Beginning during the Trump administration and continuing under President Joe Biden, the United States’ technology competition with China – conducted through major policy decisions, regulatory actions and think tank position papers – has become an important aspect of bilateral relations. This intensifying competition is likely to have profound regional security implications, particularly for the potential for military confrontation over Taiwan but also more generally for the long-term strategic and economic security of the Asia-Pacific.
In both Washington and Beijing, strategic goals have sometimes been driven by forces intent on over-securitising the technology dimension of bilateral relations. This process has been fraught with risks and potential collateral damage for bilateral US–China relations and the Asia-Pacific, as well as more generally for global supply chains in key technology sectors, such as semiconductors.
Taiwan is increasingly a flashpoint in the bilateral relationship, a development linked to changes in the semiconductor industry, which has been increasingly affected by US–China competition. US efforts to restrict Chinese companies’ ability to use Taiwanese firms as a manufacturing platform – particularly global foundry leader Taiwan Semiconductor Manufacturing Company (TSMC) – and Washington’s pressure on Taipei to support US over Chinese supply chains creates a new red line for Beijing, though it is unclear what might trigger a Chinese response.
US–China technology competition in emerging sectors such as artificial intelligence and quantum computing could lead to the decoupling of supply chains and research and development, reducing US (and US allies’) understanding of China’s progress in these areas.
SINO-AMERICAN TECHNOLOGY COLD WAR
The author first used the hashtag #USChinaTechColdWar in January 2018,1 though the term ‘technological cold war’ was first employed by China technology watcher Paul Mozur of the New York Times in July 2015.2 The context of the 2015 article – important for the future direction of US–China technology competition – was the attempted acquisition of Micron Technology, the leading US semiconductor-memory company, by China’s Tsinghua Unigroup. The latter had support from the China Integrated Circuit Industry Investment Fund (CICF), established in 2014 as part of a series of major national technology-related initiatives actioned under Chinese President Xi Jinping.3 Tsinghua Unigroup would later abandon the deal as the Committee on Foreign Investment in the US (CFIUS) came under increasing congressional pressure to review it. Over the next five years, China became increasingly authoritarian in the eyes of Washington policymakers as Xi ramped up efforts to support domestic technology companies. This context set the stage for a spiralling dynamic that would culminate in 2021 and 2022 with geopolitical risks erupting around the entire semiconductor sector and one of its epicentres (Taiwan). To understand how this happened one must follow several interrelated threads, including the failed Micron deal, the expansion of US regulatory purviews and powers and lastly Beijing’s reactions and countermeasures, both current and future.
CFIUS is an inter-agency committee that reviews proposed acquisitions of US companies by foreign firms through the prism of national security. It typically employs a mitigation strategy to address any concerns while allowing some portion of a deal to go forward. The committee had already been reviewing semiconductor-related acquisitions as far back as 2012. However, Xi’s focus on building China into a cyber superpower (on par with the US4) and the establishment of the CICF under his aegis generated greater concern in Washington regarding the potential that Micron’s acquisition could be used to assist Chinese firms. CFIUS said it was part of ‘an effort among foreign governments or companies to acquire U.S. companies involved in research, development or production of critical technologies’.5 Both sides abandoned discussions of the deal after receiving clear signals from CFIUS that it would not be approved.
CFIUS then adopted a much harder line on Chinese acquisitions of semiconductor companies. Another major milestone – in what would be an increasingly central theatre of the US–China ‘Tech Cold War’ – was president Barack Obama’s decision in December 2016 to block Chinese investors from acquiring Aixtron US, a subsidiary of a German maker of manufacturing equipment (such as gallium arsenide) for compound semiconductors that have military applications. The company had been targeted by a provincial Chinese investment fund with ties to the CICF. In this case, the very unusual route of issuing a presidential executive order,6 based on a recommendation from CFIUS, indicated that the US government would no longer hesitate to act directly and promptly on this type of deal. In the past, firms involved in a transaction had backed down merely as a result of the threat of a CFIUS recommendation to deny the deal in question; only in rare cases was a presidential action recommended or needed to kill a deal.
The coupling of concerns over Beijing’s subsidies for China’s domestic semiconductor industry and the potential for the CICF to fuel multiple acquisitions of US semiconductor-technology companies became a hallmark of the Trump era, which also saw broader US–China technology-related tensions spread to many other technology sectors, including artificial intelligence (AI), quantum computing and biotechnology. This focus on the potential impact of government subsidies generated several key strands that remain part of the ongoing US–China technology-competition narrative.
