The export-control debate has moved from what is allowed to what is bought. Friday's paper showed how H200 licenses cleared before chips shipped and how Huang's Beijing seat was a chip-export instrument. Saturday's policy fact is that Beijing's preference for Huawei domestic chips is no longer a leak — it is stated posture.

Reuters reported the United States cleared H200 sales to ten Chinese firms on May 14, and Chinese firms did not buy. [1] USTR Greer told reporters that export controls were "not a major part" of summit talks. [2] Beijing is steering buyers toward Huawei's Ascend line; the steering is now described in Chinese state-media coverage as industrial policy, not commercial coincidence.

That flips the analytic frame. The Washington question for two years has been which chips Commerce should license. The operative question now is whether Chinese firms will accept what is licensed. A cleared H200 with no buyer produces no revenue for Nvidia, no leverage for Washington, and no chip on a Chinese training cluster. The license becomes a press release.

The Huawei pivot has a denominator problem the policy debate has not solved. Ascend volume is constrained by SMIC fabrication capacity, by yield on advanced nodes, and by the absence of HBM equivalents. China can prefer Huawei without producing Huawei at the scale Nvidia could ship. Beijing is accepting that ceiling as the cost of the policy.

The unsold H200 license is the artifact. Saturday's silence on the order book is the policy fact.

-- DAVID CHEN, Beijing