Ken Griffin has told a version of this story before. At Stanford Business School in May, the Citadel founder admitted that he’d recently gone home one Friday “actually fairly depressed” after realizing what AI could now do — a stunning reversal for an executive who’d dismissed the technology as “garbage” at Davos just months earlier.
What he didn’t say then was exactly what triggered it. Speaking with Goldman Sachs’s Raj Mahajan at the firm’s Apex Symposium on June 2 — audio Fortune has exclusive, embargoed access to — Griffin filled in the blanks.
The dinner story, inverted
The hedge fund billionaire has previously talked about having dinner at a table with a full roster of CEOs. At the Milken Institute conference, he described asking the table to share how AI was transforming their businesses and getting “six or seven extraordinary stories” back.
He offered a new twist on the tale to Mahajan, saying he was with a number of major CEOs at a dinner “about two years ago,” and his guests were effusive as to how AI was transforming their business.
Griffin added that he basically didn’t believe it. “I couldn’t help myself. I’m like, ‘Let’s go around the table and share stories about how AI is transforming your business.’” When he got four or five “incredible stories” back about productivity gains, he told Mahajan, he deduced that “not one involved AI.” The distinction, he added, is that these gains involved “machine learning,” “optimization” and “digitization,” which are not quite the same thing.
The distinction matters because the terms get used interchangeably even though they describe different things. Machine learning, in Griffin’s framing, refers to systems trained to recognize patterns in data and get incrementally better at a narrow task — the kind of technology Citadel has used for about a decade, citing examples like reading radiological reports or powering self-driving cars. Those systems can produce real productivity gains without exhibiting anything like general reasoning: they’re pattern-matching at scale, not making judgment calls or synthesizing new conclusions the way a human analyst would.
What Griffin means by “AI” — at least the version that alarmed him — is something closer to agentic systems that can read a task, plan a multistep approach, execute it, and check their own work, the way a finance-paper project did when it read an academic study, reproduced its methodology, and verified the results independently. A company crediting “AI” for a chatbot that “rewrite[s] the email you drafted for the last two minutes,” as Griffin put it, may really be describing a narrower, well-established machine learning tool wearing a fashionable label.
Within the C suite, he told Mahajan, he thinks this “nuance between AI and technology writ large gets a little bit lost.” In other words, most of what gets credited to “AI” in corporate America, in his telling, isn’t AI at all. It matters because this lack of clarity on thought has huge implications for big questions about how AI will change the economy. What if it’s literacy in machine learning, or optimization, instead? “There is a technological revolution happening, of which AI is a component of the story,” he told Mahajan, “but it’s just a piece.”
The project he wouldn’t name before
This crucial distinction is what makes the case that actually rattled him so notable. Griffin told Mahajan that one of his own team members built an AI system to reproduce and verify academic finance papers — the kind of work that Citadel’s quant researchers do routinely, testing whether published findings, like whether stock buybacks predict outperformance, hold up out of sample.
That process, Griffin explained, normally takes an expert-level researcher six to eight weeks. The agentic system did it in two to three hours per paper: reading it, reproducing the methodology, verifying the published results, then testing them out of sample.
“Here’s the key point,” Griffin said. “This is not just a white-collar job; this is a master’s or PhD-level job.”
Griffin’s message to workers and students has been that they should become lifelong learners and use these new AI tools to learn in different ways. In the Goldman conversation, he offered a fuller structural theory behind that advice. He described two ideas coexisting: real concern for workers whose skills are narrow and hard to redeploy — he singled out translation work as one case demanding serious retraining infrastructure — and, on the other side, a belief that competitive moats between companies are “being filled in at lightning speed.” The result, he predicted, will be a “golden age of entrepreneurial activity,” where small teams running agentic systems can challenge incumbents that once needed 30 or 40 employees.
