What can Brussels and Washington do about the semiconductor problem?

First, pursue very narrow export controls and reduce tensions over the issue, seeking to “de-national security-ise” the semiconductor industry.  The key players whose countries lead globally in SME must decide about whether or not to control the export of advanced tools to Chinese companies. The US has already squeezed Dutch lithography leader ASML to halt shipment of an extreme ultraviolet (EUV) lithography system to Semiconductor Manufacturing International Corporation (SMIC), likely citing national security concerns (although in March this year ASML extended a 2018 agreement relating to older deep ultraviolet lithography (DUV) machinery). However, most semiconductors used in military specific applications are not at advanced nodes and are typically manufactured at much more mature nodes for which equipment is not typically controlled for specific end use companies.  SMIC does not appear to be a major contributor to China’s military modernisation, compared to say Phytium, tagged for the US Commerce Department’s Entity List. Phytium’s advanced CPUs, manufactured by TSMC, are used in supercomputing systems that among other things are used to model weapons of mass destruction and hypersonic vehicles. Export controls should be narrowly targeted for this very specific type of application, though this would be a complex undertaking given the dual-use nature of the technology. This could also be done with controls targeting Phytium’s customers as end users.

A deft application of export controls and industrial policy and trade negotiations involving the EU, US, and Japan, should include a major rethinking of how the three can confront China on the subsidies and national security issues, without enacting policies that will further disrupt delicate supply chains and result in closing off China as a major market for leading Western tool makers.  Its elements could include the following:

1) allow ASML EUV equipment to be shipped to SMIC. Over time, allowing SMIC to expand its capacity to manufacture advanced semiconductors would help increase the global capacity at the most advanced nodes and work to reduce the potential for industry crippling shortages. This move would likely have to be accompanied by removal of SMIC from the Entity List, or eased licensing restrictions, allowing it to continue to buy tools from US suppliers.[i];

2) enforce end user agreements around the equipment.  Any agreement around licensing for new tooling the EUV gear would have to include verifiable stipulations that SMIC would not use EUV systems to manufacture semiconductors for specific military end uses. This of course would be challenging, but this type of control has been implemented before, and the stick here would be the threat to cut off SMIC from ASML support, without which it could not continue to maintain and operate the highly complex EUV equipment; and importantly

3) roll back or narrow the use of unprecedented extraterritorial export controls like the foreign direct product rule (FDPR) used against Huawei. The semiconductor industry has already urged the US government to consider narrowing or rethinking the rule, (see also here) concerned about the disruptive effective it has had and will continue to have on the global sector.

At the same time, the EU, US, and Japan, perhaps as part of a technology alliance of the type being pursued at the G7, could seek to negotiate with China around broader semiconductor industry subsidies. The long-term concern of particularly the US semiconductor industry, that Chinese government subsidies would eventually result in distortions of global market driven system, are valid.  Yet there has been no attempt to discuss the issue with Beijing in the context of how to avoid having semiconductors become an issue like solar panels, where overcapacity in China resulted in the shutting down of competing companies globally as prices were driven ever lower. Chinese officials assert that they understand the problem and say the semiconductor industry is different and subsidies are unlikely to distort markets. Under this scenario, Beijing would have to become more transparent about how the National Fund works and agree to some limits on government subsidies—again, there are plenty of precedents for this type of agreement, based on WTO frameworks, with verifiable compliance mechanisms. Doing this at a time when both the US and EU are considering putting in place their own major subsidy programmes would appear to be prudent and provide fertile ground for compromise.

Admittedly these approaches and measures would have to be based on a major improvement of US China relations, and the re-establishment of some trust on technology issues between Beijing, Washington, Brussels, and Tokyo, which seems highly problematic in the short-term. But attempting to reduce both the national security and broader competitiveness concerns about semiconductor industry development and end use in China would help avoid unintended consequences of specific policy choices and improve the long-term health of the global semiconductor industry. The 2020 US export controls on Huawei, for example, contributed significantly to the recent semiconductor shortages, as the firm ramped up orders before the restrictions kicked in, and Chinese competitors, concerned about being targeted themselves, increased orders to boost inventory, exacerbating the situation created by cancelled auto industry orders, and huge demand increases stemming from the pandemic.  This was a case where political intervention in a highly efficient sector without consideration for the potential second and third order consequences has ended up having massive unforeseen consequences.

In addition, real progress in this arena would also potentially have a major impact on the broader problem of over concentration of advanced manufacturing in Asia, specifically Taiwan and South Korea. Currently, TSMC and its senior leaders are not eager to site and operate complicated and costly fab facilities in the US, and Taipei is equally as unenthusiastic, believing that keeping the status quo gives Taiwan significant leverage—to both ensure the US would come to Taiwan’s aid in a potential conflict with China, and to ensure that the intellectual property of key national champions like TSMC is more easily protected.

The easing of US-China tensions, brought about by a de-escalation of conflict over the semiconductor sector, would almost certainly also reduce pressure on players such as TSMC, giving senior company officials time to develop a long-term view of the benefits of diversifying a significant level of manufacturing capacity to developed country markets where governments are offering generous subsidies, and avoiding the false choice of having to choose “blue” or “red” supply chains.

The EU and US would need to ensure that structures existed to manage long-term investment and subsidisation of advanced manufacturing clusters in Arizona and Berlin, for example, and that there were measures in place to ensure investment in education for a well-trained workforce—one of the factors cited by TSMC for being reluctant to cite fabs outside of Taiwan is finding sufficiently trained personnel, in addition to higher costs for construction and key materials and other inputs like water and power.

Such a holistic and long-term strategy for the semiconductor industry should seek to minimise disruptions to supply chains and avoid cutting off the fastest growing market for semiconductors and SME in China from both major players such as TSMC, Samsung, and Intel, which need to consider growth in their China business to ensure alignment with the economics of building out more advanced fab capacity; and hardware and software tool makers, who are also counting on growth in China to boost long-term revenue, much of which is plowed back into R&D and helps drive innovation, also an economic and national security issue for the US and EU. Industry groups continue to make this point in arguing against export controls targeting individual companies for often unclear policy objectives.  

Clearly the China semiconductor problem is one of the most complex policy issues facing governments in the US and EU. In fashioning new policies and approaches, government players must account for the global and complex nature of semiconductor industry supply and value chains, and avoid taking action that destabilises these structures, developed over many years, while at the same time determining how best to accommodate new players such as China, and ensure the long-term health of this critical global industry.

-----

[i] SMIC was also added to the Department of Defense 1999 NDAA Section 1237 list, linking it to China’s military. SMIC has denied that it is owned or associated with China’s military. Chinese smart device maker Xiaomi challenged the US designation earlier this year, and a US District Court judge ruled that the US government had not proved compelling evidence of a military link—the Pentagon agreed to remove Xiaomi from the list in May. SMIC and other designated companies are likely to similarly challenge the ruling.

The author wishes to thank Rogier Creemers of Leiden University for reviewing and commenting on the draft of this article.

By Paul Triolo, Eurasia Group, Paulson Institute

Article category: Digital Technology and R&D