Managing Semiconductor Exports to China

It may seem counterintuitive to say that the United States should continue to allow semiconductor exports to China, but this is because the goal may be misstated. The goal is not to prevent China from building its own semiconductor industry. The United States can slow this, but it cannot stop it. The goal is to keep the U.S. semiconductor industry strong.

As the Trump administration moves forward with expanded restrictions on semiconductor exports to China, it should base decisions on two general criteria: (1) the measures should not harm U.S. companies more than China, and (2) the measures should not be symbolic, attempting to restrict sales or transfers of commodity technologies or technologies available from foreign sources.

These points may seem obvious, but there are important examples where attempts to block technology transfer did major harm to U.S. companies while only accelerating the growth of suppliers outside the United States. These include satellites, machine tools, and encryption (for a time). When the United States banned the export of satellite components to China, all that foreign companies did was design U.S. components out of their products and build alternate sources of supply.

We risk making the same error with semiconductors. The desire to restrict exports is understandable. China is still dependent on the West for advanced technology, although this dependence is shrinking. But overly broad restrictions will harm the United States more than China. Here are seven conclusions that can help guide U.S. policy on semiconductors and China.

1. The United States is not going to stop China from making basic semiconductors any more than it stopped Japan or Korea. China’s decades of investment and strong science and technology workforce mean that it will eventually be a chip producer. The United States is also not going to put Huawei out of business. Huawei has been harmed by the U.S. effort to call attention to its years of untrustworthy behavior, but the Chinese government will not allow it to collapse—Huawei is too important. U.S. restrictions on exports to Huawei of commodity end-items such as chips and operating systems have only accelerated the company’s efforts to develop alternative sources of supply. Its subsidiary’s (HiSilicon) strong chip design capabilities and access to fabless production services have allowed Huawei to begin to design out some U.S. chips.
 

China has always wanted a strong semiconductor industry. It began its efforts to develop an indigenous industry more than 30 years ago. Progress has been uneven, but the Chinese commitment has remained steady. China hopes to make all categories of chips and displace Western suppliers. To do this, China has used massive investments in semiconductor production and commercial espionage to steal semiconductor technology. Its efforts have not paid off, but that is beginning to change for some categories of chips, most notably in memory chips, where China is becoming competitive in producing “NAND” chips (“NAND” refers to the operation the chip performs). However, China cannot yet make more advanced categories of chips.

2. China is not going to play by the rules in the competition over semiconductors. Some years ago, an internal policy debate between those in China who wanted a more internationalist approach to trade versus those who wanted a unilateral approach was settled in favor of the latter. Under Xi Jinping, this tendency has grown even more pronounced. The United States would need an astute diplomatic strategy to change this, but this would take years and will not happen while Xi is in power. This means we can expect to see increased pressure from China on Taiwanese and Japanese suppliers, more Chinese acquisition efforts and other financial dealings in Europe, and increased espionage against the United States in the battle over semiconductors.
 
3. Fujian Jinhua is a good precedent for U.S. policy. This Chinese chip company stole memory chip technology from an U.S. firm and was placed under sanction by the United States as a result, ultimately forcing the company to stop production. If Chinese companies are allowed to sell products made using stolen technology—and there are examples in the aircraft, railcar, automotive, battery, and green energy industries (to name a few)—there is no penalty for intellectual property (IP) theft. Traditional remedies using World Trade Organization processes are inadequate for holding Chinese companies accountable, as they are slow and were designed to deal with individual cases, not a major national program of IP theft. Promises in the trade deal are viewed skeptically.
 

U.S. export controls were not designed for a hostile power with which the United States has an exceptionally close economic relationship. Cold War controls were developed at a time when the interconnections between Western and communist economies were almost non-existent. In contrast, China encouraged foreign investment and was a lucrative market for Western companies. The result was a deeply interconnected supply chain.

The 2016 decision to allow the U.S. chipmaker AMD to enter a joint venture with a Chinese firm and supply it with technology for logic chips is a good example of the problems with outmoded controls. It gave China access to a chip technology it had long sought unsuccessfully. The AMD transfer was allowed to go forward because it was legitimate under existing export control rules. While approval has since been reversed, it is clear these rules were designed for the Cold War, not for competition with China.

