Here is Ginn's WSJ letter:
Matt Pottinger, Liza Tobin and I share a common goal: The U.S. must win the artificial-intelligence race. Yet their argument for doubling down on failed GPU export controls falls apart under scrutiny and real-world conditions.
In their Aug. 11
letter, they write that “the controls on
Nvidia’s H20 chips appear to have been working until CEO
Jensen Huang’s lobbying secured a reversal.” But Nvidia hasn’t shipped H20s to China since the ban. The Commerce Department approved exports this week under a new deal, which returns 15% of sales revenue to the American people. As I noted in my
op-ed, Z.AI’s breakthroughs rely on more-advanced chips than the H20, and DeepSeek released new models despite U.S. policy.
Mr. Pottinger and Ms. Tobin suggest that Nvidia’s product can be reverse-engineered easily. Yet the company is more like
Apple or
Boeing than Intel or AMD, with its value in its software stack and integrated design. Without CUDA—the company’s computing platform—Nvidia hardware is an expensive paper weight, like an iPhone without iOS. Buying a PlayStation doesn’t guarantee that you have games to play.
Mr. Pottinger and Ms. Tobin likewise confuse buying a semiconductor with making one. Citing a foundry equipment maker as proof that China can replicate Nvidia is like saying owning a PlayStation means you’re
Sony. This oversimplification is a trillion-dollar category mistake.
Both lament the Trump administration’s repeal of the Biden-era “diffusion rule”—which lumped Portugal and Switzerland in with Yemen and Ukraine for GPU access—and oppose Mr. Trump’s policy of exporting American AI. Yet to lead in this space we need policies rooted in technical reality and strategy, not ones that equate sandy beaches and luxury watches with failed states and war zones that weaken American soft and economic power around the world.
Aaron Ginn
Boulder, Colo.