Citadel CEO Ken Griffin Says AI ‘Almost Certainly’ in a Bubble; Trump Has 'Mixed Scorecard'

Summary

• Ken Griffin sees “almost certainly bubbles” in AI, but investors still need to identify who can retain economic rents. If GPU supply exceeds demand, “marginal pricing will actually go into freefall,” like an airline selling a spare seat for $20. He also questions year-seven-to-nine chip economics when extraordinary pricing on 5–6-year-old GPUs probably will not persist another 5–6 years.
• AI vendors’ spending plans and customers’ cost discipline are pulling in opposite directions. Dani Burger cites Anthropic’s reported $750 billion S-1 spending plan; Griffin says customers increasingly distinguish expensive frontier-model tasks from work cheaper open-source models can handle. The unresolved bull case is that models become indispensable—for example, unlocking cancer cures—so buyers may pay “any price.”
• AI is weakening incumbents’ moats, but Griffin still credits people with most of Citadel’s alpha. Work that took highly trained staff 4–6 weeks can now finish in hours, with moats “filled in by bulldozers” and reskilling still lacking action. His depression has shifted toward optimism about startups, while Citadel’s model remains human-led research: “trading is how we monetize our research.”
• Citadel’s Situational Awareness trade was a liquidity-and-risk-transfer opportunity, not Griffin’s template for explaining every selloff. Hedge-fund deleveraging belongs at “answer number ten”; first work out what other investors know. Citadel bought the portfolio in less than 24 hours and spent days reducing key risks, betting that removing a forced-sale overhang would restore confidence; Griffin says another such episode soon would be “a bit surprising.”
• Trump has a “mixed scorecard,” and Griffin says Republicans will “almost certainly” lose one or both chambers in the midterms. He credits border closure but criticizes allied tariffs as inflationary, treatment of working undocumented families as “disgraceful,” and the explanation of the Iran war as inadequate. If voters deliver a defeat, he wants Trump to listen rather than double down.
• Productivity is Griffin’s fiscal escape route, but he is already worried about the bond market’s warning. He says pandemic spending exhausted the room for another tax-cut renewal, links a $40 trillion federal deficit to crowding-out pressure, and prioritizes AI, energy technology, education, entitlement reform and skilled immigration. Washington is “not yet” listening; without change, he predicts higher inflation, a weaker dollar and lower living standards.
• Griffin’s Miami thesis is a decades-long technology-cluster build, not an overnight migration of finance jobs. He cites his $3 billion Carnegie Mellon gift and Pittsburgh’s 100-company robotics cluster while forecasting that a Miami campus with thousands of students will transform the city. Citadel remains anchored across London, Miami and New York; he says the New York office has 2,500 employees and wants teams together where possible: “nothing beats a whiteboard and a good debate.”

Deep dive

1. AI erodes moats, but people still assemble investment theses

• Ken Griffin’s shock was seeing AI deliver in hours “the exact same answer” that PhDs and master’s-trained staff needed 4–6 weeks to produce. He saw threatened white-collar jobs, competitive moats “filled in by bulldozers,” and a “golden opportunity” for entrepreneurs. Dani Burger asks if depression has become optimism; he agrees, without dismissing the transition’s human costs.
• His prescription separates displaced workers, who need reskilling where he sees “no action,” from incumbents, which must innovate and cut costs faster, benefiting consumers. Startups offer another channel: hiring hardworking people with imperfect pedigrees gives them a new career lane.
• For Citadel, research is “almost like journalism”: get the business model, TAM, economics and management incentives right, then value future cash flows against alternatives. AI helps with pieces, but a person still assembles the narrative. He credits people with the “vast majority” of alpha; automated trading monetizes that research and supports bigger positions.

2. AI’s investment risk is rent capture, not just adoption

• Asked about semiconductors’ longest losing streak since the financial crisis, Griffin sees investors finally “doing research.” Buying every beneficiary of hundreds of billions or $1 trillion in data-center spending is giving way to questions about who earns economic rents and whether scarcity margins survive two or three years.
• His GPU analogy is an airline’s $20 incremental seat: low marginal cost makes oversupply brutal. If GPU provisioning exceeds demand, “marginal pricing will actually go into freefall.” Extraordinary pricing on 5–6-year-old GPUs “probably doesn’t persist for another 5 or 6 years,” challenging revenue and capitalization assumptions for chip years seven, eight and nine.
• Burger juxtaposes Anthropic’s reported $750 billion S-1 spending plan with customers scrutinizing bills. Griffin agrees there is tension: corporate returns on AI are uneven, and sophisticated companies are routing some work to cheaper open-source models rather than paying OpenAI or Anthropic for frontier capability. The unknown is whether models unlock cancer cures and become tools pharmaceutical companies will pay “any price” to use.
• There are “almost certainly bubbles,” he says, but transformation and speculation coexist. His dot-com lesson is not blanket rejection: “I just wish I’d kept my Amazon stock from that moment in time.”

