WASHINGTON — President Joe Biden’s motorcade pulled up to a 40-acre construction site dotted with some two dozen red, blue and gold cranes. An American flag hung from one of the site’s buildings alongside a banner that read “A Future made in America: Phoenix, Arizona.”
The site will feature a new facility of the Taiwan Semiconductor Manufacturing Company. Biden had arrived to tout the company’s newly announced $40 billion investment in U.S. microelectronics manufacturing following the July enactment of the CHIPS Act, which included $52 billion in subsidies and tax incentives for businesses that produce semiconductors within the country.
“American manufacturing is back, folks,” Biden proclaimed.
The president’s speech in December focused on creating news jobs in a swing state — the bread and butter of economic and political victories.
But he also announced a “gamechanger.” In 2026, the Taiwan Semiconductor Manufacturing Company would begin building the world’s most advanced microelectronics in Arizona: 3-nanometer chips.
That’s on top of the approximately 20,000 5-nanometer chips per month TSMC intends to start producing in Arizona next year.
READ MORE: IISc researchers build a platform for energy-efficient computing
Mounting tensions with China have spurred a new push in Washington to lessen U.S. reliance on East Asia — the global epicenter of semiconductor production — for the vital microelectronics needed to create both conventional arms and the artificial intelligence algorithms integral to building the weapons of the future.
But left out of Biden’s Arizona speech is a similarly urgent effort the U.S. government is undertaking to slow China’s advancements, including new restrictions that limit Beijing’s access to these highly sophisticated chips.
In October, the Commerce Department’s Bureau of Industry and Security announced a sweeping range of export controls that severely curtail China’s ability to obtain some of the world’s most cutting-edge microchips.
The bureau had argued that Beijing could use them to “produce advanced military systems” — although these semiconductors are also used in civilian technology.
Alan Estevez, the undersecretary of commerce for industry and security, said in December that the export controls will “slow China’s ability to produce the highest-end semiconductors, for a period of time.”
“They will figure this out,” he added. “But what we’ve done is pretty comprehensive.”
READ MORE: India planning hyper scale data centre’s for Armed forces Data
The Commerce Department’s restrictions put the U.S. on the offensive against China’s ability to procure the advanced microelectronics needed to meet its long-term military modernization goals.
The export controls prohibit businesses such as TSMC from continuing to produce these sophisticated microchips in China — even as the Taiwanese company begins manufacturing its advanced semiconductors in Arizona.
Several days after the regulations were unveiled, TSMC obtained a one-year U.S. waiver to continue operating its semiconductor production site in Nanjing, China — with one condition.
Its China-based operations must remain limited to producing lower-end semiconductors and not the advanced chips integral to China’s AI development, which are subject to the new U.S. export controls.
China has set 2030 as its target date to become a global leader in artificial intelligence, with the subsequent goal of putting the People’s Liberation Army on par with the U.S. military by 2035 — a goal the export controls intend to complicate.
“This severely limits China’s ability to access advanced nodes or build its own capabilities,” said Jay Goldberg, a specialist on Asia’s electronics supply chain and the chief executive of D2D Advisory, a consulting firm. “I don’t see any way that they can build their own advanced processors for like a decade. There’s no avenue there.”
The Biden administration in December added Yangtze Memory Technologies Corp. and 21 other Chinese companies involved in the nation’s AI microchip sector to the Commerce Department’s export control list.
That same day, Congress passed the fiscal 2023 National Defense Authorization Act, including a provision banning U.S. defense companies from using Chinese semiconductors in critical weapons systems — legislation that will force them to keep better tabs on their lower-tier suppliers.
Coupled with the export controls, “it’s a one-two punch,” Majority Leader Chuck Schumer, D-N.Y., who introduced the semiconductor provision in the defense bill, said in December on the Senate floor.
“The 21st century will be largely shaped by our two giant economies,” he added. “For decades, the Chinese Communist Party has cheated their way to the top by piggybacking off American technologies and American [intellectual property], with huge implications for our national security and for our edge in AI, cybersecurity, telecom and other major technologies.”
The groundwork for restricting Chinese exports dates back to the second half of former President Donald Trump’s term. The administration’s export controls prohibited two Chinese microelectronics producers — Huawei and the Semiconductor Manufacturing International Corp. — from obtaining U.S.-produced semiconductor equipment and software.
