China dreams of competing with global superpowers in the semiconductor industry. Whether its efforts will succeed is far from clear.
In 2022, American politicians cannot agree on much. One of the few things they do see eye to eye on is the importance to U.S. national interests of the semiconductor industry, which undergirds technological progress in everything from the humble household robotic vacuum to the most sophisticated military hardware. This consensus has resulted in one of the crowning legislative achievements of Joe Biden’s presidency so far: the CHIPS and Science Act. The law allocates over $50 billion to support the U.S. semiconductor industry, even though the United States already dominates entire industry sectors. Why is America so anxious about chips? The answer is that mastering semiconductor manufacturing has proven both an economic engine and an indispensable strategic asset for the states that have managed to do so. The few who have gained such a position are determined to keep it; the rest covet a spot among their ranks. China, in particular, has made growing a well-rounded domestic industry a national priority, motivated in large part by its techno-economic grappling match with the United States. Whether its efforts will succeed remains to be seen.
Semiconductors — chips — are the foundational components that underlie almost all of the modern electronics industry. They are what allow computers to compute — and in today’s world, almost everything is a computer. Conventional semiconductors are made from silicon, typically refined from sand, and produced by one of the most sophisticated, laborious and globalized value chains of any industry. It is traditionally segmented into three steps: design, manufacturing and packaging. These three steps in turn rely on fancy gadgets galore, built, operated and serviced by initiates of technical arts from software engineering to materials science, and raw materials ranging from highly pure silicon to xenon gas.
Designing a chip requires arranging billions of components on a postmark-sized bit of metal and wrestling against quantum mechanics to get them to work together properly. All those billions of components must be carved out of silicon with lasers accurate enough to hit a pingpong ball on the moon. Chip factories, referred to as foundries or “fabs,” are some of the most fastidiously controlled environments on the planet. The precautions taken to avoid letting the faintest trace of contamination into fabs are rivaled only by those taken to avoid letting pathogens out of high-security biolabs. Staff must wear full protective equipment. Moisture is removed from the air that enters the facility to filter errant particulates; the air is then re-humidified to optimal levels for manufacturing. Coordinating the resources and talent to produce chips at scale is a subtle art not easily mastered or distilled to a repeatable method.
Conveniently for American policymakers, the semiconductor industry happens to be almost entirely dominated by the United States and its close partners and allies in Japan, South Korea, Taiwan and Europe. Most semiconductor design is done by American firms. All three companies offering the specialized software tools used for the process are based in the U.S. Taiwan Semiconductor Manufacturing Company, or TSMC, is responsible for over 90% of manufacturing of the most advanced chips; South Korea makes the rest. The machinery required to make those cutting-edge chips is produced by only one company: the Dutch firm ASML. Though it has lost a leading position in other parts of the industry, Japan today almost entirely controls photoresist processing, a key step in manufacturing, and plays an important role in photoresist materials as well.
For China, depending on its main competitors for critical components is unacceptable. The Chinese Communist Party knows a thing or two about economic leverage, having itself used its dominant position in the processing of rare earth metals to threaten or punish other countries since 2010.
As tensions have risen in recent years, the United States has increasingly moved to restrict China’s access to semiconductor technologies, challenging the CCP’s dreams of technological empire. Beijing’s “no-limits” partner to the north, Russia, has tasted firsthand what it feels like to have semiconductor imports decimated by American sanctions. Allegedly, Russian military equipment contains chips salvaged from dishwashers and refrigerators, while Yale researchers describe the outlook for Russia as “economic oblivion.” China’s dependence on the semiconductor supply chain likely moderates the country’s willingness to escalate international conflicts with the United States and its partners.
Conversely, if China were to achieve an entirely domestic supply chain for all its semiconductor needs, it may be emboldened to throw its weight around on the world stage. Taking the lead in chips would generate leverage that may make other countries even more hesitant to anger Beijing. In short, the business of making sand do math may be the most important civilian industry for the future of China’s “comprehensive national power,” to put it in CCP jargon.
