Sam Altman sat down with Lex Fridman for a sprawling two-hour conversation, touching on everything from GPT-5 and Sora to the boardroom drama that nearly pushed him out of OpenAI, and the long arc toward AGI. He wasn’t there to lay out a product roadmap; he was preparing the ground for a world where compute is the most precious resource, and the race to build artificial general intelligence is inseparable from a giant power struggle.
OpenAI began in 2015 as a nonprofit promising to develop AGI for the benefit of humanity. But the sheer cost of compute forced the organization into a capped-profit structure—a hybrid that allowed it to attract the billions needed for training runs while theoretically capping returns to investors. That uneasy compromise set the stage for the explosion in November 2023: the old board fired Altman, citing a lack of candor. Within days, nearly all employees threatened to quit, Microsoft applied heavy pressure, and Altman was reinstated with a reshuffled board. He now casts the episode as a painful but necessary stress test, arguing that surfacing governance tensions early builds immunity for the even tougher fights ahead.
At the heart of his argument are scaling laws: as you pump more compute into models, their capabilities improve in remarkably predictable ways. For Altman, compute isn’t just a cost—it’s the currency of the future. Whoever controls the chips, the energy, and the massive data centers will set the pace in the AGI race, and that lead translates into extraordinary power—commercial, geopolitical, and beyond. He is blunt about it: the first mover will have outsize control. OpenAI’s own structural experiment—a safety-obsessed nonprofit surrounding a commercially aggressive engine—could, he hopes, provide lessons for global AI governance.
Critics see plenty of holes in this frame. Framing the AGI journey as an existential arms race can all too easily justify secrecy and monopoly, shutting out competitors and the public alike. Nor is it clear that enduring a boardroom putsch genuinely strengthens an organization; it can just as easily hemorrhage talent and erode trust. The “compute as currency” narrative also glosses over real-world constraints: chip fabrication capacity, energy supply, and the risk that concentrating compute in a handful of hands will lock in global inequality. More fundamentally, the November crisis raised an uncomfortable question: should a single private company hold a decisive say over a technology that could reshape civilization? Many argue for broader democratic oversight, robust open-source efforts, and a far more inclusive governance model.
Altman has turned the ordeal into a story of personal growth, describing himself as tougher and more resilient. But the structural cracks remain. The tension between a nonprofit mission and breakneck commercial expansion hasn’t gone away. Accountability is still murky. And the deeper question—how to design governance that is both innovative enough to push forward and robust enough to steer AGI safely—is as unsettled as ever. The five-day drama may have strengthened one leader’s resolve, but the real test will be whether OpenAI, and the world, can design institutions that survive much bigger shocks.
