The Purges That Upended China’s Semiconductor Industry

 

The Purges That Upended China’s Semiconductor Industry

 

by Geoffrey Cain

The American Affairs, Winter 2022 / Volume VI, Number 4

 

Once a technology star, Zhao Weiguo rose fast and fell hard. For the last eight years, Zhao’s semiconductor manufacturer, the Tsinghua Unigroup, had fanfare, ambition, large-scale state backing, and an affiliation with China’s most prestigious institution of higher learn­ing, Tsinghua University. All this made Zhao the face and future of China’s semiconductor industry. Frequently appearing alongside Presi­dent Xi Jinping, he helped convince the nation that China would rise as a technological superpower that could rival American hegemony.

 

In 2014, the Chinese government accelerated its semiconductor industrial policy with the formation of a massive investment consortium, the National Integrated Circuit Industry Investment Fund, or the “Big Fund.” Zhao, whose childhood was spent herding goats and who made an early fortune in coal and real estate, was suddenly leading a national champion. The consortium gave him a blank check to transform Tsinghua Unigroup from an obscure seller of herbal medicine drinks and computer scanners into a semiconductor maker that the government hoped would rival Intel, AMD, and Samsung.

 

“When your heart isn’t strong, no matter how big you are, you’re not really strong,” President Xi told semiconductor manufacturers, ac­cording to state media. He often called these core technologies the “heart” of China’s industrial efforts. In 2015, Xi declared the goal of “Made in China 2025,” demanding that China become a leading pro­ducer in self-driving cars, microchips, and automation technologies.

 

With state largesse, Zhao’s Tsinghua Unigroup began talks to acquire the Idaho-based Micron Technologies, then the second-largest maker of dynamic random access memory (DRAM) chips—a low-cost technology used in computers and graphics cards—for $23 billion. The bid failed, but Zhao scored a victory when Intel pledged $1.5 billion for a 20 percent stake in Tsinghua Unigroup.

 

Awash in cheap capital, Tsinghua Unigroup undertook relentless expansion. In May 2015, it acquired a Chinese data networking com­pany called H3C, previously owned by Hewlett-Packard, for $2.3 billion. The next year, it partnered with Western Digital, a California-based hard drive company, to set up a joint venture for big data storage equipment in China. Tsinghua Unigroup then founded a memory sub­sidiary, Yangtze Memory Technologies, along with two new memory chip factories. It also bought a 6 percent stake in an Oregon-based chip company, Lattice Semiconductor, and built a $1.9 billion plant that partially opened in 2020.

 

Added up, these ventures cost untold billions, with the full numbers still unavailable to the public. Zhao earned the nickname “semiconduc­tor madman” for his indulgent and reckless spending sprees. Publicly, Zhao denied running a government front. “Many people suspect I’m a ‘white glove’ for the government,” Zhao told the Wall Street Journal in 2015. “But we’re really just a very market-oriented company.” In reality, he was following an old playbook for East Asian industrialism, pioneered by Japan, South Korea, and Taiwan, though applied most aggressively by China in recent decades. The Chinese playbook called for rapid expansion, acquisitions, industrial espionage, and intellectual property theft under the euphemism “technology transfer.” If it suc­ceeded, advanced semiconductor companies in China would overtake American and East Asian competitors and put some of them out of business.

 

Within six years, however, Zhao saw his aspirations fall apart. Tsinghua Unigroup has hemorrhaged money and resources in a desper­ate attempt to fulfill President Xi’s grand ambitions. In July of this year, Chinese authorities placed Zhao under arrest for what local media called “procurement irregularities.” At least eleven other semiconductor exec­utives, including two from Tsinghua Unigroup, also disappeared from public view. Chinese authorities accused Zhao of large-scale mis­manage­ment and billions of dollars in losses. President Xi, it seemed, was unhappy with the slow progress. And China’s once-vaunted ambitions to build a self-sufficient semiconductor industry were in danger of collapse.

 

Considering China’s remarkable successes in critical industries such as rare earth mining, software, and artificial intelligence, this failure of industrial strategy is a puzzling anomaly. With state subsidies and government intervention, China has overtaken the United States and other powers as the largest producer of steel, automobiles, machine tools, electrical machinery, and computers. In 2019, Huawei over­took Apple as the world’s second-largest smartphone maker—second only to South Korea’s Samsung, the largest producer by volume—and is a global leader in server equipment.

Everywhere we look, China’s industrial policies have delivered a strong position in key sectors and supply chains. But not so in advanced semiconductors, which are complicated devices with expertise centered in a handful of countries outside China. Huawei might be able to catch up to Samsung in smartphone sales by volume. But unlike Samsung, it will continue to struggle building leading-edge semiconductors with enough processing power for AI and other novel applications.