Table 7.1: Chinese semiconductor companies on the US Commerce Department Entity List and Treasury Department NS-CMIC List
*Better described as a supercomputing company. Note: ‘NS-CMIC List’ refers to the Treasury Department’s Non-SDN Chinese Military-Industrial Complex Companies List, where SDN stands for ‘Specially Designated Nationals’.
Sources: US, Federal Register, www.federalregister.gov; US, Commerce Department, Bureau of Industry and Security, www.bis.doc.gov; US, Department of the Treasury, home.treasury.gov.
Tentative US steps to use the subsidy issue as a justification for applying export controlsConcern about the potential for the CICF to distort the highly competitive and global value chains critical to the semiconductor industry also became part of the US debate late in the Obama administration. The president’s last commerce secretary, Penny Pritzker, claimed in late 2016 that China was using the fund to try to ‘appropriate’ the global semiconductor supply chain.7 Though Pritzker was referring primarily to the fund’s role in backing acquisitions, she was also addressing its fuelling of overinvestment in the sector in China and undermining of more market-driven players, such as commodity semiconductor producers in the memory sector. This concern became the basis for the 2018 Commerce Department Entity List action against Chinese semiconductor-manufacturing start-up Fujian Jinhua (see Table 7.1).8 The start-up had been involved in a case involving intellectual-property theft from Micron. However, after receiving significant subsidies from the CICF, it was blacklisted for potentially undermining established US dynamic random-access memory producers that might be future suppliers to the US armed forces. One Washington trade lawyer called the move a ‘dramatic expansion of the use of the Entity List for economic purposes’.9
Expansion of US concerns to alliesAs Chinese firms determined that the US would be a much less fertile ground for acquisitions in the semiconductor and other high-technology sectors, they turned to other countries and regions. However, they have found that European and Asian countries have also stepped up their efforts to monitor transactions in sensitive technology sectors, including those involving semiconductors, AI and robotics. In addition, the US government under both Donald Trump and Joe Biden has worked hard with so-called ‘like-minded’ allies to share information on transactions deemed sensitive and related to national security. This allied effort intensified following Chinese electronics-maker Midea’s late-2016 acquisition of German robot-maker Kuka.10 The acquisition, which Kuka claims obtained CFIUS approval, drew significant criticism in Europe after being approved by Berlin.11 In the geopolitical climate of 2022, there is little chance that such a deal would be allowed to proceed.
EXPANSION OF EXPORT CONTROLS’ EXTRATERRITORIAL APPLICATION
Following the Kuka sale, the tightening of mergers and acquisitions and investments involving Chinese entities was increasingly a focus for both CFIUS and an emerging investment-review process in the European Union. The late Obama era also saw critical and unprecedented export-control actions targeting Chinese telecoms firm ZTE kick off another key strand of the confrontation over semiconductors. Under Trump, the ZTE action also spawned a spate of Commerce Department Entity List actions, including those against China’s leading technology company Huawei (a competitor to ZTE) and Sugon and Phytium – Chinese makers of central processing units (CPU).
Placing large multinational Chinese companies such as ZTE and Huawei on the Entity List came with a host of unanticipated second- and third-order effects. Although the 2016 listing of ZTE12 was put on hold after the firm agreed to pay a fine and discipline executives, in April 2018 then US commerce secretary Wilbur Ross imposed a denial order on the firm for allegedly violating the original provisions of the 2016 deal.13 Only a phone call between Trump and Xi in July 2018 – when the firm faced an increasing threat of bankruptcy – saved ZTE by again holding the Entity List action in abeyance (although new fines and monitoring requirements were imposed).14 Huawei was hit with an Entity List action in May 2019, soon after US–China trade talks had collapsed.15 This development sparked a months-long process during which US subcontractors to Huawei continued to ship some products to the Shenzhen firm (complying with the letter of the underlying Export Administration Regulations) while the Commerce Department’s Bureau of Industry and Security (BIS) lacked clear guidance on how to adjudicate incoming licensing requests. In part responding to this cat-and-mouse game, US political appointees at the Commerce Department developed a way to strengthen the restrictions on suppliers to Huawei by crafting a change to the Foreign Direct Product Rule (FDPR), issued initially in May 202016 and then amended to provide further clarity in August 2020.17 With this amendment, the rule now extended US export controls in an unprecedented manner to all manufacturers globally that used any US technology to manufacture semiconductors on behalf of Huawei.
