Restrictive policy destabilizes industry and supply chains

Washington wants the world's chipmakers to build factories in the United States, but increasingly insists that they leave behind the commercial relationships that make those investments viable. Two recent developments involving Nvidia and the Republic of Korea's SK Hynix expose the contradictions in this ambition.

Nvidia has reportedly told some data-center partners they must ensure that at least 51 percent of their services are sold to non-Chinese companies. If implemented, the requirement would effectively introduce a nationality test into commercial arrangements involving access to important AI infrastructure, making Chinese participation in data-center ventures more difficult.

Competition is therefore no longer confined to which company produces the fastest chips or offers the most efficient computing services. Market access is becoming linked to political alignment. The predicament before SK Hynix exposes the contradictions more starkly. The chipmaker is seeking US export licenses to acquire equipment needed to expand production at its Dalian NAND facility in China. The additional revenue, it argues, would help finance a proposed $45 billion investment by its Solidigm unit in two memory-chip plants in New York, with the investment potentially reaching $100 billion over time.

However, the US Department of Commerce remains reluctant to allow US equipment to upgrade or expand production capacity in China, while the Department of the Treasury is waiting for the Department of Commerce's decision before considering the relevant investment restrictions. Nearly 10 meetings between Solidigm executives and commerce officials since late August have failed to resolve the impasse.

Washington wants the investment, the jobs and the technological capacity that SK Hynix could bring to New York. But it is reluctant to permit the company to generate the resources needed to make that investment commercially viable. New York officials are understandably anxious, particularly with an investment tax credit due to expire at the end of the year. For the ROK company, the uncertainty raises a more fundamental question: why commit billions of dollars to a project in the US if its existing global operations can be subjected to unpredictable political restrictions?

SK Hynix agreed to acquire Intel's Dalian NAND operations for $9 billion in 2020, before the tightening of US export controls. Its production networks, customers and investment decisions have since developed within an international industrial ecosystem that cannot be rearranged at the stroke of a regulatory pen.

As China's Ministry of Commerce has rightly pointed out, using export controls to pursue narrow national interests undermines the stability of the highly globalized semiconductor industry and supply chains.

Chinese Foreign Ministry spokeswoman Mao Ning stressed, in reply to a question on the report about the Nvidia requirement, that cooperation should be based on fairness, transparency and nondiscrimination.

Controls that become increasingly broad, unpredictable and discriminatory can discourage investment, distort commercial decisions and weaken the very supply chains on which US manufacturers depend.

Nvidia's reported requirement and SK Hynix's licensing predicament point to the same emerging problem: political considerations are increasingly dictating the terms of commercial participation. The more Washington turns ordinary business relationships into tests of geopolitical loyalty, the more difficult it becomes to convince international companies that investing in the US offers a predictable business environment.

A semiconductor industry divided into politically approved supply chains may sound reassuring to strategists in Washington. To the companies expected to finance and operate those chains, however, it looks more like an expensive experiment in economic fragmentation. The US may succeed in making its technology ecosystem more exclusive. Whether it can make it more competitive is another matter.