The AMD case points to the need to accelerate export control modernization. Recently announced Commerce Department regulatory changes could be a good first step if they are implemented in ways that allow continued exports of commodity chips to China or the export of semiconductor manufacturing equipment (SME) to U.S. firms operating in China. If the goal of new controls is to acquire knowledge of PLA supply chains and front companies, earlier rules on encryption where companies were required to notify an export but did not need a license unless otherwise informed by the Commerce Department might be a good precedent.

4. Semiconductor manufacturing equipment is a chokepoint, but the United States does not want to accidentally choke U.S. or allied companies. The United States produces about half of SME, and Japan makes another third. The most advanced equipment comes from these two countries (and one Dutch supplier, ASML). Restrictions on SME exports to China, if used carefully, would slow China’s semiconductor growth. A straightforward approach would block export to Chinese companies while allowing sales to U.S. and Japanese firms, even if they are located in China. Complex and expensive, SME is impossible to divert from one company to another, and in more than 20 years of SME exports to U.S> firms in China, there have been no cases of diversion.
 
5. The Taiwan Semiconductor Manufacturing Company (TSMC) is a problem. It is the primary source of indigenously designed, advanced Chinese semiconductors. If the United States blocks exports of commodity chips, China (and Huawei) will design replacements and look to TSMC to make them. This is not an easy problem to resolve since U.S. companies also depend on TSMC for “fabless” production (where a company designs a chip but has a contractor make it). Apple’s announcement that it would make chips of its own design at TSMC is an example of this.
 

Semiconductors are not easy to make. This has been the primary obstacle to China’s progress in developing an indigenous industry. Chip making requires advanced manufacturing equipment, complex IP, chip design capabilities, lithography “masks,” and software design tools, many of which are proprietary. The manufacturing process has become progressively more complex as the chips’ feature size has become microscopically small, making semiconductor manufacturing a question of working at the cutting edge of physics and materials science. It is not enough to buy production equipment and steal IP and technology. You need “know-how,” the skill sets developed through years of experience.

China tried to overcome the “know-how problem” in several ways. A few years ago, instead of buying semiconductor technology, it attempted to acquire entire companies to gain know-how. Decisions by CFIUS to block sales and changes in U.S. foreign investment regulation closed off this avenue. A second approach is to encourage individuals who have acquired know-how by working for a foreign chip company to return to China and start a business there. SMIC, for example, was founded by a long-time employee of Texas Instruments. This approach has had some success, but it is difficult to scale, as you need more than a few individuals to make semiconductor manufacturing work. The most successful solution has been for China to develop semiconductor design capabilities and then contract out manufacturing to companies such as TSMC.

TSMC and other foreign producers create a “threshold” above which any U.S. restrictions will be ineffective. A solution to TSMC will require a combination of product retransfer controls for SME, which would stop TSMC from supporting Chinese companies, but this could be politically difficult for Taiwan and very damaging for TSMC. It is unusual to impose export restrictions on the products of transferred technology, but there are precedents from the world on nonproliferation and munitions.

6. Increased counterintelligence efforts are essential. To be fully effective in stopping illicit technology transfer, restrictions on exports and Chinese investment are essential but must be accompanied by more aggressive efforts to stop the illicit acquisition of U.S. technology. This does not require banning all Chinese students, researchers, or tech workers, but it does require greater efforts at denying access to identified Chinese technological intelligence collectors. The FBI and the Department of Justice have made good strides in this direction, but the United States may underestimate the scope of the problem and how much it is likely to grow as new restrictions take effect. This means the United States must devote more resources to counterintelligence.
 
7. The United States should not impose new restrictions on commodity chips. All chips are not created equal. Most chips fall into three broad categories: (1) memory, (2) logic, and (3) specialized chips, such as FGPA (Field Programmable Gate Arrays) or chips for telecommunications or AI. Chips come in generations, determined in part by Moore’s law, which accurately predicted a doubling in performance every two years, driven by improvement in manufacturing technology and materials. The chips vary in complexity of design and production. Memory and logic chips are essentially commodities, manufactured in the millions at relatively low cost and are widely available from manufactures and resellers around the world.
 