The FDPR’s extraterritoriality effectively dragged Taiwan and its semiconductor-manufacturing champion and global foundry leader Taiwan Semiconductor Manufacturing Company (TSMC) into the growing US–China technology competition. In May 2020, TSMC announced that it was suspending cooperation with Huawei, which at the time represented at least 15% of TSMC sales and was a fast-growing customer for the firm’s foundry services. It also announced that by September 2020 Huawei would be unable to source from TSMC advanced semiconductors for any of its product lines, undermining Huawei’s business model.18 While Huawei continued in business, drawing on significant stockpiles of TSMC-made semiconductors, by early 2022 it was clear that it could no longer compete in its two major product lines and revenue sources: smartphones and critical parts of its advanced mobile-telecommunications-equipment line.19 These developments set the stage for the potential broader use of the FDPR.
PUSH TO PROTECT FOUNDATIONAL TECHNOLOGY: SEMICONDUCTOR MANUFACTURING EQUIPMENT IN THE CROSS HAIRS
As concern grew within the Trump administration over China’s rise as a technological and military power, academic studies and growing debate within Congress on these issues culminated in the 2018 effort to revamp the US export-control system. The so-called Export Control and Reform Act (ECRA), passed as part of the 2019 National Defense Authorization Act,20 called (among many other new provisions) for the BIS to draft lists of foundational and emerging technologies that should be considered for new export controls and also monitored for transactions falling under CFIUS. The semiconductor manufacturing equipment (SME) sector, where US companies are among global leaders, was an important part of the overall debate, which centred on how and whether to control (or reassert control over) a broader range of SME going to specific Chinese end users deemed to be providing support for China’s military modernisation.
The debate around ECRA, SME and foundational technology helped to fuel new policy initiatives during the Trump era that have continued into the Biden era (see Table 7.2 for details on the United States’ options to pursue policy action against China). The most important aims of these initiatives are discussed below.
Table 7.2: The United States’ options for pursuing policy action against China in the context of US–China technological competition
Preventing companies in China from obtaining advanced equipment related to semiconductor manufacturingAll the technologies specific to producing advanced SME – such as extreme ultraviolet lithography (EUV) – are controlled under US export controls and the multilateral Wassenaar Arrangement, which replaced the Cold War era Coordinating Committee for Multilateral Export Controls in 1996. A company wishing to buy EUV equipment from the sole supplier, Dutch firm ASML, must get a licence approved by the Netherlands government. Beginning in early 2018 and reflecting growing US concern over China’s rise as a technological power, the US government has applied pressure on the Netherlands to deny a licence to Chinese foundry leader Semiconductor Manufacturing International Corporation (SMIC).21 The effect of this restriction has been to freeze China’s domestic semiconductor-manufacturing capability somewhere between the seven nanometre (nm) and 10 nm nodes (see Tables 7.3 and 7.4). (‘Nodes’ refers to the feature size of individual elements of a semiconductor that depend on the level of advanced processing; currently the most advanced levels in commercial operation or at the research and development (R&D) stage are pushing down towards 5 nm, 3 nm and 2 nm). In addition to restrictions on SMIC’s ability to obtain EUV equipment, in late 2021 US officials vetoed the installation of EUV equipment in a manufacturing facility in Wuxi, China, operated by South Korean semiconductor giant SK Hynix, probably citing concerns about the potential for diversion of the technology.22
Expansion of restrictions from semiconductors to more of the equipment used to manufacture chipsIn addition to imposing restrictions on EUV equipment and making slow progress on developing a list of foundational technologies likely to include SME capable of manufacturing at more mature nodes, late in the Trump administration the Commerce Department moved to add SMIC to the Entity List. As a result, US suppliers to the firm would need to apply to the department for a licence, often with the presumption of denial. The language of the listing specified that equipment ‘uniquely required to produce semiconductors at advanced technology nodes 10 nanometers or below will be subject to a presumption of denial’.23 In early 2022, there was continued discussion within the Biden administration about a Department of Defense proposal to change the language from ‘uniquely required’ to ‘capable of’. The reasoning was that typically no SME is designed specifically to produce a particular node. Such a change would potentially subject a wider range of equipment to licence denial. It was opposed by career officials at the Commerce Department and by US semiconductor-equipment manufacturers.24
Table 7.3: Number of semiconductor-manufacturing firms at key technology nodes, early 2022
Sources: SEMI, www.semi.org; Semiwiki, semiwiki.com; anandtech.com, www.anandtech.com
Table 7.4: Major industry players and the process-node levels at which they are working, early 2022
Sources: SEMI, www.semi.org; Semiwiki, semiwiki.com; anandtech.com, www.anandtech.com
SEMICONDUCTORS AND THE RISK OF A TAIWAN CONFLICT: PUSHING ON UNKNOWN RED LINES
The rolling actions taken by Trump and continued under Biden in the semiconductor and other high-tech sectors – coupled with other elements of the ‘Tech Cold War’ dynamic – have contributed to a broad cross-party consensus in Washington among certain elements of the foreign policy community, conservative think tanks and within the US intelligence community.25 The consensus perspective holds that the US and China are locked in a long-term struggle to dominate the ‘technologies of the future’, including both some already in existence – such as advanced semiconductors and next-generation mobile-telecommunications technologies – and also emerging sectors, such as AI, quantum computing and biotechnology. There is an important subtext to this perspective that holds that Beijing, left to its own preferences, will eventually misuse or exploit these technologies (or its companies’ role in particular technology sectors) to damage or undermine Western institutions and values, either by using them in ways that would contribute to human-rights violations; by exporting them to authoritarian governments; or by seeking to exploit them for China’s military modernisation, particularly in relation to a potential attack on Taiwan. By the time Biden took office in January 2021, this narrative was firmly entrenched in Washington policy circles, as well as in the Department of Defense and the US intelligence community.
