U.S. Floats Export License for Nvidia’s H200 AI Chips to China

Washington’s dilemma echoes the 1987 Toshiba-Kongsberg episode, when the U.S. tightened COCOM rules after Japan sold submarine-quieting machine tools to the USSR; then, as now, commercial allies sold second-tier tech that still had strategic bite. The debate sits at the confluence of three long arcs: (1) the century-long pattern of the U.S. using export controls to slow rival industrial catch-up (from the 1949 Coordinating Committee to the 2019 Huawei ban); (2) the relentless two-year cadence of GPU leaps that quickly turns “cutting-edge” into “previous-gen,” blurring security thresholds; and (3) the shift from single-supplier chokepoints to a multipolar semiconductor landscape where China, the Gulf, and others court alternatives. Whether the H200 license is granted or blocked may matter little in 2125—the performance gap will vanish within product cycles—but today’s decision will signal how the U.S. balances short-term leverage against fostering foreign self-sufficiency, a tension that historically tends to push targeted nations (e.g., post-embargo France’s 1960s force de frappe) toward rapid indigenous capability.