 

Understanding these failures will help America and its allies navigate the renewal of American industrial planning, now begun in earnest with the chips and Science Act and other legislation. Why has China failed after a decade of intensive and lavish state backing, while other countries have built successful semiconductor industries in roughly the same amount of time?

 

The biggest reasons for failure don’t come from the challenges of developing the technology but from China’s missteps in foreign and industrial policy. Collapse was brought about by China’s diplomatic isolation, an oppressive top-down mandate of selecting national champi­ons, and the weak position of starting generations behind industry lead­ers in America, Taiwan, South Korea, and Japan.

 

Global Isolation in the
Globalized Semiconductor Industry

On October 7, the Biden administration announced wide-ranging sanc­tions that crippled China’s semiconductor industry for guided weapons, supercomputing, and other critical military technologies. Companies are now required to apply for licenses before selling advanced chipmaking equipment, components, and semiconductors, and most of these licenses will be denied. Social media reports from inside China suggest that the sanctions incited chaos overnight, with mass resignations of American chip executives based in China.

 

As of this writing, the sanctions were the latest salvo against a country that made the strategic error of isolating itself diplomatically while it has rushed into the fragile, complex, global semiconductor industry, in which success rests on decades-long alliances and ironclad supply chains. Ever since President Xi rose to power in 2012, China has embarked on an aggressive foreign policy abroad and alarming levels of repression at home.

 

The government has issued threats, in the form of close-flying fighter aircraft and naval incursions, against Taiwan, Japan, and in the South China Sea, harassing ships from the Philippines, Indonesia, and Vietnam. It has put down pro-democracy protests in Hong Kong, which is sup­posed to be a separate “special administrative zone” under an agreement with the United Kingdom, but is now effectively annexed with national security laws that outlaw criticism of the state. It has detained 1.8 million Uighurs and other Muslim ethnic minorities in a network of concentration camps, building a technological surveillance state un­precedented in scale and sophistication.

 

All this points to a devastating miscalculation. President Xi believed that China could somehow become a self-sufficient chipset maker while he angered all the leading semiconductor-producing countries. In 2019, the United States and EU began imposing sanctions and export controls on China, ensuring that snooping Huawei servers and Hikvision camer­as did not make it to Western shores, and that Western semiconductors and equipment did not fuel China’s military and surveillance state, either.

 

As with everything in the secretive CCP, whatever President Xi discussed privately with his advisers and how these internal conversations shaped his decisions remain mysteries. But from his speeches and written works, an image emerges of President Xi as a leader with an irrational belief in patriotic spirit as the most important driver of technological innovation. Though nationalism can inspire innovation efforts, he was wrong overall about the complexities of global supply chains and the need to secure cooperation from foreign powers. Semi­conductor mastery has always required a period of copying, imitation, and catching up with outside help.

 

Every prosperous East Asian country has demonstrated, with remarkably consistent results, the importance of carefully managing international relations in addition to domestic industrial support. South Korea is just one remarkable case study: the government leveraged its military alliance with America and postcolonial relationship with Japan to kickstart the industry—two diplomatic gateways absent in China.

 

In the late 1970s, South Korea, like China more recently, was an impoverished and oppressive authoritarian state that exported steel, garments, and petrochemicals. The country built its semi­conductor industry in the early 1980s, after President Chun Doo-hwan, a military dictator, seized power and directed giant family-run conglomerates Samsung and Hyundai, called chaebols, to compete in manufacturing basic DRAM semiconductors, used in calculators and early personal computers. The goal was to eventually catch up with Japan and the United States. Chaebol groups that failed to meet the government’s demanding export quotas lost their state funding and were later forced to sell their semiconductor businesses.

 

South Korea’s authoritarian governments were skilled at exploiting their strategic geographic position bordering hostile countries North Korea and China. South Korea was indispensable to America’s Cold War alliances and benefited from American largesse. South Korea’s leaders secured American protection in the form of military bases, development assistance, and access to the U.S. market, all of which allowed the chaebols to develop homegrown technologies.

 

The other key piece was Japan. In the 1970s, South Korea seized on newfound trade relations with its former enemy and colonizer, raising capital and importing technology from Japanese investors who needed cheap labor. Chaebol groups supplied basic components in exchange for technology transfers and, somewhat underhandedly, invited Japanese engi­neers to moonlight on weekends, smuggling technological secrets and selling prized chip manufacturing equipment.

 

These days, China, heavily sanctioned, has lost the diplomatic access that would have propelled a homegrown industry.

 

But like South Korea, Chinese companies have turned to industrial espionage and infiltration of competitors. For the past decade, South Korea and Taiwan have seen increasing arrests and court cases against people accused of stealing semiconductor secrets. In April, Taiwan opened investigations into one hundred Chinese companies suspected of illegally poaching Taiwanese semiconductor talent, after successfully prosecuting seven companies in 2021.