Another important element of the emerging conventional wisdom regarding the US–China Tech Cold War was that US companies – particularly those in the cutting-edge sectors now part of the struggle for the so-called technologies of the future – should not be assisting China’s military modernisation. Increasingly during the Trump administration, rhetoric from Washington policymakers and think tanks focused on the Chinese leadership’s civil–military fusion and integration initiative, which is an attempt to exploit technologies developed by the private sector for military purposes.
From late 2021, several of these strands converged in ways that are likely to affect the Asia-Pacific by raising pressures on supply chains and innovation ecosystems and by driving further wedges between China and Taiwan – with unknown consequences for regional stability. Washington was angered by revelations in late 2021 that Phytium CPUs manufactured at TSMC may have been used in high-performance computers that modelled advanced Chinese weapons systems, such as a hypersonic glide vehicle apparently tested in July–August 2021. Some congressional hardliners on China wrote to Commerce Secretary Gina Raimondo in October 2021 following reports of the hypersonic-missile test. They urged her to extend the FDPR to all ‘entities that enable the CCP’s military capabilities and human rights abuses’.26 Then, in early 2022, the semiconductor element of US–China technology competition saw renewed emphasis on Taiwan and TSMC, specifically the latter’s continued viability as a platform for Chinese firms to manufacture advanced semiconductors. From Beijing’s point of view, the US had already cut off its leading technology from TSMC, along with a major CPU-maker, Phytium, which was added to the Entity List in April 2021.27
US officials are approaching what may be called Beijing’s unknown red lines, given the prospect that the further extraterritorial extension of US export controls could cut off more Chinese technology firms from using TSMC and other Taiwanese semiconductor-manufacturing companies, such as United Microelectronics Corporation (UMC). The importance of the semiconductor sector for China has been highlighted by Xi’s launching of the CICF and his role in leading Chinese Communist Party (CCP) Politburo ‘study sessions’ on semiconductors and other high-tech sectors. Developing these sectors and reducing China’s reliance on foreign sources are key parts of the 14th Five-Year Plan – published in March 2021 – and Xi’s overall push for ‘self reliance’, which was boosted in 2021 with Beijing’s ‘dual circulation’ and ‘common prosperity’ initiatives.
So far, the Chinese government has not even threatened to target US technology-company operations in China in retaliation for US actions that have undermined Huawei’s business model and future viability. However, it remains unclear how Beijing will react to the prospect of having most or all of its technology leaders cut off from Taiwan and TSMC. The prospect that SMIC will not be able to obtain EUV equipment to enable the firm to move to production at more advanced nodes domestically – as a hedge against Chinese firms being delinked from TSMC – creates more uncertainty. US officials deliberating changing the Entity List language on SMIC appear to favour first holding discussions with international allies on the issue, probably as part of semiconductor supply-chain and export-control collaboration taking place in the ‘Quad’ format (a grouping comprising Australia, India, Japan and the US) via its Emerging Technology Working Group and at the EU–US Trade and Technology Council. For China, the worst-case outcome of these discussions would be a united front among the US and its international partners on export controls covering SME and other advanced and emerging technologies.
In early 2022, Beijing was concerned about another Washington discussion on extending the FDPR. The talks took place in February when the Commerce Department extended FDPR restrictions to Russia broadly and added dozens of Russian firms to the Entity List as retaliation for the Russian invasion of Ukraine.28 The new rules apply the FDPR against Russia in a blanket fashion and specifically require an export licence – likely to be denied – for any information and communications technology (ICT) equipment that includes US technology intended for transfer to specific Russian firms or organisations.29 It is the first time that Washington has effectively weaponised the entire US technology supply chain to target an entire country (rather than a single company), in a manner similar to the use of the dollar in US Treasury Department sanctions. A senior official with the Semiconductor Industry Association, Jimmy Goodrich, who is familiar with the US Entity List and FDPR actions, noted: ‘We could be in uncharted waters with such a potentially broad export control measure. We are still trying to assess what the ripple effect may be to global supply chains.’30 A major issue that remained unclear in late March 2022 was whether China would go along with such an action and, if not, whether Chinese firms shipping electronics to Russia could be the target of secondary sanctions. The broad use of the FDPR in this manner would set a dangerous precedent, according to industry officials, by disrupting supply chains, driving countries to ‘design out’ US technology, and adding to concerns in Beijing that China or a much broader set of Chinese companies could be subject to the same treatment, particularly under a potential Republican-dominated Congress later in 2022.