China’s silicon ambitions — and struggles — are not new. By the 1970s, Mao Zedong was gone and the tumult of the Cultural Revolution had ended, but the country was in shambles. In 1978, Deng Xiaoping’s top priority as newly instated leader of the CCP was development, and development required technology. China launched several state-led semiconductor projects between the mid-1980s and the mid-1990s, but they were hampered by administrative delays, while international progress forged ahead. When the IT bubble popped in 2001, some international firms cut their losses in the Chinese market, leaving their partners to wither away. Piecemeal investment from domestic firms and government created factories that were outdated almost as soon as they got up and running. Perhaps worst of all, the generation of engineers who should have led the industry in the ’80s and ’90s had spent their formative years doing hard labor instead of training in their field — if they had survived the Red Guards’ persecution at all. By the turn of the 20th century, China was left with little to show for its efforts.
With these projects, CCP leaders had been hoping to follow a script that had proven successful for its neighbors. Japan, having clawed its way into the chip industry through state-led industrial policy during its postwar boom, was a model.
By the 1980s, American chipmakers were bemoaning the loss of the memory market to Japanese shops whose process innovations had cut defect rates, minimizing waste and boosting profitability. Other of East Asia’s burgeoning economic powerhouses had begun to get in on the action as well. In South Korea, support from both government and business conglomerates helped the semiconductor industry grow at breakneck speed in the mid-1990s. In 1987, Morris Chang (張忠謀 Zhāng Zhōngmóu), then head of Taiwan’s Industrial Technology Research Institute, founded Taiwan Semiconductor Manufacturing Company with 21% of the company’s startup capital provided by the Taiwanese government. But besides avoiding the autolobotomous chaos of the Cultural Revolution, these countries had another advantage China lacked: decades of commercial and intellectual trade with the United States, where many engineers had been trained at cutting-edge academic and industrial labs before returning to build suppliers, customers and competitors to their former employers in their home countries. Tacit knowledge developed through hands-on experience has been crucial for success in the semiconductor industry.
Over 20 years, Deng’s policies gradually replenished China’s pool of technical talent. Reform and opening revived China’s educational institutions such as Tsinghua University, which would become a key crucible of Chinese semiconductor talent. It also meant that Chinese students began to flow into classrooms at places like MIT and Berkeley, and later into the labs where alumni of such schools often end up. As China’s economic growth took off in the early 2000s after the country was admitted to the World Trade Organization, many Chinese educated abroad took a chance and returned to China, forming a great migration of “sea turtles.”
This new generation of world-wise leaders had the know-how to establish themselves in the industry; some founded firms that would later become major players. However, the government support that had midwifed semiconductor industries around the world was not forthcoming. Through the first decade of the new millennium, bureaucrats had a bitter taste in their mouth from the failures of large-scale funding to achieve results in attempts of the previous decades.
All this changed when Xi Jinping came to power in 2013. Unlike his predecessors, who experimented with free market reforms, Xi has moved to reassert strong central control over the Chinese economy. Most notably, this has taken the form of the now-infamous Made in China 2025 industrial strategy, a trillion-dollar initiative to upgrade China’s high-tech manufacturing capacity. Xi has also spoken at length about the importance of the “real economy” (实体经济 shítǐ jīngjì) — solid, utilitarian sectors like manufacturing and infrastructure rather than social media, financial speculation or video games. It is no surprise that the CCP’s enthusiasm for industrial policy came to semiconductors. In 2014, the Chinese government laid out an ambitious set of goals to develop a world-leading semiconductor industry by 2030. The same year, the government established the China Integrated Circuit Industry Investment Fund (国家集成电路产业投资基金 Guójiā Jíchéng Diànlù Chǎnyè Tóuzī Jījīn),, more often referred to simply as the “Big Fund,” targeting $19 billion of direct investment over five years. Semiconductors are both the most critical sector of the real economy for China’s resilience and the apotheosis of Xi’s economic aesthetics: high technology, tangible objects, and strategic leverage.
Policy support under Xi has focused on remedying China’s weaknesses in the technically demanding, high-margin sectors of design and manufacturing. One of the country’s greatest successes is its analog of TSMC: Semiconductor Manufacturing International Corporation. TSMC disrupted the industry in the 1990s with the “pure-play foundry model,” focusing exclusively on manufacturing over design. SMIC is another pure-play foundry. Like TSMC, it was founded by an American-educated returnee with a decades-long career at the U.S. semiconductor firm Texas Instruments. In July 2022, reverse-engineering firm TechInsights reported that a chip produced by SMIC had been made with technology just one generation behind the commercial state of the art — a significant accomplishment even if not yet indicative of a full commercial-scale production capability.
