China’s chip chokepoints

Earlier this year, American investigators were tipped off about a plan to ship semiconductor manufacturing equipment to China.

The alarm was sounded after two Chicago-based buyers of a Dynatex DTX-MDB 150 "scribe and break" machine asked the California manufacturer to file export paperwork that omitted the machine’s final destination: Chengdu GaStone Technology Company, a firm in southwest China under U.S. sanction for an illegal 2021 scheme to purchase military-grade integrated circuits. In April, Lin Chen was arrested and charged while Han Li escaped to China.

The case illuminates the complex U.S.-China competition to produce ever-more-advanced chips. CGTC sought the Dynatex machine to handle a final step in making military-grade gallium arsenide chips, which can better withstand extreme temperatures and radiation than ordinary silicon devices. Had they succeeded, the Chengdu firm could have improved their chip-manufacturing "success rate" and therefore their ability to meet customer demand—including from China’s military.

While companies such as SMIC have turned China into a chip-manufacturing powerhouse, the country remains unable to mass-produce the most advanced chips on its own. Despite massive government investment, China continues to rely on foreign companies for the chips that power everything from the AI revolution to the internet of things to the most advanced military equipment. This problem remains a huge vulnerability for China’s economy, its military, and its global ambitions.

Commentators inside China identify advanced chips as one of the “chokepoint” technologies that keep the country from technological parity with the world’s most advanced economies. In 2018, the government’s S&T Daily publication named 35 such technologies, most in the fields of information technology, electronics, manufacturing, biotech, and new-energy vehicles. Eight of these chokepoints related to the design and manufacture of advanced chips, casting doubt on the country’s ability to produce such chips anytime soon.

Western leverage

While China’s struggle is hardly secret, few stop to think about what exactly it means, why it is so difficult to overcome—and what leverage it might give U.S. and allied governments.

To paraphrase Boromir, one does not simply insert silicon into one end of a machine and wait for an advanced chip to emerge from the other. Producing these chips is a complex, multi-stage process. And each stage, in turn, provides numerous avenues for U.S. and allied governments to apply pressure on China’s efforts at its own advanced chip production.

Before the chips can even be manufactured, they must be designed. This necessitates the use of electronic design automation software, which brings order to the billions of transistors that go into a modern chip, a task impossible for human engineers to accomplish on their own. Here lies China’s first chokepoint. Chinese-made EDA software from companies such as Beijing Empyrean and Primarius Technologies can handle analog designs and simpler IoT applications but appear unable to compete at the high end. Thus, the Chinese market for this specialized software is dominated by the three largest U.S. developers: Synopsys, Cadence, and Siemens EDA, who collectively hold 90 percent of the PRC market. In 2022, the U.S. Commerce Department placed export controls on the most advanced versions of this software, which is needed to design top-of-the-line 3nm chips.

Next, raw materials must be turned into the silicon wafers that will eventually become chips. Silicon is extracted from sand, refined to 99.9999999% purity, melted, and fashioned into a single-crystal ingot. This ingot, called a “boule,” is sliced into thin wafers and polished to create an ultra-flat, mirror-like surface. To achieve the extraordinary nanometer-level smoothness that is required, ultra-precise polishing equipment is needed. Herein lies a second chokepoint: China appears to lack companies that can produce this equipment, although their research is showing some promising results. The leading manufacturers are in the U.S. and Germany. In 2020, the U.S. government slapped export controls on the most advanced versions of this equipment.

Once the wafers are ready, one side is coated with a light-sensitive chemical called photoresist to prepare it for photolithography. Production of this chemical is a third chokepoint. “Extreme ultraviolet” photoresist is needed for the densest chips below 20nm, and this market is dominated by Japanese companies bound by strict export-control policies. Even South Korean and Taiwanese semiconductor manufacturers depend on Japan’s advanced resists. Chinese companies make less-advanced photoresists, which are in high demand worldwide for manufacturing inexpensive commodity chips such as those found in automobiles and consumer products. But their products are insufficient for leading-edge devices.

Chinese firms such as Jingrui, Red Avenue, and its subsidiary Beijing Kempur, a U.S.-China joint venture, are pouring research into making advanced photoresists. This will not be easy, because older photoresist chemicals based on bis-azide photoresists cannot create circuits below two micrometers (2,000 nanometers). Further, the most advanced core technologies, including fine-particle filtering for new chemicals such as fullerenes (a zero-dimension carbon nanomaterial for high-resolution photoresists) may not be exported from the U.S. and its allies to China without a license.

Next, the light-sensitive wafers are sent through a photolithography machine, which projects light through a series of reticles to produce microscopic circuit patterns. Herein lies yet another chokepoint. The most advanced machines, manufactured by Dutch company ASML, have more than 100,000 parts and cost hundreds of millions of dollars. They have carried a “presumption of denial” in the export application process led by Washington since 2022.

While China manufactures huge amounts of less advanced chips, it remains incapable of reproducing the complex photolithography machinery needed for the most advanced devices. Based on past performance, the prospects for quickly catching up with the world’s photolithography industry leaders are not good: China has striven to do so since the 1980s and has historically been a decade or two behind. China’s leading producer, SMIC, can reportedly make 28nm chips, far from the global leading edge of 2 to 3nm. And while China is reportedly now producing a 7nm chip for the latest Huawei smartphone, this is believed to use a less advanced and less efficient form of lithography from a machine acquired from ASML before U.S. controls went into effect. SMIC’s success rate for chips on each wafer may be low and commercially unacceptable, keeping costs high and production limited. And Washington is trying to further restrict sales and service of the lithography machines SMIC used to make them.

There are almost certainly other chokepoints buried in the intricate process that produces modern semiconductors, perhaps in the stages of etching, ion implantation, metal deposition, dicing, and final assembly and testing. For instance, Chen and Li’s efforts in Chicago indicate that China depends on foreign “scribe and break” machines to dice military-grade gallium arsenide chips, making this yet another potential chokepoint not named in S&T Daily’s 2018 list.

The multiple chokepoints China must contend with provide a dual picture. They show the difficulties it faces in its quest for world-class chips. Meanwhile, they show how Washington’s efforts to keep cutting-edge semiconductor technology out of China have achieved some successes, but still face significant hurdles. Export controls have only been partly updated since the Cold War and rely on a network of nations and companies. The United States needs a new multilateral export regime powered by consensus among key partners that make semiconductor manufacturing tools, many of whom would prefer to continue selling to China at some level.

China will continue to improve its chip-making through overt and covert means, for the one certainty about the future of the economy and warfare is that they will be powered by semiconductors. The question is where those technologies will come from, and how much volume in chips and equipment clandestine actors can manage to smuggle into China.

Matt Brazil is a China Analyst with BluePath Labs and the co-author (with Peter Mattis) of Chinese Communist Espionage, An Intelligence Primer (2019). He is a former employee of the Commerce Department and of Intel Corporate Security. His views do not represent those of any institution.

Matt Bruzzese is a senior Chinese-language analyst for BluePath Labs.

P.W. Singer is a best-selling author of such books on war and technology as Wired for War, Ghost Fleet, and Burn-In; senior fellow at New America; and co-founder of Useful Fiction, a strategic narratives company.