Ken Griffin on Bloomberg: AI Bubbles, GPU Gluts and a Bond Market Warning

Citadel founder Ken Griffin tells Bloomberg’s Dani Burger that the AI trade is turning discerning, that a GPU glut could send compute pricing into freefall, and that the bond market is warning Washington it is not yet listening to. The conversation took place at Citadel Securities’ annual Future of Global Markets conference in New York, two weeks before the 2026 midterms, and covers AI and white-collar work, the summer purchase of the Situational Awareness portfolio, Griffin’s scorecard on President Trump, his feud with Illinois Governor J.B. Pritzker, and a growth-first plan for the federal deficit. Watch the full interview on Bloomberg Television’s YouTube channel.

TLDW

Griffin says AI now does in a few hours what used to take Citadel’s PhDs four to six weeks, which threatens white-collar jobs and corporate moats but should unleash a golden age for startups. Investors who bought every AI name are now asking how long chip margins last, what happens if GPU supply outruns demand, and how long a GPU stays useful. On politics, he expects Republicans to lose one or both houses, calls the tariff battles with allies a disaster and the treatment of long-settled immigrants disgraceful, and says pandemic spending left no room to renew the tax cuts. His fix for the fiscal hole is growth: AI, energy, robotics, better schools, entitlement reform and green cards for foreign graduates.

Thoughts

Growth is Griffin’s whole answer to America’s fiscal problem, and he means it literally. Asked whether growth is the way out of the deficits, he says, “That’s the only direction,” and when Burger asks whether Washington hears the bond market’s warning, he says, “Not yet.” If it does not, he expects higher inflation, a weaker dollar and a falling standard of living. The plan he lays out (more data centers, fission and fusion, batteries, robotics, fixed schools, a Simpson-Bowles style vote on entitlements, and a green card handed out with every elite degree) is coherent, and it sits awkwardly with the party he funds. Two of his three biggest complaints about the administration, the tariff wars with allies and the treatment of working immigrants, are complaints about policies that shrink exactly the growth he says is the only exit. He is effectively telling Republicans that their economic program is working against their own fiscal math.

The sharpest single point in the interview is his explanation of why the 2018 tax cut playbook stopped working. He thinks the original cuts were “spot on,” but says renewing them a year and a half ago ignored one huge difference: the pandemic, during which the US spent roughly as much as a share of GDP as it did to win World War II. That left no fiscal space, and he ties it directly to why mortgages and the 30-year yield are so high. Coming from one of the largest Republican donors in the country, this is a meaningful concession. It also reframes the deficit debate away from a simple tax versus spending fight and toward a sequencing argument: the same policy can be right in one fiscal position and wrong in another.

His account of the Situational Awareness trade is a clean lesson in how markets turn fear into fact. Once fast-money traders sense a fund is in trouble, they start trimming the same positions or front-running the sales they expect, and that “transform[s] a possibility into a reality.” Citadel’s edge was being able to take the whole portfolio in under 24 hours, then hedge the key risks cheaply before selling individual positions as confidence returned and prices snapped back. Griffin is careful to frame it as a solution for the fund’s investors, who he says are still up for the year. The more revealing line is his reminder that Citadel itself was nearly finished in 2008: every firm that survives long enough has a footnote about the moment its back was against the wall.

The airline analogy for GPUs is the most useful framing of AI risk he offers. Airlines have near-zero marginal cost for an empty seat, so the moment there is overcapacity, pricing collapses; the 2000 telecom fiber build-out did the same thing. If GPU supply ever exceeds demand, Griffin says, marginal pricing goes into freefall and “upends all the investment assumptions” behind today’s spending. Pair that with his second worry, that nobody should assume a GPU earning strong pricing in year five or six will still do so in year nine, and with corporate buyers routing more work to cheap open-source models, and you have a coherent bear case that does not require AI to fail. It only requires AI to be oversupplied.