As semiconductor technology has advanced, the ranks of foundries capable of producing the smallest, most sophisticated chips have thinned from dozens in the late 1990s to a mere trio today: TSMC, Samsung and Intel. Although SMIC will need to run even faster to catch up with, much less overtake, the leaders, maintaining a fast-follower position is impressive in its own right.
Although American firms such as NVIDIA and Intel dominate design, China has managed to cultivate a number of firms capable of dreaming up advanced chips. HiSilicon, a subsidiary of Chinese telecoms giant Huawei, produced logic chips on TSMC’s 5-nanometer production node before falling under U.S. sanctions.
A swath of Chinese firms aim to challenge NVIDIA’s vise grip on GPUs, with some producing designs only a few generations behind NVIDIA’s. In the somewhat less challenging memory market, China has produced strong players such as Yangtze Memory Technologies Corp, which grew from 1% of global market share at the start of 2020 to 5% as of summer 2022, and secured a deal to supply memory chips for the newest iPhone.
China is strongest in the less sexy but no less critical final link of the semiconductor value chain: assembly, testing and packaging (abbreviated collectively as ATP). ATP generally requires a lower level of skills, more akin to the low-end manufacturing and device assembly work that has been China’s bread and butter for decades. It also garners lower margins. Nevertheless, it is a critical sector in which China made up almost 40% of the global market as of 2021.
Furthermore, advanced packaging is an area of active innovation in the industry where China may be well-positioned to lead if it is so inclined. Though it has to date employed primarily low-skill labor, greater complication and precision required by fancy packaging methods such as three-dimensional stacking means the industry will likely move toward heavier use of industrial robots. As the largest consumer of industrial robots for almost a decade running, China could potentially parlay its robotic manufacturing capacities into a significant moat.
Chip mania reached a fever pitch throughout the Chinese economy by 2019, following a set of new tariffs introduced by the Trump administration and the subsequent realization that U.S.-China trade tensions would neither go away nor be quickly and easily resolved through a trade deal. The second tranche of investment from the Big Fund was announced in October of that year, promising another $20 billion focused on equipment and upstream materials. Tellingly, Xi also placed top economics official and close confidant Liu He, previously responsible for China’s end of the U.S.-China trade war, in charge of semiconductors policy in June 2021. Giving semiconductor manufacturing a high position among policy priorities also means the party is more willing to pay the political costs of encouraging industrial espionage, protectionist policies and other measures that may benefit the industry.
As is characteristic in China’s highly policy-driven market, private investment has stampeded to position itself in this “wind tunnel” of opportunity.
According to Chinese state sources, over 58,000 new chip firms were registered in China between January and October 2020 alone. Some firms signal their red bona fides with patriotic marketing decisions: one advertises a “Self-Strengthening” line, which uses only Chinese domestic components. Nearly every major Chinese tech company has made a foray into the sector, no matter how far their core consumer-facing software business might be from the primarily B2B computer hardware market. Even TikTok’s parent company, ByteDance, has thrown its hat into the ring. In some cases, designing hardware that can be co-optimized with the rest of a company’s technology stack can provide real benefits; in others, these firms may be hoping to cash in on government incentives, or simply signaling their piety in following the CCP’s leadership.
China has a few natural advantages to give all these new companies a head start. As in many domains, one is its sheer size. To remain in the lead, chip companies must spend massively on R&D, shoveling up to 15%-30% of revenue back into the engine of technological progress.
China’s uniquely large domestic market means firms can establish a flywheel of sales to R&D to more sales without needing to venture into other markets. In recent years, China has been the world’s second-largest consumer market for semiconductors, comprising a quarter of global demand in 2019, about on par with the United States. In addition to demand, China has a major strength in supply — of labor, at least. China is projected to graduate over 77,000 STEM Ph.D.s in 2025, compared to only 40,000 in the United States. Chinese universities have established a swath of semiconductor-related degree programs over the last years to train more specialists in the arcane arts of building thinking machines.