While the U.S. is no longer a major player in producing semiconductors — something Biden hopes to rectify — American companies still play a leading role in designing these chips and exporting the equipment needed to physically manufacture them. This gives the U.S. significant leverage over the global semiconductor supply chain.
Biden’s export controls greatly expand the Trump-era restrictions; they require U.S. microelectronics companies, or foreign businesses that use American technology, to obtain licenses from the Commerce Department to sell to Chinese firms the equipment needed to produce the most advanced semiconductors — licenses Washington is unlikely to grant.
The restrictions also block blacklisted Chinese microchip companies from employing American citizens; those who remain with the companies risk losing their citizenship.
On aggregate, these restrictions strike a major blow to China’s ability to produce supercomputers needed to generate AI algorithms.
Goldberg said the export controls hinder the Chinese military’s ability to build “advanced capabilities into their systems.”
“Especially for the Chinese military, the really limited things will be around supercomputing, which is really important for weapons design, for nuclear weapons simulation, advanced weather modeling, those kinds of things,” he explained.
China has the largest number of the world’s most powerful supercomputers. It surpassed the U.S. in quantity in 2016, according to Top500, which tracks the world’s most powerful supercomputers. These computers are necessary for developing artificial intelligence algorithms, and they can also simulate nuclear and missile tests. But they require the advanced semiconductors now subject to U.S. export controls.
For instance, China seeks to develop a new military construct it calls Multi-Domain Precision Warfare, which is roughly analogous to the Pentagon’s Joint All-Domain Command and Control initiative. These systems require advanced AI algorithms and supercomputing to rapidly sift through data that informs battlefield decisions under very compressed time frames.
The restrictions will also hamper China’s ability to use supercomputers for dual-purpose activities like space exploration, vaccine development and weather modeling.
Jon Bateman, a senior fellow at the Carnegie Endowment for International Peace who previously served as a special assistant to former Chairman of the Joint Chiefs of Staff Gen. Joseph Dunford, said the export controls effectively amount to an economic policy of “quasi-containment.”
“With each year that goes by, AI algorithms at the cutting edge require more and more computing power, and therefore need more and more advanced chips,” Bateman told Defense News. “If the rest of the world moves forward and China can’t, then it’s going to face bigger and bigger problems.”
But Bateman said he’s “skeptical about the immediate impact” of the export controls on China’s defense-industrial base, noting they wouldn’t thwart a possible Chinese attack on Taiwan within the next few years.
China considers Taiwan a breakaway province, and Beijing has said it will take back the island, by force if necessary. Such an attack would rely on more conventional weapons systems that use less advanced semiconductors.
“If there were to be war over Taiwan in the next five or 10 years, the outcome of that war wouldn’t depend on artificial intelligence,” Bateman said. “It would depend on China’s amphibious landing capabilities; it will depend on China’s political will — really traditional stuff.
“China has a lot of the chips that it needs already and probably has a certain level of stockpiling and will be able to do a lot of the same things with chips that fall slightly below the threshold for export controls just by buying more of them.”
Suspicions of sabotage
China does not yet have the capacity to domestically mass manufacture the world’s leading semiconductors, but it still produces plenty of cheap, less-sophisticated microchips — many of which it exports to the United States.
The National Defense Authorization Act, which became law in December, includes Schumer’s provision that will eventually ban U.S. defense companies and other federal contractors from using these semiconductors in critical national security systems. However, it remains unclear to what extent Chinese chips are currently used in U.S. weapons.
But even low-end semiconductors could potentially disrupt U.S. weapons systems.
“China’s capable of building low-end semiconductors, and those tend to be low-value, overlooked chips that sell for a few dollars or a few pennies a piece,” Goldberg said. “From a national security perspective, I think we should be paying more attention to that and being more concerned about that.”
Proponents of the ban fear a faulty semiconductor could sabotage an entire weapons system, or that Chinese chips could give Beijing access to technology, systems and classified information relevant to U.S. national security.
The congressionally mandated U.S.-China Economic and Security Review Commission’s 2022 report to Congress warned “low-quality, counterfeit microelectronics” could compromise U.S. military equipment. For instance, Air Force Times reported in September that several faulty semiconductors in a malfunctioning ejection seat — which caused the 2020 death of F-16 pilot Lt. David Schmitz — may have been counterfeits, though it’s unclear if they came from China.
