MATT SHEEHAN
In early October, the U.S. government rolled out extensive new restrictions on China’s access to advanced semiconductors and the equipment used to make them. The restrictions require a hard-to-get license for the sale of advanced semiconductors to entities within China, largely depriving the country of the computing power it needs to train artificial intelligence (AI) at scale. The rules also extend restrictions on chipmaking tools even further to industries that support the semiconductor supply chain, cutting off both the U.S. talent and the components that make up the tools that make the chips. Together, these restrictions amount to the single most substantial move by the U.S. government to date in its quest to undermine Chinese technology capabilities.
The new restrictions also attempt to settle a long-running debate within U.S. technology policy. That debate centered on a perceived trade-off between two competing goals: damaging Chinese capabilities today versus maintaining American leverage in the future. With the latest rules, the U.S. government is betting that it can so deeply undermine China’s semiconductor fabrication capabilities that it won’t matter how motivated or well-resourced China’s efforts are to create its own semiconductor industry—they simply won’t be able to catch up.
Whether the U.S. government wins that bet will go a long way toward determining the future balance of global economic and technological power. Understanding the different sides of that bet requires looking at the different layers of restrictions and the knock-on effects they’re likely to have around the world.
HOW WE GOT HERE
For more than four years, the U.S. government has experimented with and iterated on new restrictions for China on semiconductors. These policies have plugged holes in existing regulations, expanded the scope of companies covered, and—perhaps most importantly—deepened how many layers into the supply chain restrictions go.
The first layer involved the direct sale of American-fabricated semiconductors to specific Chinese companies, initially ZTE and then Huawei. This happened when each of the firms was added to the Commerce Department’s Entity List, which bans companies from buying controlled U.S. exports without a license. (ZTE was later removed.) But that move proved ineffective at cutting off Huawei’s access to the advanced semiconductors used in 5G equipment, because the company could still purchase advanced logic chips from foreign fabrication facilities, known as fabs, in Taiwan or South Korea. American firms were also sometimes granted licenses to sell low-end chips to Huawei.
To plug some of these holes, controls on Huawei were then expanded to include a new Entity List Foreign Direct Product Rule (FDPR), which requires companies that use controlled U.S. technology in their production processes to comply with U.S. export restrictions. In this case, the change meant that any company in the world that uses U.S. technology to fabricate chips must apply for a license in order to sell to Huawei. Almost every part of the global semiconductor supply chain has significant U.S. inputs, so FDPR effectively allowed the U.S. government to force the rest of the world to comply with its own export controls on Huawei. Over the following years, the U.S. government added many more Chinese institutions to the Entity List, including many with ties to military or surveillance activities, though it did not apply the expanded FDPR to all of them.
These restrictions dealt a heavy blow to Huawei and lesser blows to the other companies targeted, but the rules only remained meaningful so long as China was unable to fabricate its own cutting-edge chips. In order to maintain that status quo, the United States had to slow down China’s efforts to indigenize semiconductor fabrication capabilities.
To do that, the U.S. government dealt with the second layer of the supply chain: semiconductor manufacturing equipment (SME), the highly specialized tools that fabs use to make chips. In a key early moment in 2018, the U.S. government successfully lobbied the Dutch government to not grant the necessary licenses for Dutch firm ASML to sell its most advanced equipment to Chinese fabs. Over the following years, Washington placed China’s leading fab on the Entity List, further restricting its access to crucial SME. Restrictions on SME proved particularly durable because these are often extremely specialized tools, requiring years or even decades of development. They are also produced by just a handful of firms in countries that are all U.S. allies, meaning coordination with allies can prove particularly potent.
For almost two years, different versions of this two-layer approach to semiconductor controls remained the status quo—but it was never fully stable.
CAPABILITIES VS. LEVERAGE
As U.S.-China tensions continued to worsen, a debate emerged in national security and foreign policy communities over whether the United States should pull the trigger on more drastic restrictions that would limit the sale of any advanced AI chips to China. That line of reasoning was bolstered by revelations that, despite long-standing bans on sales for military applications, the People’s Liberation Army used an array of intermediate and shell companies to continue purchasing advanced AI chips that were designed by American firms and fabricated in Taiwan and South Korea. But other foreign policy analysts pushed back, arguing that far more extensive bans on advanced chip exports to China could accelerate the time line for China to develop fully indigenous supply chains for chips.
Understanding this linkage requires appreciating the interdependence of chip design firms, fabs, and SME providers. For any one of these players to significantly advance their own capabilities, they need to be working with other advanced firms. Fabs have a hard time improving their chips if they aren’t working with good design companies, and SME providers struggle to improve their equipment if they can’t sell them to quality fabs. This is because all of these processes involve co-engineering, in which buyers and sellers give each other feedback and steadily refine their products together.
The existing setup prevented China from achieving this type of collaboration because it allowed most of the key Chinese players—their best chip design start-ups and fabs—the option to work with superior foreign partners rather than subpar domestic firms. Most advanced Chinese chip design companies could still fabricate in Taiwan, meaning they didn’t work with inferior Chinese fabs. And Chinese fabs still chose to buy foreign SME whenever possible, meaning Chinese SME firms struggled to get off the ground. In essence, allowing China access to greater capabilities today helped maintain longer-term U.S. leverage over China’s semiconductor sector by keeping it dependent.