Key Takeaways

AI models now produce in a few hours work that took Citadel’s PhDs and master’s graduates four to six weeks, which Griffin says left him depressed one Friday before he came around to optimism.
He sees three effects at once: threatened white-collar jobs that need serious reskilling, corporate moats “filled in by bulldozers,” and a golden era for entrepreneurs and new business formation.
Startups matter because they hire hard workers without perfect pedigrees, giving people with tarnished resumes a new lane for their careers.
Citadel’s alpha still comes from human fundamental research; trading, the most automated part of the business, is how that research gets monetized.
After semis logged their longest losing streak since the financial crisis, Griffin says the market is finally “doing research” instead of buying every AI name.
If GPU supply exceeds demand, marginal pricing could collapse the way airline seats and 2000-era telecom fiber did, and the useful life of a GPU is a growing concern.
Large companies are splitting work between expensive frontier models from OpenAI and Anthropic and cheap open-source models, which creates tension with labs planning huge spending.
Big technological transformations are usually intertwined with bubbles, but what survives, like the internet and Amazon, leads for decades.
“Hedge fund deleveraging” should be answer number ten on any list of why prices moved, because the excuse stops you learning.
Citadel bought the Situational Awareness portfolio in under 24 hours and spent several days of trading managing its risk; Griffin does not expect a repeat soon.
He expects Republicans to lose one or both houses in the midterms and calls the Iran war poorly explained, the tariff battles with allies a disaster and the treatment of long-settled immigrants disgraceful.
He says he offered J.B. Pritzker public support for a tax increase in exchange for a constitutional amendment on pension reform, and was told to “pound sand.”
Pandemic spending, roughly equal as a share of GDP to the cost of winning World War II, meant there was no fiscal space to renew the tax cuts.
His priorities are growth, fixing schools, a Simpson-Bowles style entitlement vote and green cards for foreign graduates of elite US schools.
Citadel keeps each team in one location because “nothing beats a whiteboard and a good debate,” and its center of gravity sits between London, Miami and New York.

Chapters

0:39 AI and the Future of White-Collar Work

Burger opens on how much has changed in a year, with AI now solving decades-old math problems. Griffin recalls a Stanford talk where he described PhD-level work collapsing from weeks to hours, and the depressing Friday that followed. He argues the country needs real reskilling, pointing to English-to-German translators who have likely already lost their jobs, while corporate America will have to innovate faster. The upside is a wave of startups that hire people without perfect pedigrees, and he says his gloom has turned to optimism.

5:07 Why Research, Not AI, Still Produces Citadel’s Alpha

Griffin describes Citadel’s core business as research, “almost like journalism,” where the job is getting the story right about a company’s model, market and management. AI can shed light on pieces of that, but a person still assembles the narrative and decides the multiple. Trading is how the research gets monetized, and it has become the most automated part of the firm.

7:09 The AI Trade Gets Discerning: GPUs, Airlines and Dark Fiber

With semis on their longest losing streak since the financial crisis and IPOs delayed, Griffin says investors are moving past “buy them all” and asking how long high margins in DRAM and other constrained parts can last. He compares GPUs to airline seats and the 2000 fiber glut, where overcapacity sent pricing straight down. He also flags the useful life of GPUs, warning that strong pricing on five- or six-year-old chips probably does not persist another five or six years.

10:31 Corporate America Pushes Back on AI Bills

The most sophisticated companies are deciding which work needs costly state-of-the-art models from OpenAI and Anthropic and which can run on open-source models at a fraction of the cost. Burger raises a Wall Street Journal report on Anthropic’s S-1 citing $750 billion in planned spending, and Griffin agrees there is real tension. The deciding question, he says, is whether labs make their models a must-have, for example by helping unlock cures in life sciences.

13:27 Bubbles, the Internet and the Deleveraging Excuse

Griffin says there are almost certainly bubbles inside this moment, as there were in 1999 and 2000, and wishes he had kept his Amazon stock. His son once marveled that he lived before the internet existed, which he uses to show how transformative the last “bubble” was. He then dismisses “hedge fund deleveraging” as everyone’s favorite explanation for losses, saying it washes your hands of accountability.