The broader technology-policy environment is also increasingly hostile for Beijing. The US-organised Summit for Democracy held in December 2021 and the associated Alliance for the Future of the Internet, launched in early 2022, use language such as the need to promote ‘democracy-affirming’ technologies and to control technologies that could be used by ‘techno-authoritarian’ states, meaning principally China and Russia.31 As the debate intensified over how to handle sales to China of advanced and emerging technologies that could contribute to its military modernisation and government surveillance operations, the issue of the role of Taiwan (and particularly TSMC) generated media speculation in late 2021 in relation to the potential that China will use military force to achieve ‘reunification’. Taiwanese President Tsai Ing-wen has emphasised the importance of Taiwan’s ‘silicon shield’32 – in other words, the dominance of TSMC and other Taiwan-based companies in the semiconductor-manufacturing sector – as a potential deterrent to Chinese military action. Some US commentators have even suggested that Taiwan should threaten to destroy semiconductor facilities in the event of Chinese military action against the island.33 That the discussions reached this point indicated US and Taiwanese policy circles’ growing recognition of the centrality of Taiwan and TSMC to the technology sector in general, to China and to the global economy.
The US and its allies in Europe and Asia are working to reduce their companies’ dependence on TSMC via new funding for domestic semiconductor manufacturing. However, it will be at least a decade before there is significant change in the sector. The Biden administration’s CHIPS for America Act,34 which is unlikely to receive funding until May 2022 or later, will provide incentives for TSMC, Samsung and Intel to locate advanced facilities in the US. The EU Chips Act will attempt to provide similar subsidies to incentivise these firms to consider placing facilities in Europe. It is likely that over the next five years Washington and its allies will make progress towards reducing the centrality of Taiwan and TSMC in semiconductor manufacturing. However, as a result of the risks around US export-control and other policies, there will likely be increased tail risk that the semiconductor issue will play a larger role in Beijing’s calculus around making a military move against Taiwan – especially if there is a substantial reduction in Chinese firms’ ability to use Taiwan as a manufacturing platform. Therefore, the consequences of US–China technology competition blowing up over Taiwan are real and growing, with major implications for the wider Asia-Pacific.
COMPETITION IN EMERGING TECHNOLOGIES: ARTIFICIAL INTELLIGENCE AND QUANTUM COMPUTING
In addition to semiconductors and semiconductor manufacturing, US–China technology competition has intensified under both the Trump and Biden administrations in two other important sectors that Washington identifies as ‘technologies of the future’: AI and quantum computing. AI-related competition intensified following the publication of China’s ‘National AI Development Plan’ (AIDP) in 2017, which highlighted Beijing’s efforts to develop key technologies. Subsequent plans and strategies, such as the 14th Five-Year Plan and 14th Five-Year Plan for Informatization, have continued to focus on AI and other key technology areas, such as quantum computing.
Following the AIDP’s release, Chinese government support for the AI sector grew with the establishment of industry alliances, an AI Strategic Advisory Committee and preferential policies.35 In addition, many AI start-ups and established companies found ready funding for application development from government security organisations. This was particularly true in relation to facial-recognition and natural-language processing, the latter being the core of ‘perception AI’. A large number of companies have emerged in this field, which have provided an impetus for advances in the sector. These companies include Hikvision, iFlytek, Megvii, SenseTime and Yitu. Other key areas of commercial AI development in China include healthcare, autonomous vehicles, retail applications, education and data science as a service (‘DaaS’).