 

But industrial espionage has proven not nearly enough to make up for China’s dwindling foreign access. Global supply chains are capital-intensive, multi-decade projects, many of which continue to rest on Cold War relationships. Today, semiconductor manufacturing mostly takes place in Taiwan, home to TSMC, and South Korea. The most advanced semiconductor equipment comes from the Netherlands, where the firm ASML controls the machinery that packs the largest amounts of circuitry onto chips, elevating their processing power. In 2019, the Netherlands joined U.S.-led sanctions and blocked shipments to China.

 

The United States still leads the world in semiconductor design with Intel, Nvidia, and AMD. But no matter how much America tries to reshore the other industry pillars—the manufacturing and the equipment—the global reality rules out self-sufficiency. The United States needs to avoid China’s mistake of alienating its partners and allies.

 

While China’s ambitions appear to extend to self-sufficiency and global dominance, the United States can be content to pursue a less antagonistic goal of diversifying manufacturing and supply chains. Though American and Taiwanese firms may still find themselves at odds, this is a small price to pay to ensure supply chains remain spread out and safe from geopolitical or military meddling by China. Rebuilding some domestic manufacturing capacity need not entail concentrating the entire supply chain in the United States, a project which would take decades and involve a high chance of failure anyway.

 

The Pitfalls of the Command Economy

Another reason for China’s failure is the country’s centralized, top-down approach of choosing national champions. In a diverse and often provincially managed nation of 1.2 billion people, China’s command approach has not been able to achieve the success of smaller neigh­bors such as South Korea and Taiwan, tadpoles in size and population next to China. In particular, China’s heavy-handed efforts to correct for an unbalanced labor market, including slave labor, have harmed its semiconductor ambitions.

 

It is true that the semiconductor industry is reliant upon state support. Here is a common scenario that guarantees the failure of almost every young and undercapitalized chip company: Advanced semiconductor firms typically set their next chipset lines three or more years in advance, the amount of time needed to build up R&D and intensive capital investment. The market, however, moves in dramatic fluctuations that are impossible to predict. By the time a chipset maker manufactures and sells the next generation of chips, many other competitors with state funding will have converged on the same market. With swelling supply and cutthroat price wars, all those years of nose-to-the-grindstone design and manufacturing will prove a disaster—unless the state can subsidize a firm’s losses. Because of these inherent risks, semiconductor manufacturing isn’t the realm of Silicon Valley start-ups and VCs. It favors large, legacy enterprises with the scale to absorb shocks and the political influence to get state funding.

 

The centralizing forces of the semiconductor industry worked well in smaller, rapidly developing nations from the 1960s to 1990s. National governments in Japan, South Korea, and Taiwan oversaw workforces that were young, socially homogenous, and concentrated in a few technological or population centers. South Korea manufactured most chips in the cities of Suwon, Giheung, and Cheongju. Taiwan maintains facilities in Hsingchu, Tainan, and Taichung. Traditionally, American design and manufacturing were weighted toward the West and Southwest, along with a few fabs in the Northeast.

 

China, however, has failed to achieve anything near a healthy, regionally balanced labor market. It suffers from demographic and labor imbalances that are damaging to its long-term prospects. Technology industries are concentrated in a few superstar cities, notably Beijing, Shanghai, and Shenzhen, which are home to disproportionate numbers of fab plants. With affluence, overcrowding, and overdevelopment along China’s eastern coast, and a wrenching and rural underbelly further west, China has attempted to correct disparities in its usual authoritarian fashion: by enlisting slave labor.

 

China has brought in forced labor from the minority Uighur and Kazakh groups from the oppressed Xinjiang region, who toil away making polysilicon, a material used in solar panels and semiconductors. The Semiconductor Industry Association believes that the polysilicon made with forced labor doesn’t meet the standards required for America’s semiconductors. Luckily, this polysilicon does not appear to be at risk of infecting our supply chains.

 

Of course, American and Taiwanese firms would not enlist slave labor. But the gaping and inhumane inequalities in China’s technological ecosystem teach a lesson: that deeply uneven regional development makes it impossible to manufacture semiconductors and components at cutting-edge, global standards. The regional workforce, lacking educa­tion, empowerment, and a viable future, can’t keep up.

 

Though America’s regions are nowhere near as impoverished as China’s, the United States must not repeat China’s mistakes. Historically, it has concentrated technological development in Silicon Valley, but it should diversify semiconductors in ways that benefit different regions and enrich local populations. On October 3, the U.S. Department of Commerce made community building a strategy for implementation of the chips Act. It wrote that it will “grow a diverse semiconductor workforce and build strong communities that participate in the pros­perity of the semiconductor industry.”