Griffin was explicit that Citadel itself hasn’t cut a single job over these gains. “There’s no reduction to headcount at Citadel on the back of this breakthrough,” he said, framing the efficiency gain as expanding what his existing staff can tackle rather than a reason to shrink the team. Implicitly invoking what’s known as the “lump of labor fallacy,” Griffin argued that his “incredibly talented people” will take all of the productivity gains from technology (if not necessarily AI) and “we have just a huge swath of problems that we’re trying to attack and go after.”
Distinguishing signal from noise
The same instinct Griffin displayed on AI vs. technological advancements — separating the real signal from the narrative — extended to his reading of the rest of the world right now.
On markets, he argued that the S&P’s record highs, despite active wars in the Middle East and Europe, aren’t the product of investors ignoring risk but of a few specific, underappreciated shifts — the U.S. being partly insulated from the energy shock, a surprising collapse in Chinese oil demand, and a steady if irregular flow of oil still leaving the region.
His analysis of China followed the same pattern of separating hype from hard numbers. He calculated that China now leads in roughly 67 to 68 of the 75 most important global technologies. “As an American I get frustrated by this,” he said, arguing that the fix lies in education, not tariffs. On Taiwan, he was similarly concrete: a blockade cutting off access to TSMC’s chips could shrink US GDP by 8% within six months, with devastating effects. “Simply put, we go into a Great Depression in the blink of an eye. Unlike any we’ve seen before.” Boeing would stop making planes, automakers would stop making cars, “everything freezes,” he said, adding that it was a reason why he saw “no winners” from military escalation in Taiwan.
Even his approach to portfolio risk reflects the same worldview: distrust the abstract, focus on the concrete, measurable case. Asked about hedging, Griffin described a framework centered on stress-testing for a “definable, tolerable” worst-case loss, rather than trying to insure against every possible tail event — the same instinct that led him to dismiss all the vague AI enthusiasm until he saw a specific, verifiable case that convinced him otherwise—and helped him understand where it was really coming from.
The Mamdani Feud
Griffin didn’t stay confined to markets and AI. Asked by Mahajan about a viral moment involving New York City Mayor Zohran Mamdani and his New York City home, Griffin sidestepped naming the city’s mayor directly, before declaring: “I believe Citadel will be a principal player in financial services for far longer than [the mayor] will be mayor.” He argued that Citadel’s decision-making operates on a decades-long horizon that outlasts any single administration. “We intend to be here for decades. And he will be here for a few years.”
Griffin used the moment to renew his criticism of New York’s tax burden relative to city services, calling the gap “breathtaking” and asking “where is all this money being wasted?” He pointed to Citadel’s and Goldman Sachs’s expansions outside New York — Goldman’s new headquarters in Dallas and Citadel’s 1.7 million-square-foot tower under construction in Miami — as direct responses to high taxation and what he called “mediocre services.” He has not spoken to the mayor, he revealed, before pivoting to a debate about the merits of capitalism versus socialism and his befuddlement with an increasingly anti-capitalist Gen Z.
“How have we ended up with so many 20- and 30-year-olds who actually think socialism is the path to prosperity?” he asked rhetorically, arguing that there’s a “simple lesson” to draw from China’s history, as a transition to open markets lifted 1 billion people out of poverty. (The World Bank put the number at 800 million over 40 years, as of 2022.)
“It’s the greatest success story in the history of humanity,” Griffin said, urging the self-identified socialist politicians, “whether it’s Bernie Sanders, whether it’s Mamdani,” to “read a damn history book for once and then tell us how to run our country.”
Wealth, Art, and the Limits of Legacy
Asked by an audience member about his personal capital allocation, including his real estate and art holdings, Griffin was candid about art as an investment: “It’s fine at best. Its returns are uninteresting over the last 20 years,” he said, adding that collectors tend to publicize their winning purchases while staying quiet about the ones that lost money.
On intergenerational wealth, Griffin offered a strikingly unsentimental view. He said he “quibbles” with the idea of planning wealth for descendants four or five generations out, arguing that by that point, “there’s not much of you left in that DNA” and that distant descendants “may have the same last name” but are otherwise “like the stranger on the street.”
This story was originally featured on Fortune.com