US Considers Chip Tracking to Curb AI Hardware Smuggling to China

The U.S. and China are engaged in a race over chip development -- both countries are racing to build infrastructure

Both superpowers are building up chip development capabilities while trying to reduce dependencies upon one another. The U.S. and China are locked in an unprecedented race over semiconductor technology, driven by national security concerns, economic competitiveness, and the belief that control of advanced chips will determine global leadership in the decades ahead. Chips are the foundation of everything from smartphones and data centers to fighter jets and the energy grid, as well as powering the AI tech both Washington and Beijing see as critical for future military and economic power. The rivalry has intensified in recent months as both sides have recognised vulnerabilities in their supply chains. The U.S., home to most chip design and high-end equipment companies, grew wary of depending on global manufacturing concentrated in East Asia - especially as China poured hundreds of billions into building its own chip industry, and constantly threatens invasion of Taiwan, where over 60% of global semiconductor production and more than 90% of the most advanced chips (below 7nm) passes through. China, in turn, saw its reliance on foreign (often American or U.S.-allied) technology as a strategic weakness, especially after repeated sanctions blocked its access to the most advanced chips and manufacturing equipment. This escalating contest has led to historic subsidies, new export controls, technical bans, and a scramble to onshore production. We reached out and interviewed multiple industry analysts to get a further understanding of how this dynamic could impact the industry. The two countries are engaged in an arms race to see who can stand up their own chip development first, which has resulted in a global divide between major superpowers. While neither side can yet claim decisive victory, their race is already reshaping trade, geopolitics, and the global tech landscape. "We're definitely seeing more bifurcation in the chip industry, as China pours money into building a self-sufficient supply chain," says Chris Miller, professor at Tufts University and the author of Chip War: The Fight for the World's Most Critical Technology. It's an assessment that William Matthews, senior research fellow for China and the world in the Asia-Pacific Programme at Chatham House, agrees with. "I think the long-term trend will be toward bifurcation as China catches up. Beijing is clear about its desire for independent technological capability across the board - I'm not sure we're on the 'brink' of this, but that does seem to be the direction of travel." But both believe the rupture hasn't yet arrived. Each country remains dependent on the other - at least for now. "China is still reliant on importing chipmaking tools and materials, especially for high-end production," explains Miller. China relies on imported lithography equipment, particularly EUV machines from ASML, which are restricted by U.S.-led export controls. It also depends on foreign EDA software and ultra-pure materials essential for leading-edge chips. These tools and technologies are extraordinarily complex, protected by IP, and take decades to master - areas where domestic alternatives are still underdeveloped. As a result, China cannot yet manufacture chips at the most advanced process nodes ( 5nm or below) at scale or efficiency. "I think the export control has been effective in controlling China's access to advanced chipmaking from TSMC, Samsung, and Intel and advanced chipmaking tools from ASML," says Ray Wang, research director for semiconductors, supply chain and emerging tech at The Futurum Group. It's an approach that some of the U.S. president's most senior advisors have given full-throated support to. "Restricting the sale of EUV lithography equipment to China is the single most important export control we have," David Sacks, the administration's AI and crypto czar, posted on social media in May. And beyond those struggles, the reality is that China is still a chip importer of its own. "It is also still importing large volumes of high-end chips because of their struggles to ramp up production capacity," says Miller. As of 2024, China's monthly chip production capacity reached around 8.85 million wafers, with most output coming from mature-node fabs (28nm and above). This figure is forecast to rise to 10.1 million wafers per month by the end of 2025, largely driven by government funding and 18 new fabs under development. But that's changing. Market research and tech consulting firm Yole Group predicts that China will hold 30% of the world's global foundry production capacity by 2030, making it the world's largest hub of semiconductor production. In 2024 alone, it bought 40% of all wafer fabrication equipment worldwide - recognising it needed to step up its production. However, when that does happen, and China manages to ramp up production, there will be significant changes for us all. "It will have global implications - and be yet another example of China successfully innovating and catching up under pressure from the U.S., while also exploiting a market for its own technology abroad, especially across the Global South," says Matthews. "I suspect we will see a pattern which is now quite familiar, of growing Chinese indigenous capability leading to deeper separation of ecosystems as the U.S. responds by attempting to restrict China further." That's concerning experts who are studying the area. "With U.S.-China tensions rising and technological competition intensifying, a deeper decoupling of the semiconductor supply chain now looks inevitable," says Wang, "spanning chip design, manufacturing, equipment, and other critical segments. If U.S.