Export controls on China have reached a point of no return. A new and expansive package due this month or early next from the U.S. will mean that the semiconductor manufacturing sectors of the two countries are set firmly on a path towards complete decoupling. U.S.-China relations are likely to take a major hit, with Beijing ratcheting up its own controls on critical minerals exports, threatening a major tit-for-tat escalation. 

Despite all the rhetoric from Washington about small yards and high fences, and its stated desire to impose controls only on the most advanced semiconductors with potential military end uses, the reality of what the American government is doing is much more complex — with much deeper risks for leading U.S. technology companies. 

To add insult to injury, there has been little effort to run the numbers on all of this. In particular, the Biden administration has still not provided any real cost-benefit analysis of its export control measures. Nor has there been any attempt to show whether the government is achieving its national security objectives, nor indeed how it is measuring success on this score. In addition, the disconnect between export controls and industrial policy embodied by the CHIPS Act is becoming ever more apparent. 

In short, we are in a whole new world, but key parts of the system have not done sufficient homework. 

The Biden administration’s first export controls, put in place in October 2022, were designed to slow China’s ability to develop advanced artificial intelligence and the semiconductors that could be used to train and run AI models and algorithms. The justifications for this move have been all over the map. As I wrote in early 2024, the Biden administration has changed the goalposts repeatedly since October 2022. 

Already before October 2022, U.S. export controls prohibited Chinese firms from obtaining the most advanced lithography equipment, the most important tool needed to make advanced semiconductors. So if the U.S.’s goal was to prevent China from being able to produce chips below 5 nanometers in size, that train had left the station well before 2022. (The smaller chips are in size, measured in nanometers, the more advanced they are.) 

At that point, though, Chinese firms could still buy in the latest and greatest chips, including advanced graphics processing units (GPUs) to train advanced models; hence the 2022 package of controls included restrictions on the performance level of GPUs permitted for export, measures which were updated in October 2023 and are due for another change, probably in October 2024. All this after GPU leader Nvidia modified its GPU performance levels to comply with the restrictions in place at the time, only to see the goal posts moved. 

We are in a whole new world, but key parts of the [U.S.] system have not done sufficient homework.

The most significant aspect of all the packages since October 2022 concerns the tools used to make semiconductors, in processes such as etching, deposition, implantation, metrology and cleaning. Here, U.S., Japanese, and Dutch companies dominated the scene prior to 2022. When we wake up in October 2024, it will be a much different world. Two years of export controls have transformed the landscape, bringing Chinese toolmakers to the fore — and advantaging Japanese and Dutch toolmakers to some degree — but also angering key allies who do not share all of the U.S.’s concerns, all while massively disadvantaging U.S. toolmakers, heretofore world leaders in one of the most complex and important parts of the semiconductor industry supply chain. 

How did we get here and why? What are the costs and benefits? Inquiring minds want to know. 

The October 2022 controls were issued unilaterally, without full agreement from critical U.S. allies, and rested on some what now seem dubious assumptions — that U.S. firms were the only ones who made certain technologies, that Chinese toolmakers could not catch up, and that American allies would eventually fall into line. 

In addition to being unilateral, the controls included novel restrictions on domestic U.S. persons servicing tools in China and end-use controls that attempted to keep Chinese firms below a certain level of development. The controls also inexplicably included memory, both DRAM and NAND chips — think of the type of memory in your computer that always allows you to run an operating system and applications, and the type in your phone that keeps your data intact when you turn it off. All of these novel controls have proven to be either extremely difficult to implement, or have resulted in significant collateral damage to U.S. companies, or both. And the costs will continue to mount. 

First, the ‘domestic persons’ controls have not been applied by all the key countries with major companies in the game. Not surprisingly, the Netherlands and Japan have resisted imposing these types of controls, concerned about their leading technology firms losing access to their advanced tools that are being used at Chinese facilities, and the risk this will result in their technology being transferred to their Chinese competitors. For both governments, this is a real national security issue. 

This process has in fact already happened for U.S. toolmakers, who were forced to pull all their support personnel from Chinese facilities operating at technology levels now prohibited by so-called end use controls, which were designed to draw a line around how far Chinese semiconductor manufacturers could move up the technology ladder to more advanced products. In addition, there has been much confusion over which companies or facilities within Chinese manufacturing complexes are actually working at the level targeted by the end use controls. 

Drawing a line around Chinese technological advancement has proven to be very difficult too, because Chinese firms have been able to develop the tools they already had beyond their original capabilities. For example, China’s domestic foundry leader SMIC has been able to manufacture advanced semiconductors for Huawei’s new advanced smartphones despite U.S. export controls, using all of the technology and industry knowhow that it had developed before October 2022. 

The new rules have in turn massively incentivized Chinese front-end manufacturers like SMIC, YMTC and others to work more closely with domestic toolmakers. In the space of two years, this has transformed Chinese toolmakers from also-rans to becoming far more competitive. At the same time, Chinese semiconductor manufacturers are now eagerly excluding products and technologies made by U.S. companies across the semiconductor supply chain, due to concerns about the American firms’ ability to be reliable, long-term suppliers. 

The ultimate impact of the U.S.’s attempt to draw a line around China’s technological development will thus be to completely decouple U.S. toolmakers from Chinese manufacturers. This has already created a windfall for Japanese, Dutch and Chinese firms that compete with U.S. toolmakers. As these companies gain market share in China they will be able to leverage that additional revenue to compete with U.S. companies on price outside of China.

For the long-term viability of the U.S. semiconductor industry, this is a huge development, given that leading American technology firms had in the past hugely benefited from access to the China market, plowing revenue from their sales there back into R&D, thereby maintaining their technological edge. That virtuous cycle is now irreparably broken, with serious consequences for U.S. technology leadership. In a recent letter to Undersecretary of Commerce Alan Estevez, Congresswoman Zoe Lofgren (D-CA) and Senator Alex Padilla (D-CA) warned that “some [U.S. semiconductors] companies are even at risk of a ‘death spiral.’”

 A similar dynamic is happening with companies such as Intel and Nvidia, which are seeing their China sales eroded in ways that impact their cash flow. For Intel — which recently saw some of its licenses for shipment of commodity chips to China revoked — this has contributed to serious financial problems and major layoffs, plus delays in the firm’s buildout of fabs in the U.S. as part of the CHIPS Act. 

The real kicker here is that the supposed national security gain from all this turmoil in the industry remains and will remain, startlingly unclear. The one cost-benefit analysis so far done by a credible third party institution, the New York Federal Reserve, was issued in April. It concluded that “by forbidding U.S. firms to export to a selected list of Chinese firms for national security reasons, export controls aim to generate a selective decoupling of U.S. firms from China.” 

Many both inside and outside the industry are now questioning the supposed national security gains from end-use and memory controls in particular. Even so, the Biden administration is preparing to double down on these controls, expanding them to include high-bandwidth memory and more Chinese companies — and also putting the health of U.S.-China relations further at risk. With the controls appearing to be driven by a small number of White House officials who will be out of office in January, the question arises: What has been accomplished by the process started in October 2022 other than substantial damage to U.S. industry and technology leadership, the disruption of complex supply chains, the acceleration of the design out of U.S. technology in China and soon elsewhere — all with no measurable national security gain? Whoever wins the election in November will need to answer these questions.

---