But while the inputs to the industry driven by state support are impressive, it is far from guaranteed that their outputs will be equally so. Mandarin Chinese offers a pun to encapsulate a major issue for the industry: The second character in the word for “chips,” xīnpiàn, can be swapped for a homophone 骗 piàn meaning “to swindle,” creating a neologism 芯骗 xīnpiàn, “chip cheating.” Of the legions of “semiconductor” firms registered in recent years, it is unclear not only how many will succeed, but how many are even contributing to the industry in any meaningful sense. The story of one of China’s older companies, Wuhan Hongxin Semiconductor Manufacturing Company, provides a case in point.
Sometimes referred to as China’s “semiconductor Theranos,” HSMC was a scam led by a career con man and two accomplices with no technology expertise who succeeded in roping in the founding CTO of TSMC, acquiring a top-of-the-line lithography machine, and raising billions of dollars before plundering the firm’s assets and making off scot-free. Even the Big Fund itself has been beset by graft scandals, with at least four current or former senior executives involved in managing the fund coming under investigation since July 2022.
American policymakers, recognizing the geopolitical strategic importance of semiconductors, have started to take action to stymie China’s efforts. They have levied or threatened sanctions on multiple Chinese semiconductor players, including Huawei, ZTE and YMTC. They have also made use of the choke points within the industry. Dutch firm ASML is the sole manufacturer of the extreme ultraviolet lithography equipment necessary to make the most advanced chips, and it has already agreed to American entreaties not to sell its most advanced kit to China. In recent months, Washington has introduced a bevy of new measures to hamstring China’s semiconductor industry, including restricting exports of even older semiconductor manufacturing equipment, banning American companies from servicing equipment in China, requiring export licenses to ship top-shelf GPUs to China and planning to block certain Chinese firms and research labs from purchasing anything made with a drop of US technology. These new policies are designed to restrict the progress of China’s industry past the generation of technology it has already mastered, locking the country out of high-end chips for applications like AI and supercomputing entirely.
Beyond Beijing, a similar chip mania has infected governments around the world whose ambitions in the space further threaten to crowd out China’s place in the silicon sun. In the United States, the CHIPS Act has finally passed, with measures to bolster the United States’ strategic position in the industry, particularly in leading-edge fabrication. Taiwan, South Korea, Japan and Europe have all explored new policy support for the industry as well. The leading firms have geared up for the melee too. TSMC announced in January 2022 it would spend $100 billion over three years. Samsung funneled over $40 billion into semiconductors in 2021 alone. Meanwhile, the United States continues to dominate in R&D, making up more than half of total industry spending. What was already a competitive market may become even more cutthroat. China will not find itself uncontested in chasing the various prizes that await those who solve the engineering puzzles holding back progress today.
Flows of technical knowledge into China are also being constricted. As University of Toronto professor Jun Zhang put it in a recent talk: “To overtake the leaders, you have to learn from them. But will they give you opportunities to learn from them?” China’s success to date has been catalyzed by knowledge and experience gained from foreign industry leaders. They are no longer so generous in offering such education. Taiwan’s government has introduced anti-poaching regulations to cut down on the trend of Taiwanese semiconductor engineers being lured to the People’s Republic, where salaries may be multiples of their compensation on the island. Investment flows from China into global semiconductor firms have met with increasing resistance since the Trump administration blocked a Chinese private equity firm from acquiring Lattice Semiconductor in 2017. Now even investments into China are garnering closer scrutiny as avenues for leakage of technical know-how. All but the most oblivious international semiconductor firms recognize they are targets of Chinese state-affiliated hacking groups and are actively hardening their defenses against espionage.
