Washington's AI Strategy Is Betting Against China's Playbook

Washington has poured trillions of dollars into keeping America's AI lead intact — underwriting frontier labs, sprawling data centers, and export controls meant to starve rivals of advanced chips. Beijing, confronted with those same restrictions, has taken the opposite tack: shipping “good enough” open-weight models like Moonshot's Kimi K3 at a fraction of the cost and letting them proliferate across markets the United States has largely written off. The contrast unsettled even hawkish observers this week, with Harvard's Graham Allison warning that a rival running such a different playbook “should make Americans pause and think.” The bet built into Washington's approach is that raw capability will keep compounding faster than China can copy it, and that scarcity of frontier chips will always outweigh abundance of merely adequate ones. But cheap, widely distributed models carry a different kind of power: they set technical standards, seed dependency in markets the U.S. isn't bidding for, and normalize a Chinese stack in places American labs have no presence at all. If frontier supremacy and mass adoption turn out to be different games, the side optimizing for the wrong one may find its lead was never quite what the benchmarks suggested. That is the quiet anxiety running underneath a week of triumphant chip announcements.