Pages

1 February 2021

Lagging but motivated: The state of China’s semiconductor industry

Christopher A. Thomas

With advanced semiconductors key to powering a wide range of potentially transformative technologies, cutting edge computer chips have become a heated area of geopolitical competition for the 21st century. Despite their importance, semiconductors represent a rare area in which the Chinese economy is dependent on the rest of the world—rather than the other way around. Every year, China imports more than $300 billion of semiconductors, and most, though not all, major American semiconductor companies pull in at least 25% of their sales from the Chinese market.

This mutual dependence has benefitted the technology sectors in both countries. Every major Chinese technology company relies on U.S. chips: Tencent or Alibaba would not be the powerhouses they are today if they had relied on Chinese microprocessors during their formative years or had developed and manufactured their own. Many U.S. companies, meanwhile, have benefited from Chinese customers, markets, and innovations. The scale and cost reductions enabled by system and device manufacturing based in China and Asia more broadly has helped make information technology ubiquitous. Despite the harsh rhetoric on both sides of the Pacific, American semiconductor companies and their Chinese counterparts today are working together on hundreds, if not thousands, of product designs and joint technology development efforts.

Yet these collaborations have not prevented semiconductors from becoming a central faultline in tensions between the United States and China. In a post-COVID, post-Trump world, many in Washington would like to see the American economy less dependent on China and are exploring new restrictions on imports of Chinese hardware and exports of both cutting-edge semiconductors and the equipment required to manufacture them. Meanwhile in Beijing, Chinese officials are pursuing a clearly stated, though ambiguously defined, goal of “technology independence,” as articulated in the 14th five-year plan outlined last year.

But how to achieve that independence—and whether pursuing it makes sense in the first place—represents a question of profound uncertainty. As U.S. officials weigh their policy options in that regard, they first need to level-set on the state of the Chinese and global semiconductor industries, and how Beijing has approached its goal of building a domestic chip-making industry. Though it has made major advances, most segments of China’s semiconductor industry remain behind its foreign competitors, and its efforts to catch up face major economic obstacles. How the United States approaches its policy toward that industry will have major ramifications not only for the U.S. relationship with China, but also for the American semiconductor, systems, and internet services industries, which remain deeply intertwined with China.

Understanding the semiconductor industry

Although the United States is home to a majority of the world’s leading semiconductor firms, no country has true independence in the semiconductor value chain. The United States depends on critical foreign inputs and manufacturing capacity in the rest of the world. The manufacturing chain for any given semiconductor is extraordinarily complex and relies on as many as 300 different inputs, including raw wafers, commodity chemicals, specialty chemicals, and bulk gases; all are processed and analyzed by upwards of 50 different types of processing and testing tools. Those tools and materials are sourced from around the world, and are typically highly engineered. Further, most of the equipment used in semiconductor manufacturing, such as lithography and metrology machines, rely on complex supply chains that are also highly optimized, and incorporate hundreds of different companies delivering modules, lasers, mechatronics, control chips, optics, power supplies, and more. The “installed base” within a semiconductor factory today represents the cumulation of hundreds of thousands of person-years of R&D development. The manufacturing process that integrates them into a single manufacturing chain could represent hundreds of thousands more.

The types of products for which these manufacturing processes are designed are nearly as varied as the manufacturing inputs themselves. There are at least 20 major semiconductor product categories (from optical sensors to battery management modules to CPUs) and each category usually contains hundreds of different stock keeping units—distinct items for sale—for specialized applications. This complexity leads to a large market filled with myriad niches, in which specialized world-class companies have built defensible market positions through decades of targeted research and development.

Complexity also makes semiconductors a winner-take-all industry. The top one or two players in any given niche—whether a small one, such as furnaces, or a giant one, such as server CPUs—earn all the economic profits in that niche due to scale, learning efficiencies, and high switching costs for customers. It is rare to see newcomers break into these oligopoly positions. For instance, the market leader in graphics processing units (GPUs), Nvidia, invented the segment in 1999 and never relinquished its lead. While China has early-stage startups in the GPU segment, its market share is essentially zero. TSMC, based in Taiwan, was the first dedicated competitor in the foundry segment and has not relinquished its lead in its 33-year history. Indeed, SMIC, China’s leading competitor in the foundry segment, remains four or five years behind TSMC in technology, despite almost two decades of investment.