An important part of the 14th Five-Year Plan was concerned with so-called ‘hard technologies’.36 This included AI and quantum computing as well as semiconductors and advanced manufacturing. In a major speech in October 2021, which was later published as an essay in the CCP journal Qiushi, Xi emphasised the importance of AI, cloud computing, big data and blockchain for China’s future economic growth.37
AI is clearly a priority for Beijing and increasingly the sector is being pulled into the US–China technology competition. Most Western analyses of the AIDP highlight the Chinese government’s role in promoting AI development, including applications relevant to surveillance and internal security, and its potential applications for military command and control. This emphasis has given rise to the narrative that Beijing is intent on using AI as a critical part of its military modernisation, with significant potential security consequences for the Asia-Pacific. However, in reality, it is the Chinese private sector that is responsible for much of the R&D related to AI, driven by foreign and domestic venture capital. Meanwhile, the Chinese government is primarily ‘playing catchup’, supporting rather than leading China’s AI charge.38
The narrative that US–China technological competition includes an ‘AI arms race’39 obscures the reality that US companies currently lead their Chinese counterparts in almost every category of AI development and sectoral application, including what former Microsoft Research Asia chief and venture-capital investor Kai-fu Lee has called ‘internet AI’, ‘business AI’, ‘perception AI’ and ‘autonomous AI’.40 Nevertheless, spurred by its innovative private-sector companies, China is rapidly closing the gap. Yet companies in both China and the US have benefited from the current high levels of collaboration: China is able to leverage open-source and openly published research and is increasingly contributing to cutting-edge R&D, while US companies benefit from being able to recruit large numbers of qualified software engineers in China and from collaboration with Chinese companies to gain exposure to Chinese datasets.
Table 7.5: Chinese emerging-technology companies on the US Commerce Department Entity List and Treasury Department NS-CMIC List
Note: ‘NS-CMIC List’ refers to the Treasury Department’s Non-SDN Chinese Military-Industrial Complex Companies List, where SDN stands for ‘Specially Designated Nationals’
Sources: US, Federal Register, www.federalregister.gov; US, Commerce Department, Bureau of Industry and Security, www.bis.doc.gov; US, Department of the Treasury, home.treasury.gov.
Advanced technologies increasingly pulled into competitionDuring the Trump administration, large Chinese AI companies came under increasing pressure from export controls and later from financial sanctions, having been targeted for selling hardware and software to support public security organs involved in monitoring the Uighur population in China’s Xinjiang province. In October 2019, iFlytek, Hikvision, Megvii, SenseTime and Yitu were added to the Entity List,41 followed in June 2020 by other AI companies including Cloudwalk, Intellifusion and SenseNets (see Table 7.5).42 Under Trump and in December 2021 under the Biden administration, these companies were also added to the Treasury Department’s so-called non-SDN43 Chinese Military-Industrial Complex Companies (NS-CMIC) list,44 barring US investors from holding the securities of the firms and precluding them from stock listings on US markets.
The measures have restricted these companies’ access to US technology, such as semiconductors, but also affected their ability to raise money on capital markets. For example, Megvii postponed a planned listing in Hong Kong following the October 2019 Entity List action but by early 2022 was planning to list on the Shanghai STAR market.45 SenseTime briefly postponed its initial public offering (IPO) in Hong Kong in early January 2021 as a result of the NS-CMIC listing. All the companies on these US lists face reputational challenges and are unlikely to be able to operate globally as a result. Megvii appears to have appealed to the Commerce Department regarding its Entity List action in 2019. It claimed that even before the October 2019 Entity List action it did very little business in Xinjiang and was phasing out remaining support for public security organs there, and had installed an internal AI ethics committee, which reported to the board, that included several external members.46 Megvii’s reported inability to secure a clear answer from US officials about what it would take to be removed from the list, despite apparently taking action to address the underlying reasons it was placed there, suggests a continuing lack of a clear US strategy for listing firms.47 Another result of the growing restrictions on Chinese technology companies’ ability to tap overseas capital markets (and the intensified efforts of US financial regulators to require Chinese firms listed on US markets to undergo third-party audits48) is that the Chinese government is boosting its support for domestic stock markets, notably Shanghai’s STAR and a new bourse in Beijing.
Chinese AI sector under increasing pressureWhile China’s big push into AI has encouraged Western observers to think in terms of a bilateral US–China ‘AI arms race’, the reality is that China’s AI companies face major challenges to further development on several fronts:
Investment in the sector has slowed considerably, and much touted Chinese government funding is squandered in dubious investments. Foreign investors remain concerned about further US pressure on the sector as well as the inability of large players like iFlytek, Megvii and SenseTime to list on US capital markets. Government funding in the form of large pools of capital, such as that announced by the city of Tianjin with much fanfare in 2018, appear to be largely real-estate schemes.49 A close watcher of Chinese AI developments, Jeffery Ding, noted in early 2022 that much state-backed AI investment has been ‘reckless and redundant’,50 with private-sector companies devoting a significant amount of revenue to AI R&D focused on optimising their business operations, but often then experiencing difficulties when trying to integrate AI applications due to lack of sufficient qualified personnel or management understanding of the potential benefit.
