was mocked as a waste on a zero-revenue Go-playing team, but the deal paid for itself through data-center air-conditioning efficiency before AlphaGo and AlphaFold changed the world.
- 23:40
Watching AlphaGo Crush the Koreansstory
Schmidt flies to Korea for the AlphaGo match, sits in the Google room watching an RL prediction monitor climb from 50 to 51 to 52 percent, and the architect says they planned for it to go to infinity. The Korean players are crying.
- 35:59
RSI: Scientists Still Disagreeinsight
After spending a week doing recursive self-improvement reviews, Schmidt reports that scientists do not agree on the exact approach yet. Lab demos show limited cases, but the query 'start now, learn everything, tell me what you learned' does not work yet.
- 39:13
A Modest Chernobyl to Wake Usdebate
Schmidt says he is not endorsing a tragedy but describing one: a biological or nuclear attack spawned by AI may be needed to make world leaders sit down and cooperate on safety. He frames it as descriptive, not prescriptive.
About this episode
This episode was filmed at the 2026 Abundance360 Summit. Learn more at a360.com Eric Schmidt ignites Abundance Summit 2026: AI's reasoning boom crushes code (20-80 human split), scaling laws unbound, orbital data centers, China's robot edge, energy/power crunches, recursive self-improvement asymptote, and steering ASI to American values amid geopolitical races. Get access to metatrends 10+ years before anyone else - https://qr.diamandis.com/metatrends Peter H. Diamandis, MD, is the Founder of XPRIZE, Singularity University, ZeroG, and A360 Eric Schmidt is the former CEO of Google; Chair and CEO of Relativity Space. Dave Blundin is the founder & GP of Link Ventures – My companies: Apply to Dave's and my new fund:https://qr.diamandis.com/linkventureslanding Go to Blitzy to book a free demo and start building today: https://qr.diamandis.com/blitzy Your body is incredibly good at hiding disease. Schedule a call with Fountain Life to add healthy decades to your life, and to learn more about their Memberships: www.fountainlife.com/peter _ Connect with Peter: X Instagram Connect with Dave: X LinkedIn Connect with Eric: X Linkedin His latest book Listen to MOONSHOTS: Apple YouTube – *Recorded on March 9th, 2026 *The views expressed by me and all guests are personal opinions and do not constitute Financial, Medical, or Legal advice. Learn more about your ad choices. Visit megaphone.fm/adchoices
Transcript
7,603 words · 3 speakers
Topics
- AI Safety
- Recursive Self-Improvement
- Energy Infrastructure
- China Competition
- Robotics
- Data Centers
- Superintelligence Timelines
- AI Alignment
Page updated · transcript and summary are machine-generated