Ken Griffin on US-China Tensions and AI

Summary

In this conversation, Ken Griffin frames AI as both a productivity engine and a new strategic bottleneck: Citadel has used machine learning for years, but newer agentic systems can now reproduce and verify finance research in hours rather than weeks. He says that productivity gains have not reduced headcount at Citadel; instead, they let the firm pursue more opportunities, while also eroding corporate moats and potentially ushering in a “golden age” of entrepreneurship. Griffin also emphasizes that compute is increasingly scarce and expensive, with large market-making firms spending hundreds of millions on it. On geopolitics, he argues China is an unusually strong competitor in advanced technologies and that a Taiwan disruption could cause an immediate, severe shock to the U.S. economy through semiconductor supply-chain collapse. His policy prescriptions center on education, nuclear power, natural gas, and building data centers and related generation in the United States.

Notes

• Citadel returned to a five-day in-office model early, which Griffin says was meant to preserve apprenticeship, mentoring, and human-capital development.
• He argues remote work has damaged young workers' employment prospects more than AI has.
• He says humans learn through apprenticeship, and mentors are critical to development.
• Griffin links Citadel's early office return to both performance and talent development.
• He says Citadel moved quickly to get the FDA to approve experimental COVID drug trials in the U.S.
• He says Operation Warp Speed worked by changing incentives and pushing manufacturing before efficacy was known.
• Griffin says the vaccine program saved about half a million lives.
• He says the U.S. spent a few billion dollars and saved a few trillion dollars in GDP through Warp Speed.
• On energy markets, Griffin says China's oil-demand destruction has been larger than expected.
• He says closing the straits could have pushed oil to nearly $200 a barrel.
• He says Citadel has used machine learning for about a decade, dating back to TensorFlow.
• He distinguishes AI from the broader stack of machine learning, optimization, digitization, and technology.
• He says newer transformer-style models have extended Citadel Securities' pricing and risk-management capabilities.
• An agentic system at Citadel can recreate finance papers, verify the results, and run out-of-sample checks.
• He says this is master's- or PhD-level work, not just routine white-collar automation.
• Griffin says the AI breakthrough did not reduce headcount at Citadel.
• He argues AI is rapidly filling in competitive moats.
• He says the result could be a golden age of entrepreneurship.
• He says Citadel Securities has used TensorFlow and later ML models for 10 years.
• He says market makers are competing on compute and the variable supply is basically fully utilized.
• He says compute prices have risen beyond what people projected two or three years ago.
• He says large market-making firms are spending hundreds of millions of dollars on compute.
• Griffin says China is among the most innovative and fastest-growing economies in the world.
• He claims China leads in roughly 67 or 68 of 74 important technologies.
• He says China's scale and STEM pipeline are core advantages.
• He says the U.S. should respond by educating youth and competing on innovation, not tariffs.
• He says a Taiwan conflict could cut U.S. GDP by 8% in six months.
• He says the shock would freeze production across aircraft, autos, and consumer electronics.
• He says the policy answer is to re-embrace nuclear and small modular reactors.
• He argues solar manufacturing often relies on Western China and coal-intensive production.
• He says natural gas is a major U.S. advantage and has helped cut emissions.
• He says data centers should be built in the U.S. with corresponding power generation and grid ties.
• He says portfolio management should focus on worst-case loss and whether it is tolerable.
• He says the hedge fund industry's equilibrium cost of capital is roughly risk-free rate plus 4%.
• Citadel has returned $25 billion to $30 billion to LPs to keep ROE high and avoid being overcapitalized.