Talent recruitment is becoming more difficult and increasing numbers of Chinese AI engineers, particularly the top 10%, are choosing to work outside China. Top-level Chinese AI researchers are on par with their peers in the US, Europe and elsewhere and are able to find work with top-tier AI companies internationally. Many are choosing to leave China, particularly in the wake of increased US sanctions targeting Chinese AI companies, which makes it likely these firms will be shut out of major markets in the US and EU.51 In addition, a 2020 study by MacroPolo, the US-based Paulson Institute’s think tank, found that while about one-third of ‘top-tier’ AI personnel globally has been trained in China, only one-tenth is currently working there.52
IPOs are under pressure as US restrictions remove potential for US listings. The pressure on Chinese AI companies from US restrictions will continue to mean that they will be both unlikely to penetrate markets in the US and other developed countries and unable to contemplate future overseas listing to raise capital. After briefly postponing an IPO in Hong Kong in December 2021 following its NS-CMIC listing, AI giant SenseTime was able to go public in late December, raising some US$740 million.53 It is not clear whether other Chinese AI ‘unicorns’, of which there are many in the pipeline, will be as successful as SenseTime in tapping capital markets to continue their growth.
Access to cutting-edge, AI-optimised hardware may prove more difficult. The Biden administration’s continued use of export-control laws to restrict access to US technology has made it difficult for blacklisted Chinese AI firms to obtain cutting-edge CPU and graphics processing unit (GPU) hardware from US firms. In addition, Washington could increase the use of extraterritorial controls to prevent Chinese companies from developing their own AI-optimised semiconductors (many Chinese AI semiconductor start-ups are using foundry services from TSMC in Taiwan, Samsung in South Korea or US-based Global Foundries).54Due to these challenges, the future of China’s AI development is uncertain and the AIDP’s ambitious targets for 2025 and 2030 are unlikely to be met. Moreover, the decoupling pressure on the sector is likely to continue apace. The coronavirus pandemic has complicated international travel from China, making it difficult, if not impossible, for Chinese AI researchers to travel to the US or other international destinations for conferences. Due to the US China Initiative at the Department of Justice and anti-Asian sentiment in the US stoked by the pandemic (and right-wing media portrayals that blamed China for the coronavirus outbreak), the US has become a more hostile environment for Chinese STEM researchers and professionals. The result has been fewer Chinese AI researchers willing to travel to the US, fewer Chinese students at US universities and more Chinese AI researchers and software engineers returning to work for Chinese companies. The once tightly coupled US and Chinese AI sectors are becoming increasingly decoupled, with unknown consequences. The United States’ National Security Committee on AI (NSCAI), a blue-ribbon committee that included former US government officials and industry leaders, concluded in March 2021: ‘It would be counterproductive to sever the technology ties to China that benefit basic research and U.S. companies.’55 As the decoupling process continues, the risks for Asian security are likely to increase: US and other foreign researchers’ lack of visibility into China’s AI sector will reduce US (and its allies’) understanding of how the sector is developing, how the People’s Liberation Army (PLA) may be benefiting from AI, and how and when major breakthroughs with economic and national-security implications might occur.
A similar dynamic can be seen in other emerging-technology sectors, notably quantum computing. In December 2021, the US Commerce Department put Chinese quantum-computing and quantum-communications firms on the Entity List.56 The companies were not designated because they were involved in developing quantum technology in general terms but rather because they had acquired quantum-related technology that could be used for military purposes. According to the Commerce Department, the listing was intended to prevent ‘U.S. emerging technologies from being used for China’s quantum computing efforts that support military applications, such as counter-stealth and counter-submarine applications, and the ability to break encryption or develop unbreakable encryption’.57
Some of the companies have ties to China’s defence industry and PLA-funded academic organisations. One sanctioned firm, QuantumCTek, supplies equipment and devices for quantum computing, as well as quantum key distribution (QKD) systems used in telecommunications systems to establish encrypted links. The Chinese Academy of Sciences and the University of Science and Technology of China (USTC), plus leading Chinese QKD researcher Pan Jianwei, hold stakes in the company. A USTC-based institution, the Hefei National Laboratory for Physical Sciences at Microscale, was also listed. In late 2021, USTC announced that it was launching China’s first doctoral-degree programme in quantum science and technology,58 while Hefei is also home to the National Laboratory for Quantum Information Science. Quantum information sciences (QIS) are a priority for Beijing, which has a national strategy for quantum and has provided significant funding for projects such as the Hefei facility. Most of the Chinese breakthroughs announced in QIS, such as those relating to QKD and quantum computing, have come from state-backed national projects.59
The Commerce Department listed quantum computing among the technology sectors under consideration for further controls as part of the requirement mandated in the 2018 ECRA.60 However, the US export-control system has been slow to bring quantum-computing technologies under its aegis. In part, this is a result of the industry’s nascent state but also uncertainty about how to bring clarity to concerns specific to national security – especially those related to the dual-use technologies used in quantum computing and QKD systems. In November 2018, the BIS announced an advanced notice for proposed rulemaking (ANPRM), which cited 14 broad areas of emerging technology under consideration for expanded controls, including quantum computing, quantum encryption and quantum sensing. Since then, under the Biden administration the BIS has attempted to narrow the controls to focus on specific enabling equipment, such as quantum-refrigeration units and post-quantum cryptography, software used for nucleic-acid assembly and synthesis, and some specific AI-related software.61
As well as QIS (including quantum computing), biotechnology is another sector under focus – particularly in terms of its military applications and access to healthcare data. The broader risks related to further decoupling in these sectors are similar to those for developments related to AI, as Beijing continues to support advanced technologies and restrict information about domestic R&D in key sectors – such as high-performance computing – while the US government increases its controls and US researchers retract their collaboration. Loss of access and visibility into how these technologies are developed and used in China will make it more challenging for the US and its allies to keep abreast of significant breakthroughs there.