The new restrictions attempt to put an end to this trade-off by both damaging Chinese capabilities and delaying the country’s ability to indigenize the supply chain. The rules implement an extensive ban on the sale of advanced chips to any entities in China, and they apply this outside of China by using new versions of FDPR previously reserved for Huawei. (To be precise, the ban is a licensing requirement with a “presumption of denial” for those licenses.) They also impose a ban on the sale of American-made SME for making advanced chips, though without the extraterritorial aspect of the FDPR. These two moves expand the scope of previous restrictions on the first two layers of the supply chain: chips and the tools needed to make them. (For a more detailed breakdown of more aspects of the restrictions, see this report from the Center for Strategic & International Studies.)
But the new rules also added two important and novel constraints that together constitute a third and a fourth layer of supply restrictions: cutting off Chinese access to many key components that go into making SME and preventing U.S. citizens and residents from working with Chinese semiconductor companies. Each of these will have a unique, and likely substantial, impact on China’s ability to develop its own indigenous semiconductor supply chain.
SME components are often extremely specialized tools—precision lasers or mirrors—that are produced by a single firm in the United States or an allied country, giving Washington unique leverage. Now not only will China not have direct access to U.S.-fabricated chips, but also its chip design companies will not have access to fabs that use American equipment, its fabs will not have access to U.S. SME, and its SME companies will not have access to U.S. components. Each step in that process multiplies the difficulty China faces in building this wildly complex and engineering-intensive industry at home.
In addition, the rules require a license for any “U.S. persons”—which include citizens, permanent residents, anyone who lives in the country, and U.S. companies—to work with Chinese companies contributing to advanced semiconductor production in China. All of these groups have been forced to halt work with Chinese semiconductor firms. This has proven immediately damaging, as leading SME firms have had to stop all servicing of equipment at Chinese fabs and many of these machines need maintenance every couple days in order to continue running. In addition, the dearth of experienced U.S. talent will inhibit Chinese industry, since semiconductor fabrication requires the kinds of intangible skills built up over decades of engineering work and can’t be captured in a blueprint or instruction manual. Prior to the new rules announcement, many U.S., Korean, and Taiwanese engineers had responded to this need by taking lucrative positions in the Chinese semiconductor industry—echoing the way Taiwan built up its own chip industry in past decades.
Taken together, the new rules are attempting to eliminate the capabilities-versus-leverage trade-off by making a bet: that even though the extensive chip and SME bans will finally force Chinese companies to work with each other, the additional layer of blocking SME component sales and foreign talent will more than offset any newfound gains from that collaboration—for the foreseeable future, at least.
WILL IT HOLD?
How well this strategy will work depends very much on time lines. In the next one to three years, these moves are likely to have a truly devastating impact on China’s access to advanced semiconductors and its ability to make them at home. China has been stockpiling some chips and SME in anticipation of such controls, but as those stockpiles dwindle and foreign support dries up, it will likely be forced to step backward in technological time and use less advanced chips that the industry has long since moved past. In the next four to ten years, the impacts will remain deep, but it seems likely that China will begin to have access to alternative supplies of key parts, namely SME and many components for SME.
This may be due less to China’s indigenous development of such tool—which still faces enormous hurdles—but rather to other international companies engineering out the U.S. inputs or components that make them subject to FDPR and thus U.S. export controls. Reports based on conversations with industry insiders allege that this scenario planning already began following earlier Huawei restrictions, with some estimating a time frame of four to six years for the construction of an overseas fab that would be free of the American restrictions. A version of this trend played out after Washington placed restrictions on the export of satellites, with U.S. market share of exports falling from 73 percent in 1995 to just 25 percent a decade later.
Many leading SME and component companies are headquartered in Germany, Japan, the Netherlands, South Korea, and other U.S. allies that often share a general concern about Chinese technology prowess. But the exact cost-benefit calculations around equipment sales look different when viewed from a smaller country that has far fewer technology juggernauts, more overall reliance on economic ties to China, and either less acute threat perceptions or greater fear of escalation risks. The recent U.S. restrictions represent a turn toward a strongly zero-sum approach to confronting China on technology, an approach that isn’t equally compelling to countries that don’t see themselves as locked in a battle to be the one dominant global superpower. At the very least, companies would aim to reduce U.S. components so that they are only subject to the veto power of their home government and not a foreign power.
In the next decade, the impacts of these restrictions are deeply uncertain. If China’s SME firms, fabs, and chip design industries are forced to work together and manage to survive the initial onslaught of restrictions, they may emerge both strong and fully untethered from U.S. controls. Many analysts (myself included) who warned of the risks of this strategy did so with this possibility in mind. Those concerns are legitimate, but they come with a huge amount of uncertainty—technological, economic, and geopolitical—baked in. In issuing the latest round of export controls, President Joe Biden’s administration chose to play the cards it had for the impact it can create today.
No comments:
Post a Comment