-China tensions persist - or deteriorate further - the global semiconductor supply chain could ultimately and completely split into two parallel ecosystems, which is occurring in front of our eyes." There are already signs of that happening. Washington is choking China's access to the tools used to design and make advanced chips. A May order from the US Commerce Department revoked existing licences for shipping electronic-design-automation (EDA) software and critical chemicals to Chinese customers, forcing vendors such as Cadence (which was fined $140 million for exporting design tools to China) and Synopsys to apply for case-by-case permits. In return, China had initially refused to approve, but then backed down on, a merger of Synopsys and Ansys, two of the largest providers of EDA software, which is essential for designing the architecture and logic of semiconductors before they are manufactured, in the world. These tools enable simulation, verification, and layout of complex chip designs, making them critical to advanced semiconductor R&D and production. China is responding to the threat of being squeezed out of the supply chain by backing its own homegrown production arms to the hilt. They're putting the full force of Chinese finance and regulation behind the plans. Big Fund III, a third-phase, state-backed war chest, aims to deploy ¥344 billion ($48 billion) into the country's weakest links of lithography and design software. The fund plans longer holding periods and aggressive consolidation to accelerate breakthroughs, according to people familiar with its mandate. Alongside that, Chinese leadership is also putting the power of the state to work in favour of the AI sector. The forthcoming 15th Five-Year Plan (2026-30) will have AI development at its core. Early drafts flagged by Chinese state media put semiconductor equipment self-reliance near the top of national priorities, alongside the "AI Plus" push to embed artificial intelligence across every sector of the economy. Despite those hints at the future direction, exact details are difficult to discern. "It's hard to interpret exactly how China's industrial policy is evolving because so much of its industrial policy spending happens at the provincial level and local level, as well as various public-private funds," says Miller. "Still, it does seem like the new iteration of the Big Fund is focusing more on EDA and chipmaking equipment. However, these are very challenging segments to enter, and it will take years for China to produce cutting-edge capabilities in these spheres." While China's chipmaking ambitions are indeed large, it has not been entirely successful in its efforts thus far, with a large number of projects and fabs entirely collapsing, despite lofty claims. What the ongoing feud between the U.S. and China means for the rest of the world and Europe is particularly vexing for experts. "That is a risk for Europe - again, following a pattern seen in other technologies from EVs to robotics," says Matthews. The problem is that "Europe has been consistently late in waking up to China's capability and challenge to the U.S.," he explains. In response, Europe ought to try and carve out niches in order to maintain competitiveness. "But that means massive focused investment and a strategy more like China's," Matthews says. "The question would be whether Europe has the political will to pursue that over the long term." Matthews argues there is a strong case for Europe to pursue chip production even if it lags behind the U.S. and China in order to strengthen its own autonomy and resilience - as it has in other areas, including its tech stack, with the development of proprietary versions of Google Docs and Microsoft Office - "but again, that needs to be accompanied by a real mindset shift on industrial policy and a plan to engage countries in the Global South as customers," Matthews explains. "The challenge will be catching up with China on these two counts." Still, given the EU's recent history in making sure it's in control of its own supply chain and regulation, and not at the whims of the United States, it seems likely. "Europe is probably going to consider a second EU Chips Act to build its semiconductor capabilities, which will probably focus on some of these unique chipmaking capabilities, as well as its niches in materials and equipment," says Miller. At the same time, Europe will have to compete in the global market for expertise and manufacturing tools. And the U.S. is making sure it's as competitive as possible to head off China. Five years ago, Washington wagered that a 25% investment tax credit would be enough to lure chipmakers home. It worked - up to a point. But a June Senate draft lifted the subsidy to 30%, and President Donald Trump's "One Big Beautiful Bill," signed in July, took it to 35% for fabs breaking ground before 2026, with sweeteners for AI-specific capacity. Industry has responded in kind: in March, before the Big Beautiful Bill was passed, Taiwan Semiconductor Manufacturing Co. (TSMC), one of the world's biggest names in the sector, lifted its financial commitment to onshore chip manufacturing capacity in Arizona from $65 billion to $165 billion, cementing the single largest foreign-direct investment in U.S. history. "AI is reshaping our daily lives and semiconductor technology is the foundation for new capabilities and applications," said TSMC Chairman and CEO C.C. Wei in a statement. "With the success of our first fab in Arizona, along with needed government support and strong customer partnerships, we intend to expand our U.S. semiconductor manufacturing investment by an additional $100 billion, bringing our total planned investment to $165 billion." "It's clear to me that TSMC in Arizona has been quite successful despite the early hurdles, and its P1 - the first fab in Arizona - is in fact in operation," says Wang. The P1 facility is expected to produce up to 30,000 wafers per month using 4nm and 5nm process technology, primarily for high-performance computing clients like Apple and Nvidia. "I expect its P2 and P3 will have [an] easier learning curve, and construction-to-operation process should be smoother." As for other investment from Texas Instruments, which announced a $60 billion U.S.-focused tech pledge in June that it said would create 60,000 jobs, and Micron, which has committed up to $100 billion to build the largest semiconductor fabrication facility in the U.S. in New York, focusing on memory chip production, with initial construction starting in the late 2020s, it's "too early to say," reckons Wang. The two countries are caught up in a delicate dance to try and one-up each other, while ensuring they keep access to the things on which they are reliant on one another for. In mid-July, the United States loosened its restrictions on the export of Nvidia's H20 chips, seemingly in exchange for China reallowing the export of rare earths that are vital in other chips' manufacturing. Until one side or the other can fully onshore all the required production processes, there's an awkward quid pro quo going on.

