Eric Schmidt on AI, the Battle with China, and the Future of America

Summary

• Schmidt’s revised China thesis is that chip controls and shallower capital markets are steering Beijing away from “crazy AGI strategies” and toward deploying AI everywhere. Chinese teams are targeting consumer apps and robotics, including an attempt to repeat their electric-vehicle playbook. While America pursues frontier intelligence, “we better also be competing with the Chinese in day-to-day stuff.”
• Open models have become a geopolitical distribution contest that Schmidt fears China could win. China is proliferating open weights and open training data like a “Belt and Road Initiative,” while the U.S. is largely focused on closed weights and data. DeepSeek R1’s reinforcement-learning work and use of 8- or even 4-bit precision make lower-cost deployment—and an American 03 model that Schmidt said could fit on a phone—strategic priorities.
• The AI revolution is “underhyped,” but Schmidt separates powerful agents from machines capable of choosing their own goals. He expects domain-specific superintelligent savants in perhaps six or seven years, versus the “three-ish years” forecast in San Francisco, but says there is “no evidence” yet of systems setting their own objective functions. Until then, AI remains “middle to middle,” with humans controlling the end-to-end workflow.
• Drone warfare flips defense economics when a $4,000-$5,000 aircraft carrying a 2-kilogram payload can destroy a $30 million tank. Schmidt first argues that reinforcement-learning battle plans invisible to the opposing side could deter attacks; in a later exchange, he rejects the host’s mutually assured destruction framing as “not a deterrence.” The host stresses that infrastructure-scale drone war has no winners, while Schmidt says humans would eventually still have to cross a line.
• Relativity Space is an execution-heavy launch bet built around physics that Schmidt found surprisingly immature. Its rocket produces 4 million pounds of thrust, while roughly 80% of launch mass is propellant, 18% the rocket and only 2% payload. The low-Earth-orbit order book is full; “we just have to launch the rocket.”
• America’s advantage rests on capital markets, universities, entrepreneurial disorder and immigration—but demographics threaten the demand base. Schmidt cited China at about “1.00 for 2 parents” and Korea at “78 for 2,” arguing humanity is “collectively choosing to depopulate”; aging then means fewer customers and structurally declining revenue. His broader premise is that America and the West should win by investing faster in people, businesses and infrastructure: “Don’t screw it up, guys.”

Deep dive

1. China is choosing deployment over the most expensive AGI race

• Schmidt’s changed view: he had considered China a peer-level AGI competitor, with U.S. chip restrictions merely slowing it. Instead, hardware limits and shallower capital markets make it difficult to raise roughly $100 million “on a wing and a prayer” for data centers, pushing China toward practical deployment.

• In Shanghai, he saw robotics companies attempting to reproduce China’s electric-vehicle playbook: strong funding without American-style valuations, extraordinary work intensity and a focus on applying AI across consumer products and physical machines.

• The geopolitical danger is distribution. Schmidt framed Chinese open weights and open training data as a new “Belt and Road Initiative”: much of the world could adopt Chinese rather than American models, and he would prefer the underlying learning to reflect Western values.

• Responding to the host’s claim that Meta had misexecuted its open-source effort, Schmidt said it was not clear the company was going fully closed. He also thought DeepSeek R1 had complicated Meta’s position through strong forward-and-back reinforcement learning using lower numerical precision—8-bit and even 4-bit versus the 16-bit training typical in America.

• Schmidt said Sam Alman had indicated that the smallest version of the 03 model would be released with open weights and told him it had been. He described it as much smaller than “10 to the 26” and said it could fit—or can fit—on a phone, offering another route to American model proliferation.

2. Agents arrive before machines can choose their own goals

• Schmidt opened expansively: agents will become powerful, work together and increasingly run computers autonomously. For the next few years, however, he expects assistants acting on human commands and prompting—not independent intelligence with its own volition.

• The “San Francisco narrative” predicts recursively self-improving systems and domain savants in chemistry, physics or mathematics within “three-ish years.” Schmidt accepts the direction but puts those savants perhaps six or seven years away; when the host noted that a savant is not general intelligence, he agreed.

