All In Podcast: Chip Crash, AI Leverage, and the Frontiers of Intelligence

**SPEAKER_1** (0:01)
When the fastest moving trade in markets turns, the unwind can be brutal. On All In, Chamath Palihapitiya, Jason Calacanis, David Sacks and David Friedberg dig into the chip stock crash, the AI boom and the ripple effects hitting hedge funds, South Korea and frontier AI labs.
The semiconductor sell-off is really the centerpiece here. The Philadelphia Semiconductor Index dropped more than 20% in a month, and Samsung, SK Hynex and the broader Cospi in South Korea were hit even harder. But the panel frames this as a violent momentum unwind, not a collapse in the AI story itself.
And then there's the Leopold Aschenbrenner situation. He's described as a 25-year-old hedge fund manager who reportedly got margin called after riding the AI and chip wave with heavy leverage. One speaker puts it perfectly. Leverage is the only way smart people go broke.
That's the critical warning about risk. If you're unlevered, a bad month can still become a rebound. But if you're three or four times levered, a small drawdown becomes a wipeout. Prime brokers can force liquidations automatically, which makes the unwind especially merciless.
The macro pressure is intensifying too. The 30-year treasury yield crossing 5.2% is a major signal because government bonds suddenly compete with risky growth bets. Add persistent inflation, a roughly $2 trillion annual deficit, and energy pressure from conflict in Iran, and you've got a much tougher environment for high multiple stocks.
Still, there's disagreement about what the market is missing. The speakers argue that the AI capex boom is real, and productivity gains are being underestimated. They point to energy transition trends, and suggest that solar, batteries, and massive buildouts could lower costs more than current forecasts assume. The efficiency angle is fascinating. The panel argues that future models may cut token usage by 50 to 75%, changing the economics of software work. They describe this shift toward on-demand intelligence, where models listen, automate, and continually improve work flows. Then, the conversation turns to safety and politics. They react skeptically to a petition signed by 1300 workers across Anthropic, OpenAI, and other labs calling for paced development. One line captures the pushback. There is no need for one savior or one Moses to take us across the desert. A major debate emerges around whether frontier AI is really a duopoly. The panel says Anthropic and OpenAI are strong today, but open source models are gaining because they are cheaper, more flexible, and better for control and privacy. They call this software freedom. There is also the copyright question. They reference Google Books and argue that converting books into searchable knowledge was treated as fair use. That logic may extend to AI training, even as lawsuits continue.
Later the episode pivots to Mamdani's proposal for city-owned grocery stores in New York City by 2029 It's presented as populist policy, offering discounts and higher wages, though critics fear inefficiency and crowding out private stores. The episode closes philosophically. They discuss a Budapest neuroscience model, using a fruit fly brain map that worked better in 64 dimensions than in 3D geometry. It sparks reflection on biology, consciousness, and how little silicon-based systems truly understand compared with living systems. The speakers describe biology as a frontier still beyond full comprehension, moving into faith, science fiction, and the stories we use to explain what we cannot yet grasp.