White House considering chip tracking to curb AI hardware smuggling to China amid enforcement gaps -- software or hardware tracking could be next step in U.S. export controls over leading-edge AI silicon

The U.S. is weighing a new approach to protect its lead in artificial intelligence: embedding location-tracking technology directly into high-end chips. The move comes as years of export controls -- and recent escalations -- have failed to fully stop smuggling into China, leaving policymakers looking for solutions that go beyond paperwork. This involves leading-edge AI GPUs such as Nvidia's H20, which are otherwise already permitted to be sold in China following lengthy bans. Talking to Bloomberg, Michael Kratsios, Director of the White House Office of Science and Technology Policy and one of the architects of the administration's AI action plan, confirmed that both software-based and physical tracking solutions are being discussed. "There is discussion about potentially the types of software or physical changes you could make to the chips themselves to do better location-tracking." The idea was explicitly included in the plan unveiled last month, which aims to keep U.S. technology dominant as AI adoption accelerates globally. The urgency reflects a persistent problem: export controls aren't closing the gap. Recent reports show Chinese companies obtained over $1 billion worth of Nvidia GPUs through smuggling in just three months, despite tightened U.S. restrictions. The challenge is compounded by third-party diversion routes through Southeast Asian nations such as Malaysia and Thailand, which U.S. officials are now scrutinizing as potential hubs for illicit shipments. Building location awareness into chips is meant to address these loopholes, but it comes with technical and commercial hurdles. Unlike smartphones or laptops, AI accelerators operate in data centers where GPS tracking is impractical. According to the report, sources familiar with early discussions suggest that delay-based verification systems -- which confirm location via secure time-stamped signals rather than constant connectivity -- are under consideration. These methods reduce reliance on external networks but introduce design complexity and must avoid any performance hit in high-compute environments. Industry concerns go beyond engineering. Adding tracking functionality could raise costs, create new attack surfaces for hackers, and even spark geopolitical retaliation. If U.S. policy mandates embedded monitoring, other regions may impose reciprocal requirements, further fragmenting global supply chains. That risk is real as Beijing recently summoned Nvidia officials over alleged "tracking functions" in its H20 chips, claims the company strongly denies. At the same time, enforcement debates have exposed differing views in the industry. Nvidia CEO Jensen Huang has publicly criticized U.S. chip restrictions as a "failure", warning that overregulation accelerates China's domestic chip development and erodes America's long-term advantage. This perspective underscores the delicate balance Washington faces -- controlling technology flow without pushing rivals to innovate faster. The scale of the problem leaves little room for complacency. China remains the largest global consumer of semiconductors and dominates legacy chip production, holding around 30% of the market and expected to account for 40% of future expansion through 2030. Even small percentages of diverted AI chips can have a significant impact, especially as they power advanced models and military systems. For now, hardware-level tracking represents Washington's most aggressive -- and controversial -- enforcement idea yet. Whether it becomes reality will depend on technical feasibility, industry cooperation, and the geopolitical fallout. But the message is clear: future breakthroughs in AI should run on American hardware, and under American oversight.