Beijing’s semiconductor strategy

Prior to the last half-decade, China spent more than 30 years and tens of billions of dollars to build a domestic semiconductor industry, showering its national champions with resources to compete with Western companies. Despite these investments, Chinese semiconductor companies make up a relatively small part of the global market. Chinese companies hold global share across the value chain at the following levels, according to data from IHS iSuppli:

Approximately 20% of fabless chip designers (the companies that design their chips and then contract out their physical production);
10% of global foundry plant capacity (the outsourced manufacturers that serve fabless chip designers);
Less than 1% of global integrated device manufacturing capacity (the companies that both design and manufacturer their own chips);
Less than 1% of electronic design software, semiconductor tools, and materials;
A less than 1% share in the most important end-product categories, such as the logic chips that are the brains of the internet or the advanced memory chips that major cloud vendors use to store trillions of photos and videos.

To counter its dependence on foreign suppliers of semiconductors, China announced a major new semiconductor policy in 2014. The “Made in China” policy, which launched the following year, included core technologies to semiconductors. The new semiconductor national policy contained two major innovations to previous industrial policy efforts: The first was to acquire technology from overseas via M&A; the second was to bring in “smart money” via private investors, such as private equity funds, to take the lead on investments. Over time, that policy has shifted toward a more traditional industrial policy model, with large manufacturing and R&D subsidies delivered to designated national champions. But with more than 50,000 Chinese entities registered as “semiconductor companies,” that investment is at risk of fragmentation.

Despite this effort to build up the Chinese semiconductor industry, the regional structure of the industry—based on the global distribution of market share using company headquarters location—was essentially unchanged in 2020 compared to 2014, and there has been no major shift to China in that time period. Chinese players remain decades behind in some of the most important manufacturing technology areas, such as lithography and the most advanced software design tools. The Chinese Semiconductor Industry Association estimates that to meet its policy goals, China will need to close a talent gap of roughly 300,000 engineers. In recent months, the Wuhan government stepped in to take over one cash-strapped manufacturer that had pledged to invest $20 billion in logic manufacturing, an example of the struggles facing the Chinese semiconductor industry.

Despite these setbacks, progress has been made. China’s share of back-end manufacturing, which is the labor-intensive process whereby processed semiconductor wafer is diced into individual chips, packaged, and outfitted with electrical connectors has nearly doubled since 2015 to 40% due to acquisitions. Market share of Chinese companies in fabless design has also nearly doubled, mainly via HiSilicon, the semiconductor arm of Huawei. These segments are more natural fits for China’s competitive advantages: Back-end manufacturing is a less technically challenging process that depends on low labor costs and operational efficiency; fabless design companies benefit from closeness to end customer applications (of which Chinese OEMs develop many) and have low barriers to entry due to widely available off-the-shelf design tools. There are now Chinese competitors in nearly every single stage of the value chain, including chemicals, materials, tools, and manufacturing, though some of their technology lags market leaders. China is turning its vast pool of venture capital and engineering talent to focus on this industry. In similar markets with lower technology barriers, such as MOCVD tools used for manufacturing LEDs, Chinese companies have developed competitive manufacturing equipment. While there are many potential outcomes of the Chinese semiconductor policy depending on tactics and the quality of execution by Chinese engineers, the Chinese semiconductor industry will undoubtedly become more competitive over the next ten years.

The 14th five-year plan appears to enshrine “technology independence” as a national strategy, though what independence means is not clearly defined. If we define technology independence as “self-sufficiency”—a fully in-country, PRC-controlled semiconductor supply chain that that serves all Chinese customer needs and does not infringe on the intellectual property rights of any global company—this is clearly not possible in the medium term. In the long term, it is also unlikely to be achievable. For one thing, the economics of an “only in China for China” supply chain do not work. Even if Chinese companies at each stage of the value chain win 80% of potential business from every potential Chinese customer, Chinese companies would collectively generate less than 15% of the industry’s overall R&D capacity—and likely less as prices in China tend to be lower, leaving less profit to re-invest in R&D. Such an indigenization strategy would still leave China behind the rest of the world: How can products developed with 15% of the world’s R&D compete with those from entrenched companies spending collectively far more? Of course, PRC government subsidies can and are closing that funding gap. But keeping such large-scale subsidies in place for the decades required to build the industry would likely generate a set of companies so dependent on government largesse that they may not be commercially viable.