17:34 Inside the Situational Awareness Portfolio Trade

Griffin places the summer deal alongside past portfolio takeovers such as Amaranth’s energy book around 2006. Situational Awareness needed to raise cash and protect a year-to-date gain, and Citadel offered a holistic solution, buying a complicated portfolio in under 24 hours. He explains how front-running by fast-money accounts can turn a feared liquidation into a real one, and why buying into that stress tends to be followed by a sharp price reversal. He says the fund was caught in “three mid-sized storms” and does not expect another such moment soon.

23:37 Midterms: Griffin’s Scorecard on Trump

Griffin calls the midterms a normal check on the party in power and expects Republicans to lose one or both houses. He notes he has backed Democrats too, including Rahm Emanuel and an early Obama fundraiser, and says a slight majority of his own staff are probably Democrats. He criticizes the unexplained Iran war, the treatment of immigrants who arrived when the border was open, and tariffs that have been “pro inflation,” while crediting the administration for closing the border. He also calls for a robotics-led reindustrialization, citing his recent $3 billion gift to Carnegie Mellon University.

32:16 Pritzker, Pensions and Listening to Voters

Griffin points to France’s recent turmoil as a warning about a state that spends 60% of GDP, and says Illinois Governor J.B. Pritzker is one candidate he fears is heading down that path. He recounts offering to publicly back a tax increase if Pritzker put a pension reform amendment on the ballot, when Citadel represented almost 2% of the state’s income tax revenue. Pritzker refused, lost his Fair Tax Amendment, and lashed out at Griffin instead. Griffin hopes that if Trump takes a beating in the midterms, he listens rather than doubling down.

35:08 Pandemic Spending, the Bond Market and a Growth Agenda

Griffin says the 2018 tax cuts were right but renewing them was not, because pandemic spending ate the fiscal space and is now crowding out borrowers through high mortgage rates and long yields. His priorities are to grow the economy and push Europe to do the same, fix schools (he cites 53 Illinois schools where not one student performed at grade level, and praises Success Academy), pass Simpson-Bowles style entitlement reform, and give green cards to foreign graduates. He calls growth “the only direction” out of the deficit and says Washington is not yet listening to the bond market.

44:00 Miami, Carnegie Mellon and Citadel’s Center of Gravity

Griffin has no relationship with New York Mayor Zohran Mamdani and says he does not want to be involved in New York politics, though Citadel’s 2,500 New York employees are. He places Citadel’s center of gravity over the Atlantic between London, Miami and New York, and says cities’ trajectories are written over decades. He credits Miami’s Cuban community for its ethos and expects Carnegie Mellon’s planned Miami campus to accelerate the city the way robotics is reshaping Pittsburgh. He closes on why Citadel keeps teams together physically, with whiteboards on the walls.

Notable Quotes

“As a business owner, you’re seeing your competitive moats be filled in by bulldozers.”
Ken Griffin, on watching AI do weeks of PhD work in hours

“If we reach a point where the provisioning of GPUs exceeds demand for GPUs, marginal pricing will actually go into freefall.”
Ken Griffin, comparing GPUs to airline seats and 2000-era fiber

“I always want the hedge fund deleveraging to be answer number ten on the list of answers.”
Ken Griffin, on why easy excuses stop investors learning

“The American people have spoken their mind, and you better damn well listen.”
Ken Griffin, on what a party should do after losing the midterms

“We spent as much money during the pandemic as a percentage GDP that we basically spent to win World War two.”
Ken Griffin, explaining why there was no fiscal space to renew the tax cuts

“Here’s your degree. Here’s your green card.”
Ken Griffin, on keeping foreign graduates of elite US schools in America

“Nothing beats a whiteboard and a good debate.”
Ken Griffin, on why Citadel keeps teams in one location

Watch the full conversation here.