If China’s plans are to succeed, they will do so through the assiduous efforts of a motivated, masterful semiconductor workforce. But so far, “hard tech” sectors have struggled to attract talent. The Chinese tech industry is infamous for its grueling conditions, often summarized as “996” — 9 a.m. to 9 p.m., 6 days a week. An entire lexicon has emerged to describe the crushingly competitive professional lives of urban white-collar workers. The dominant concept is “involution,” (内卷 nèijuǎn), a race of ever-increasing competition over finite resources that leaves all participants burnt-out and despondent. “Lying flat” (躺平 tǎngpíng) has become a rallying cry of slackers. The working masses in China sometimes refer to themselves as “chives” (韭菜 jiǔcài), growing tall only to be harvested as cheap labor by the combine of state-led capitalism. The semiconductor industry, despite its pride of place in the CCP’s eye, is far from exempt from these dynamics. The training in electrical engineering necessary to contribute in the industry — generally considered to require a Ph.D. and significant time gaining hands-on experience — is a heavier burden than the comparatively relaxed investment of time needed to break into software, where salaries are similarly high and no one needs to spend their days wrapped up in PPE in a clean room. Why, some might wonder, go through all the effort if you will only be chopped up with the rest of the chives?
Optimistically, China’s flood of semiconductor investment could cultivate a pool of human capital that will form the foundation for the industry’s eventual success. However, the country may equally just waste the time of technical talents who could otherwise be pursuing more economically efficient, market-driven opportunities. In flushing the intensely competitive and challenging manufacturing sector full of cash, the Big Fund’s Phase I investments may be fueling involution in chipmaking while failing to capitalize on opportunities for innovation elsewhere, such as in advanced packaging.
Meanwhile, the Chinese economic engine is becoming creakier as the population ages and the debt-fueled infrastructure investment that powered growth for decades runs into diminishing returns. Official state estimates forecast population growth to peak in 2031, but some researchers believe the birth rate has already begun to decline, and may in fact have peaked as early as 1991. Concerning signs have surfaced in property markets, a critical part of China’s financial stability. Strict zero-COVID policies pose a dilemma for the CCP: They incur high economic and public opinion costs, but lifting them risks an outbreak, which could itself also cause economic havoc.
Xi Jinping, whose personal influence over Chinese policy is perhaps greater than any leader since Mao Zedong, lacks a strong background in economics and is suspected to make decisions based more on his expertise in domestic politics and foreign policy. Though the People’s Republic has many competent technocrats, Xi’s word may override the wisest counsel.
Success in semiconductors would be not only a source of national pride for the Chinese Communist Party, but a major shift in the global distribution of techno-economic power.
Some think he already has with the new wide-ranging package of economic and social reforms dubbed the “Red New Deal.” The hammer has come down particularly hard on Chinese consumer tech companies, which have been subject to anti-monopoly measures and other new restrictive regulations. While some of these policies appear commonsensically technocratic, the party’s recent appetite for bold action has shown itself most clearly in policies related to children, including banning after-school tutoring services (previously a boisterous market) from turning a profit, and limiting minors to only three hours of video game play per week — Friday, Saturday and Sunday from 8 p.m. to 9 p.m. While intended to promote “common prosperity,” these policies may end up hamstringing the private sector dynamism responsible for almost all China’s economic growth. The Chinese government in late August announced a large stimulus package, but even this was not enough to buoy economists’ and investors’ hopes for the Chinese economy. The future is far from certain for the Chinese economic juggernaut.
What, then, should the savvy observer expect from China’s chip dream? The country hoped to source 40% of semiconductors domestically in 2020. Ultimately, it reached 16%. We might conclude that today’s campaign is as doomed as in the past. The semiconductor industry today requires even more exacting technical sophistication. But until quite recently China remained poor; today’s China is a vastly more formidable technological competitor. American efforts to cut off China’s semiconductor industry from global supply chains may even backfire and help it to grow strong and independent.
Most likely, if trends continue, the country will achieve a middling outcome — making a respectable showing in older technology, perhaps eking out enough progress to inch forward a few generations, and likely achieving a handful of major technical breakthroughs.
Yet, there is one other dynamic that may provide China with surprising opportunities. Moore’s law, the self-fulfilling prophecy of predictable progress in shrinking transistors to make chips faster and more efficient, has guided the industry for over half a century. It is now coming to an end as the conventional technical paradigm runs into fundamental physical limits.