• His AGI boundary is the ability to set one’s own objective function, for which “there’s no evidence” today. The technical obstacle is changing objectives: humans continually revise goals, while current systems optimize against a target. His test is whether a model, given only information available in 1902, could derive Einstein’s special relativity followed by general relativity through cross-domain analogy. “If we can solve that problem, then I think it’s over.”

• Schmidt’s own optimistic framing is that humans are “end to end,” while prompted, validated and iterated AI is “middle to middle.” Humans define the objective, prompt and validate; AI performs much of the middle work. He has seen no recursive self-improvement yet. Startups he funded claim proximity, but their optimism leaves him with a rough estimate of “5–10 years.”

3. Cheap autonomous systems overturn the cost of war

• Ukraine showed Schmidt how a country with no navy or air force could use automation while outnumbered roughly three to one. Its resistance has been innovative, but Russia’s strong second and third rounds reinforced his warning: “The enemy gets a vote.”

• The decisive metric is the kill ratio. A retail drone costing $4,000-$5,000 can carry a 2-kilogram payload capable of destroying a $30 million American tank, making much traditional armor economically untenable. More broadly, mobile systems replace much fixed military infrastructure.

• Drones are now more cost-effective than mortars, grenades or artillery; both sides then field drones, creating drone-versus-anti-drone competition centered on detection and destruction. The doctrine becomes forward drones with people behind them. Schmidt has seen Ukrainians command operations from Kev over Starlink at a distance. The host also cited Ukrainian unmanned surface vessels attacking Russia’s Black Sea fleet, helping enable grain exports from Odessa that he described as roughly 6% or 10% of Ukraine’s economy.

• Schmidt’s hypothetical endpoint is two sides with perhaps one million drones each, using reinforcement-learning battle plans invisible to the opponent. He first said that uncertainty would make deterrence against attacking each other very high. The host instead described infrastructure-scale drone war as lose-lose and asked about mutually assured destruction; Schmidt replied that this was “not a deterrence,” defining deterrence as a penalty greater than the value of an attack. He still expects destructive waves followed eventually by humans having to cross a line.

4. Relativity’s full order book still depends on one hard launch

• Schmidt confirmed he was Relativity’s first or an early investor and said he had also been invested in SpaceX, Swarm and Starlink. He picked Relativity because “rockets are really cool, and they’re really hard,” having assumed rocketry was as mature as jet engines when it is still both an art and a science.

• Relativity’s rocket produces 4 million pounds of thrust—enough that ordinary metal restraints cannot hold it during testing. Roughly 80% of rocket mass is propellant, 18% the vehicle and just 2% payload, a ratio physicists told him reflects six decades of confronting Earth’s gravity.

• Relativity is positioned as a low-Earth-orbit competitor with a full order book. Schmidt’s summary was simple: “We just have to launch the rocket.”

5. America’s advantage compounds only if talent and demand endure

• Schmidt says he favors work-life balance but opposes essentially working from home, especially for people in their twenties, who learn by hearing experienced colleagues argue in person. Against China’s nominally illegal but widely practiced 9 a.m.-to-9 p.m., six-day “996,” his message was blunt: winning in technology requires tradeoffs.

• His broader geopolitical premise is direct: “I want America to win.” He connects that view to the American dream that gave him opportunity and to earlier generations’ fight to keep liberalism and democracy alive. America’s advantages, he says, are chaotic ingenuity, intelligent capital allocation, deep financial markets, universities, entrepreneurs and an industrial base envied in Europe and Asia. “We should celebrate this. We should stoke it. We should make it go faster and faster.”

• Demography is the internal constraint. Schmidt cited China at about “1.00 for 2 parents” and Korea at “78 for 2,” arguing that humans are “collectively choosing to depopulate.” Shrinking and aging populations mean declining customers and revenue that businesses cannot overcome through innovation alone.

• He broadly favors immigration because it helps America address the problem, while saying the global mechanism requires a broader response. His prescription is continued investment in the right people, businesses and infrastructure.