 

What will this community building look like within the semiconductor industry? Semiconductor facilities, first and foremost, should be clustered around university towns. In January 2022, for instance, Intel announced a $20 billion initial investment in two new leading-edge fab plants and $100 million towards educational partnerships just outside Columbus, Ohio—the location of Ohio State University, a major engineering and sciences school. This is a sound strategy. The Wall Street Journal reports that the United States will need to add ninety thousand semiconductor workers by 2025 to obtain its manufacturing goals. This builds on an American regional strategy dating to the Cold War, when local innovation hubs developed in Durham, North Carolina, near Duke University; Madison, Wisconsin, next to the University of Wisconsin; and Phoenix, Arizona, near Arizona State University.

 

Technological Lag and the Failure to Catch Up

A further reason for China’s dim prospects is that the country has started out generations behind the leading semiconductor designers and manufacturers. Though South Korea and Taiwan once started out far behind Japan and America, they had favorable access to technology, corporate partnerships, and export markets that allowed them to pro­gress rapidly.

 

China, at times, has seemed to follow this playbook, but remains at the low end of the sector. In September 2022, Chinese state media reported that Yangtze Memory Technologies (YMT), one of China’s largest chip companies, was on the path to self-sufficiency, with oppor­tunities to supply Apple with its latest NAND flash memory chips. YMT has never confirmed the reports. Apple uses NAND flash memory in the iPhone because it is resistant to shocks and falls. Apple, reportedly, has also been diversifying its supply chain beyond usual suppliers in South Korea, as a hedge against supply-chain instability.

 

Despite being lauded by China’s state media, this is one example of an industry nonevent. China remains generations behind other producers because it focuses on old-school commodity memory chips. These chips, which include DRAMs and NAND flash memories, require smaller investments and far less risk. Chinese engineers can rely on the engineering that came before them—in South Korea, Taiwan, and the United States—and attempt to imitate existing technologies.

 

With the recent scandals and purges, it’s unclear if China has the stomach or ability to advance toward the high-risk, high-investment advanced industries, in the non-memory semiconductors, for instance, that power artificial intelligence software. In 2019, China’s share of global semiconductor sales surpassed Taiwan’s to become number one, at 35 percent of the world’s production. These numbers can be deceptive, however. China remains a net importer, despite holding the top global market share, because it cannot produce its own advanced chips, and shows no momentum toward doing so.

 

The Failure of China’s Semiconductor Strategy

China’s industrial strategy—the product of vast differences in government, history, and demographics—offers little for America to emulate. To outsiders, President Xi’s China looks like a technological juggernaut. Mainstream American media occasionally report on China’s supposedly superior STEM education, its ability to churn out semiconductor engineers and AI developers at breakneck rates, and the supposed culture of discipline that has displaced the lazy, complacent American worker. American industrial policy, critics say, is already doomed. The country cannot compete with China and its technological fervor.

 

Every time I hear these arguments, I cringe. The peddlers have typi­cally visited China for two-week, curated junkets.

 

State-approved tour guides extol Communist Party propaganda. They often push stereo­types about supposedly silent, hardworking East Asians, calling upon five thousand years of tradition, beating out self-absorbed, decadent Ameri­cans, ignorant of the wise ways of China’s technologists.

 

China’s semiconductor failures have dispelled these political and cultural myths. Its “national champions” collapsed because they were gilded and wasteful, conspiring against public interests under the veil of state secrecy.

 

Nevertheless, China’s mistakes offer critical lessons as America revives its own industrial policies amid intensifying geoeconomic competition. China’s semiconductor failures show the importance of allocating federal funding across regions and firms, rather than placing all our eggs in a few “superstar” cities such as Silicon Valley or “national champions” such as Intel. Americans must also be careful not to overemphasize federal decision-making in the centralized Chinese mold, which risks under­mining the market competition that best chooses winners and losers. While China has struggled at starting from scratch, the United States merely needs to dust off and update the Cold War infrastructure that supported technological innovation for past generations. China has mainly had success in industries that America has completely abandoned; in sectors we have continued to view as strategic, like semiconductors, the American model has proven far more resilient. Our anxieties about China must not turn us into China.

 

 

This article originally appeared in American Affairs Volume VI, Number 4 (Winter 2022): 14–22.

 

About the Author

Geoffrey Cain is a senior fellow at the Lincoln Network, visiting fellow at the National Security Institute at George Mason University, and a member of the advisory council at the Krach Institute for Tech Diplomacy at Purdue University. His most recent book, The Perfect Police State (Public Affairs, 2021), explores China’s surveillance industries.

 

 

ALSO BY GEOFFREY CAIN

• Ukraine versus Afghanistan: Lessons in National Solidarity