As in the phone and PC industries, Chinese semiconductor companies would be better off focused on winning customers rather than winning government support, and by partnering with foreign firms to co-develop technology, build ecosystems, and pursue global customers. This is what several major Chinese consumer electronics companies, such as Vivo, Xiaomi, and Lenovo, as well as large telecommunications equipment manufacturers, have done. In many years, more than half of the sales of these leading device and system companies are outside China.

But as semiconductor technology has emerged as a point of competition with the United States, such partnerships have become more difficult. Since 2016, in response to concerns about Chinese tech firms’ involvement in human rights abuses and Chinese semiconductor companies’ links to Chinese military institutes, the U.S. government has added major Chinese consumers and producers of semiconductors to the entity list. These firms include semiconductor consumers such as DJI, ZTE, and Hikvision; targeted semiconductor producers include Huawei and SMIC. The companies on the list are generally ineligible to receive any item subject to Export Administration Regulations without a license provided by the Bureau of Industry and Security. The government has also tightened oversight of acquisitions of or investments in sensitive technologies such as semiconductors and has limited joint R&D and academic engagement between U.S. and Chinese companies, labs, and educational institutions. Taken together, these moves by the U.S. government have made the already difficult task of building a competitive Chinese semiconductor industry that much harder.

Despite the economic logic that would drive Chinese companies to globalize, many Chinese companies in the semiconductor value chain, from startups to established players, are quietly pursuing supply chain “indigenization” strategies. They are leveraging the current moment to focus on import substitution and are angling to remove American suppliers from their approved procurement lists or to become that alternative supplier. They encourage the Chinese press, government officials, and investors to view them as the “Chinese champion.” Executives in these Chinese firms are not making these decisions out of “patriotic spirit” or pursuing “national policy.” Rather, they are pursuing business-continuity goals, aiming to reduce risks from future export controls and entity listings (and, in many cases, hoping for support for a lucrative stock market listing). Since many of these executives have previously worked at large multinational corporations—and some even hold American passports—indigenization strategies are not ones that they would typically pursue. Yet the prospect of new and more extensive trade restrictions has led them to embrace more of an “in China for China” model.

The Chinese government is encouraging companies to follow this indigenization path. For example, the government has offered to provide insurance to Chinese companies to protect against faulty equipment or materials from Chinese suppliers and has also made manufacturing subsidies contingent on a commitment to use local suppliers. But make no mistake, the primary driver of the move toward indigenization are business decisions driven by businesspeople on the ground, rather than national policy.

How the Chinese and global semiconductor industry evolves from here depends largely on the strategic moves and the engineering execution of Chinese companies—but also on the policies of the incoming Biden administration. The move toward indigenization within China, coupled with widespread fears of American dependence on Chinese hardware and manufacturing, will pose a unique challenge for Biden’s foreign policy team. Along with their counterparts in Congress, the Biden administration will need to carefully weigh the benefits that come from U.S. semiconductor companies competing in one of the world’s largest markets against the risk of American-sourced technology being used to endanger national security; clarify the often blurry line between “primarily state-driven” and “primarily market-driven” operators in China; balance the benefits of having the best Chinese scholars studying in top engineering schools and collaborating with our national labs against the risks of industrial espionage and the infringement of intellectual property rights; and author a semiconductor industrial policy that continues to let the market decide winners and losers.

How the Biden team should approach those questions will the subject of an upcoming article. Yet even as the future of the Chinese semiconductor industry remains unclear, there is one undisputed outcome for now: The United States and its companies are losing influence and market share in the private Chinese technology sector. These vibrant, advanced, and innovative Chinese companies are increasingly determined to forge a new semiconductor ecosystem centered on China. The repercussions of this shift in mindset will reverberate far longer than any short-term benefits derived from using access to semiconductors as a negotiating lever in the larger context of U.S.-Chinese relations.

Christopher A. Thomas is a nonresident senior fellow in Foreign Policy at Brookings, a board director at Velodyne LIDAR, and a visiting professor at Tsinghua University. He is also affiliated with the Brookings Artificial Intelligence and Emerging Technology Initiative.

No comments:

Post a Comment