3. Deleveraging is the tenth explanation, not the first

• Deleveraging should be “answer number ten,” Griffin argues, because the default explanation “washes your hands of accountability.” First ask what other investors understand that you do not. “The minute I start to use an excuse, I stop learning.”
• Situational Awareness illustrates the genuine mechanism: investors anticipating forced sales can turn a possibility into reality. A portfolio buyer generally gets a sharp reversal when the liquidation overhang disappears. Citadel completed the purchase in less than 24 hours, then spent several days reducing key risks before paring individual positions. Griffin says the solution was meant to help the seller raise cash and protect substantial year-to-date gains; Citadel “did okay.”
• Griffin says Situational Awareness was caught in “three mid-sized storms”: the CME announcement, chatter about foundation-model growth rates, and large pension plans reshaping portfolios. He finds a near-term repeat “a bit surprising.” That is not invulnerability: recalling Citadel in 2008, he says Morgan Stanley had failed to achieve funding with the Japanese and “I don’t think I’d be on stage today.”

4. Trump faces a midterm check on a mixed scorecard

• With midterms two weeks away and Burger citing Trump polling at 33%, Griffin says Republicans will “almost certainly” lose one or both chambers: the House looks particularly difficult, the Senate close. He hopes any beating prompts Trump to listen rather than double down, and notes his past support for Democrats including Rahm Emanuel and Obama.
• The “mixed scorecard” credits border closure but calls treatment of working undocumented families “disgraceful”: he wants a route to remain, not necessarily citizenship, partly to grow the economy and fund entitlements. He calls allied tariffs inflationary and destructive to supply chains. On Iran, he faults Trump’s explanation of an unpopular war while emphasizing the importance of preventing an Iranian nuclear weapon.
• Griffin is “not so worried” about Democratic socialism’s geographically limited rise, but points to France—government spending at 60% of GDP—as a warning. Burger asks whether resistance to austerity and readiness to protest make France an isolated case. He recalls a van smashing Chicago’s Dior storefront so 50 looters could empty it in “about 300 seconds”: “That dream is a nightmare. I left Chicago. That dream doesn’t belong in our country.” He names J.B. Pritzker as heading down that path.
• His Illinois counterexample is an offer to publicly support higher taxes if Pritzker pursued a constitutional fix for unsustainable pension promises; Citadel represented almost 2% of state income-tax revenue. Pritzker rejected it, Griffin says, then lost his “Fair Tax Amendment.” Griffin’s complaint is that the governor attacked him rather than accept the voters’ spending message—the response he hopes Trump avoids.

5. Productivity is the fiscal exit, but bonds are already warning

• Burger points out Republican spending increases too. Griffin defends the productivity logic of the 2018 tax cuts but says pandemic spending, comparable to World War II as a share of GDP, left no fiscal room for the later renewal. Citing a $40 trillion federal deficit, he sees crowding-out pressure in mortgages and 30-year yields, alongside inflation.
• “Grow the damn economy” is his first priority: support AI and American data centers, but also fusion, fission, batteries and robotics for the next 20 years. Rather than tariff battles, he wants allies pursuing growth reforms too: “making the pie bigger for all of us.”
• Education and labor supply complete his program: he contrasts 53 Illinois schools with no child at grade level against Success Academy seniors identifying a painting’s artist, year, country and social context. He wants a Simpson–Bowles-style entitlement vote honoring existing promises while discussing 1–2 extra working years or delaying Social Security 2–3 years. For elite colleges’ often 15–20% foreign students: “Here’s your degree. Here’s your green card.”
• Burger asks whether growth makes bond warnings less alarming. “Oh, no, I am worried,” Griffin replies; politicians are “not yet” listening. If they do not, he predicts higher inflation, a weaker dollar and a falling standard of living.

6. Citadel stays transatlantic while Miami builds for the long term

• Burger challenges Miami’s finance-professional growth—perhaps 20–50 people—with investment-professional headcount actually shrinking. Griffin says cities evolve over “decades,” citing business conditions and the Cuban community’s importance to Miami’s ethos while conceding housing and traffic constraints. Citadel’s “center of gravity” sits over the Atlantic between London, Miami and New York. Griffin says he has 2,500 employees in New York and does not want to be involved in New York politics.
• Griffin says his $3 billion Carnegie Mellon gift backs robotics and reindustrialization—not factories making Nikes and T-shirts. Pittsburgh’s 100 robotics companies around CMU supply his model: a Miami campus with thousands of undergraduate and graduate students that he says will transform the city.
• Business verticals stay geographically concentrated “as much as possible.” “Nothing beats a whiteboard and a good debate,” Griffin says; writable walls and physical proximity support collaboration while immersing younger employees among bosses and peers who act as mentors.