What kind of chips is China able to make? China’s decades of investment and strong STEM workforce mean that it will eventually be a significant chip producer, but some chips, even if they are commodities, are still beyond China’s ability to produce. China has had the most success in memory chips, and China is only a few years away from being competitive in other categories. Though China cannot yet make the most advanced chips, it aims to eventually produce all categories.

Memory chips are semiconductors used to store data. Most digital devices require several kinds of memory chips. China now buys most of these form Western companies (although their fab may be located in China). NAND is one of the most common forms of memory chips, and since their introduction in 1987, they have found their way into a wide variety of consumer devices. Currently, six companies—Samsung Electronics, Toshiba, Western Digital, SK Hynix, Micron Technology, and Intel—dominate the memory chip market (valued at an estimated $58 billion). A new Chinese entrant, Yangtze Memory Technologies Corp (YMTC) is supported directly by the Chinese government from one of its multi-billion-dollar semiconductor investment funds and is gearing up to produce NAND chips.

The arrival of a disruptive new entrant has been the norm in the memory chip market for many years, but YMTC’s connection to the larger Chinese government strategy of displacement makes it an unusual case. The question is how best to position Western companies to compete with YMTC. Restricting chip exports to give it a protected market is not the answer. New U.S. rules should not restrict exports of memory chips to China unless the United States wants to accelerate YMTC’s growth and, ultimately, the growth of the Chinese chip industry.

What does this mean for the memory market? If YMTC or other Chinese companies succeed in making commercially viable memory chips, a new source of supply will be introduced and shrink market share and revenue for the other producers. However, if YMTC no longer had access to Western SME or materials, this would slow the company’s competitiveness, growth, and its ability to produce more advanced chips.

To compete with YMTC, the United States should allow U.S. and Western companies to continue to export to ensure they retain market share and watch the Chinese government closely for illicit behavior promoting YMTC. Even if YMTC can make NAND chips, this is a highly competitive market that requires the ability to coordinate complex research and industrial processes and a global supply and marketing chain. China may be tempted to help YMTC by pressuring memory chip customers to relocate to China for preferential treatment.

Export Controls Must Be Part of a Larger Strategy

The best outcome would let U.S. companies continue to sell to China to protect market share and revenue in ways that do not give China strategic advantage. This can be done by permitting exports of commodity chips even to companies such as Huawei. It can also be done by permitting exports of SME to America and firms with facilities in China but not to Chinese firms.

The United States needs to strengthen its semiconductor industry. Export regulation is essential but not sufficient. The United States does not need to copy China’s top-down industrial policy for semiconductors, but there are areas where federal investment is necessary. First, the United States needs to increase the supply of STEM labor through support for students and research. Second, if the goal is to encourage companies such as TSMC to move semiconductor production facilities from China to the United States, the United States will need to provide subsidies. In both cases, this will entail billion-dollar or larger investment programs. This sounds expensive but is appropriate and necessary for a strategic industry. Third, there are complex IP and trade disputes in semiconductors where the United States should intervene to support U.S. companies.

Working with allies—Japan, South Korea, and the European Union—is also essential. These partners are increasingly concerned by China’s efforts to manipulate the market and are prepared to cooperate to minimize market disruption and restrict access to SME. U.S. policy is more likely to succeed if done cooperatively as part of a larger strategy to reinforce the U.S. semiconductor industry, but this strategy has yet to be fully defined.

Semiconductors are the backbone of the digital economy. Their primary use is commercial, but there are also important military applications. The ability to manufacture chips plays a leading role in the growing technological competition between the United States and China, not just in semiconductors but in other important areas such as AI. There are deep interconnections between the two economies, and China is committed to restructuring those connections to favor its interests. But this is a winnable fight. The United States needs to pull back from its close economic integration with China, but this will need to be done carefully and in ways that best support the continued strength of the U.S. semiconductor industry.

James Andrew Lewis is a senior vice president and director of the Technology Policy Program at the Center for Strategic and International Studies in Washington, D.C.

Commentary is produced by the Center for Strategic and International Studies (CSIS), a private, tax-exempt institution focusing on international public policy issues. Its research is nonpartisan and nonproprietary. CSIS does not take specific policy positions. Accordingly, all views, positions, and conclusions expressed in this publication should be understood to be solely those of the author(s).

© 2020 by the Center for Strategic and International Studies. All rights reserved.