Ken Griffin urges US data centers to hedge geopolitical risk

Ken Griffin shifts focus to survivable losses, energy tail risks, AI agent reality, and Taiwan chip shock

Size losses you can survive, not every tail risk. Stay in the game when extremes hit.

- You can’t predict or hedge all tails. First compute the worst loss you can bear. Portfolios aim to endure, not avoid all risk.

- US stocks sit near highs because the US is shielded from an energy shock. If the Strait of Hormuz shut, oil could near $200. Instead, a larger-than-expected China demand drop and continued shipments keep it near $100. Markets discount unrealized energy tail risk, not price peace.

- Don’t take “AI productivity” at face value. Many cited wins are ML, optimization, or digitization, not generative AI. Separate outcomes driven by frontier models and agents.

- Agentic AI is the change vector. One system did work that took several PhDs eight weeks in three hours. Citadel didn’t cut headcount. First-order value is solving problems that were uneconomic before, not staff reductions.

- Alt data nowcasts earnings in near real time. Edge shifts from “guess this quarter” to reading 2nd- and 3rd-order effects on industry structure, competition, and capital allocation. Surging compute costs raise barriers and sideline capital-poor, low-margin funds.

- A Taiwan semiconductor supply shock may not be a “dip,” but a sudden, depression-like scenario. Building data centers in the US functions as insurance. Domestic builds cost more, but concentrating core compute abroad carries a higher geopolitical risk cost.