The commission’s report also found “the United States faces exposure to backdoor vulnerabilities from integrated circuits that undergo final assembly, packaging and testing in China.” It noted 38% of the world’s semiconductor assembly, packaging and testing occurred in China as of 2019.
“I don’t think anyone has seen active exploits out in the wild, but they’re theoretically possible,” Goldberg said.
Some defense industry groups raised the alarm over an initial version Schumer’s legislation, arguing it was overly broad and that the implementation period was too short.
Schumer and the other senators behind the legislation, including Republican Roger Wicker of Mississippi, significantly revised the language before Congress passed the defense bill last month.
“There’s no question that when Chinese-owned companies do the manufacturing, there is a security risk which allows them to gain sensitive information that they have no business knowing,” Wicker told Defense News. “I don’t think that’s subject to debate. The only question has been the practicality of getting that done in a short time frame.”
The version that became law expands implementation from two to five years, only bans Chinese semiconductors from critical systems, and allows the Pentagon and other agencies to issue waivers. It also requires the Pentagon and several other departments to “establish a microelectronics traceability and diversification initiative” with industry.
The Pentagon is currently weighing a zero-trust policy that would assume no microelectronics are safe, thereby requiring validation for every chip in the supply chain, Defense News reported in December.
But defense contractors often have trouble tracing equipment down to the lowest rungs of their supply chains as in the case of Lockheed Martin’s F-35 aircraft, which used a Chinese-sourced cobalt alloy in its Honeywell-made turbo-engine. The cobalt alloy posed no risk to the F-35, but it did violate so-called buy American laws, prompting the Pentagon to grant a waiver for the fighter jets.
Lockheed Martin CEO Jim Taiclet noted the issue took 15 years to identify.
“Three or four layers of suppliers down, they will be sourcing from distributors. Small parts, microelectronics, magnets … those come in pools,” Taiclet said in early December at the Reagan National Defense Forum. “It’s hard to figure out where the commercial distributor that sells [to] the fourth-level-down supplier got that magnet material.”
Still, Goldberg said, the problem comes down to “people’s willingness to do the work.”
“If they have a concern about counterfeit products, that’s just a basic supply chain problem,” he added. “Forget about national security. If you’re building a system, you should actually be deliberate in not getting counterfeit parts.”
In response to U.S. efforts, China is trying to lessen the dependence of its defense-industrial base, economy and AI development on American technology. Beijing is preparing a $143 billion package to bolster its own semiconductor industry, Reuters reported in December.
Gary Clyde Hufbauer and Megan Hogan, both fellows at the Peterson Institute for International Economics, wrote in December that Chinese alternatives to U.S. advanced semiconductor technology are likely “decades away.” In Australian National University’s East Asia Forum publication, they added that Beijing’s investments will help it eventually procure this technology despite export controls.
“Many engineers and computer scientists are likely to be assigned to semiconductor design and manufacture, assisted by espionage against U.S., South Korean, Taiwanese, Japanese and European chip firms,” they wrote. “If China needs more advanced chips for AI-driven weapons systems, it can likely produce them, albeit at a very high cost. Many semiconductor industry experts agree that China has the technical capability to produce cutting-edge chips yet lacks the commercial capability to scale up production.”
Goldberg said that even with export controls, “there’s going to be some leakage” that China can use to circumvent U.S. restrictions. “Chinese companies are really good at setting up byzantine corporate shell structures,” he added.
China in December also launched a dispute with the World Trade Organization against the Biden administration’s export controls. Taiwan on Jan. 5 joined the consultations on the dispute, but stressed in a statement to Bloomberg News that doing so does not indicate “any dissatisfaction with the United States’ measures.”
Meanwhile, the U.S. is in talks with Japan and the Netherlands — both of which boast prominent microelectronics industries — in a bid to persuade them to instate their own export controls that would bar China from using Japanese and Dutch technology.
Still, most of the physical semiconductor production remains concentrated in East Asia, raising the prospect of a global economic crisis and trouble for the U.S. defense-industrial base should a war occur over Taiwan or elsewhere in the region.
“Almost all the production is moving to three places: China, South Korea and Taiwan,” said Taiclet, the Lockheed CEO. “The real issue of a nightmare scenario is not of the defense-industrial base alone, but for the entire economy.”