In addition, semiconductors, quantum-related technologies and AI have featured in discussions initiated by the Biden administration in 2021 about expanding multilateral or plurilateral export controls among the US, the EU and the Quad. The formats for these discussions include the EU–US Trade and Technology Council, which has working groups on export controls and supply-chain resilience,62 and the Quad Emerging Technology Working Group,63 which is likely to investigate opportunities for expanding export controls targeting China and coordinating semiconductor-related industrial policies.
RISKS FOR ASIA-PACIFIC SECURITY
If left unchecked, growing pressures driven by the effects of US–China technology competition on semiconductor supply chains and by decoupling in emerging-technology sectors will pose an increasing risk to Asia-Pacific security over the next five years. If the Republican Party sweeps the US mid-term elections in November 2022, hardline political views on tightening export controls and using the FDPR more expansively are likely to gain traction in Washington. If large numbers of Chinese firms are affected by an FDP-style rule – and thereby cut off from using Taiwan as a manufacturing base – the result may be that Beijing feels pushed into a corner on the issue of Taiwan and advanced semiconductor manufacturing. The US would then risk running up against Beijing’s unknown red lines on technology-related issues. Semiconductors are now part of Beijing’s calculus when it considers the timetable for ‘reunification’ with Taiwan. It is unclear what the regional consequences would be if the US attempted to cut off Taiwan’s role as a semiconductor-manufacturing base for Chinese firms. Such a development could reduce the utility of the silicon shield around Taiwan, increasing Beijing’s willingness to discount the impact of a significant disconnection from Taiwan’s semiconductor industry resulting from military action.
In addition, the lack of clear US policy with respect to dual-use-technology controls and restrictions will further complicate the development of robust, resilient semiconductor supply chains in the Asia-Pacific. It could also lead to further economic-security challenges of the kind caused by the global semiconductor shortage since 2020. For example, some industry observers believe that the US Entity List action against Huawei in May 2019 and the FDPR action in May 2020 are responsible for 15–20% of the global semiconductor shortage. These actions were starting to take effect at the same time that the coronavirus pandemic was impacting auto-industry procurement decisions (which were the prime cause of the shortage). These developments led Huawei and then other Chinese companies to stockpile, the impact of which ripped through supply chains and semiconductor brokers, worsening the shortage.64
Without a clear regulatory framework governing civilian-use technologies not traditionally covered by the US export-control regime – including in AI, quantum computing and biotechnology – the US risks taking action with unknown second- and third-order collateral impacts. These impacts could affect global and regional supply chains, as with the Huawei Entity List and FDPR actions. Highlighting the risks relating to future regulatory decisions affecting advanced technologies, in March 2021 the NSCAI made a number of recommendations on export controls, including:
The U.S. Government must clearly state the principles that will guide future U.S. decisions regarding policies to protect critical technologies. This will enable more consistent and cohesive technology protection policies and provide clarity to industry regarding how the government intends to utilize these regulatory tools in the current competitive environment, thereby reducing uncertainty for U.S. businesses. No such framework currently exists.65
US officials could consider tightening controls in the semiconductor sector, pursuing very narrow export controls and seeking to reduce tensions over the issue. They should aim to sharpen the focus on specific national-security implications for the semiconductor industry; often, these implications are not clearly articulated by government officials.66 Similar considerations could be applied to other technology sectors in order to reduce the unforeseen or low-probability (but high-risk) implications for the Asia-Pacific. If current trends continue unchecked, the regional security implications will only grow larger.
No comments:
Post a Comment