Scientists have proposed and prototyped alternative architectures for computing hardware, each with its own advantages and constraints. Many industry observers expect the future of computing to see heterogeneous components stitched together in tailored combinations to suit specific cases, rather than the handful of general-purpose architectures ubiquitous today. We should expect that China may succeed in at least a few of these directions. The resulting strategic landscape may look very different depending on which work out and which do not. If, for instance, China pulls ahead in quantum computing, where it is relatively strong, but cannot produce ultrawide bandgap semiconductors, recently placed under export control by the U.S. Commerce Department, Beijing may find itself able to crack the toughest encryption in the world but struggling to field satellites with spyware to match those of its competitors. If the United States and its partners continue to tighten the screws, the number of such successes may be very limited. On the other hand, even one major success could give China a powerful position if every international firm is clamoring to use, for example, Chinese packaging processes, much as TSMC commands great influence through its near monopoly on the most advanced production methods.
Success in semiconductors would be not only a source of national pride for the Chinese Communist Party, but a major shift in the global distribution of techno-economic power. It would be positive evidence of a strength of the Chinese model and a reinforcing buttress for it in the future. Failure, on the other hand, could lead to stagnation.
The absolute minimum purity needed for electronics-grade silicon is 99.9999999%, or one nonsilicon atom to every billion silicon atoms. Some producers offer silicon that is a full two orders of magnitude purer still, with a single nonsilicon atom to every 100 billion silicon atoms.
Chinese state-affiliated hacking groups have also recently attempted to undermine American attempts to build a domestic rare earths supply chain through a disinformation campaign against a planned rare earths mining facility in Texas.
Japan’s Ministry of International Trade and Industry, which directed much of Japanese industrial policy through the country’s economic miracle, earned the epithet “the notorious MITI” from foreign commentators and inspired the term “Japan, Inc.” due to its powerful influence in Japan’s economy.
海龟 hǎiguī, literally “sea turtle,” is a pun on the phrase “overseas returnee,” 海归, also pronounced hǎiguī.
The chip was reportedly “a near duplicate of TSMC 7-nanometer process technology,” leading some to speculate that industrial espionage may have played a role in the achievement.
UNISOC has since become the leading Chinese designer of smartphone processors, with products on TSMC’s 6-nanometer node.
China is not unique in using tactics that violate the rules and norms of international trade to build its chip industry; Japan did many of the same things in the 1980s as it strove for semiconductor success.
Fēngkǒu, literally a gap or hole through which wind flows, is a term often used in the Chinese business world to refer to promising sectors where favorable economic “winds,” often policy-generated, will blow in the sails of hopeful firms.
By comparison, R&D spending as a fraction of revenue in sectors such as energy or consumer products can be as low as 1% or even 0.1%.
Also abbreviated as HSMC, the company’s name 弘芯 Hóngxīn, “grand chips” or “promoting chips,” patriotically echoes the term “red heart,” 红心 hóngxīn, used to describe passion, especially for the Communist Party.
This includes Ding Wenwu, Wang Wenzhong, Lu Jun and Yang Zhengfan.
Prospects are better for older generations of semiconductors, which have not been subject to the same restrictions and are still crucially important for a panoply of simpler electronics.
China has to date refused to accept the more-effective American and European mRNA vaccines and has a low rate of vaccination even with its less-effective domestic vaccines. Health care capacity is also highly unequally distributed in the country, meaning the system could easily become overloaded in rural areas.
In particular, cutting off exports of chips, rather than the tools and materials to make them, opens up space in the domestic market for Chinese chip firms that would previously have been dominated by the NVIDIAs and AMDs of the world, helping them invest in the R&D necessary to get to the cutting edge. The U.S. recently imposed export license requirements on top-shelf GPUs, which could have exactly this effect.
The aforementioned field of advanced packaging, which includes methods such as simply stacking chips on top of one another, is one way the field may yet continue to squeeze more performance out of conventional silicon.
Published November 2022
Karson Elmgren is a Research Analyst at Georgetown’s Center for Security and Emerging Technology (CSET), where he works on the AI Assessment team.
By highlighting text and “starring” your selection, you can create a personal marker to a passage.
What you save is stored only on your specific browser locally, and is never sent to the server. Other visitors will not see your highlights, and you will not see your previously saved highlights when visiting the site through a different browser.
To add a highlight: after selecting a passage, click the star . It will add a quick-access bookmark.
To remove a highlight: after hovering over a previously saved highlight, click the cross . It will remove the bookmark.
To remove all saved highlights throughout the site, you can click here to completely clear your cache. All selections have been cleared.