US Explores Better Location Trackers for AI Chips, Official Says

The US is exploring ways to equip chips with better location-tracking capabilities, a senior official said, underscoring Washington's effort to curtail the flow of semiconductors made by the likes of Nvidia Corp. to China. Washington has espoused working with the industry to monitor the movements of the sensitive components, part of a broader plan to curtail smuggling and ensure American technology remains dominant. Last week, Beijing summoned Nvidia representatives to discuss US efforts around location-tracking and other alleged security risks related to its H20 chips.

Uncle Sam floats tracking tech to keep AI chips out of China

Plan would embed location verification in advanced semiconductors to combat black market exports The Trump administration wants better ways to track the location of chips, as part of attempts to prevent advanced AI accelerator hardware from getting into Chinese hands. Washington wants to equip semiconductors with location-tracking capabilities, and is keen on working with the industry to accomplish this, according to a senior White House official. The purpose is to enable the US to follow where shipments of key products for AI development, such as Nvidia GPUs, are actually ending up, so as to aid efforts to prevent smuggling of the components into China. "There is discussion about potentially the types of software or physical changes you could make to the chips themselves to do better location-tracking," Michael Kratsios told Bloomberg. Kratsios is understood to be one of those behind the White House's AI Action Plan unveiled last month. However, he said he has not so far spoken directly with either Nvidia or AMD, the other major supplier of advanced GPUs used in AI training. The move follows the introduction of legislation in the US Senate and the House of Representatives in May to require the Department of Commerce to mandate that certain advanced chips (or products containing them) be equipped with a "location verification mechanism" to detect if shipments are being diverted after export. Neither of the two bills specified exactly how the tracking mechanisms would be expected to work. However, exporting companies would be responsible for reporting violations to the Commerce Department's Bureau of Industry and Security (BIS) in the event tracking is tampered with or the chips show up in places they shouldn't. Such a move is unlikely to play well with Beijing, which has already grilled Nvidia over allegations of backdoors in the company's designed-for-China H20 GPU chips that might allow US security services to monitor or remotely disable them. America's ever more elaborate attempts to curb China's AI expertise are perhaps based on claims that it is just a year or two behind the US in such technology, and recent reports that an estimated $1 billion worth of high-end Nvidia GPUs have made it to the Chinese black market, despite Washington's strict export regulations. In a related move, Senators Elizabeth Warren, D-Mass, and Mike Rounds, R-South Dakota, wrote to Commerce Secretary Howard Lutnick and Secretary of State Marco Rubio urging the administration to retain strict rules that discourage companies from offshoring their AI infrastructure. The letter [PDF], sent July 31, argues that America's edge in AI stems not just from software innovation but from the massive domestic buildout of datacenters and GPU clusters powering AI development. "AI is infrastructure," they wrote, warning that offshoring this infrastructure - tempted by foreign subsidies - could shift the center of gravity for frontier AI development away from the US and into the hands of rivals like China. The senators are calling for the forthcoming replacement to the administration's Diffusion Rule to impose robust security requirements on any overseas AI facilities and to ensure the bulk of advanced AI compute remains on US soil. ®

China draws red lines on US chip tracking with Nvidia meeting - The Economic Times

As the US and China look for any sort of leverage in a prolonged trade fight, Beijing sees an opportunity to win over the world by taking a stand against the Trump administration's plans to track high-end chips. Chinese internet regulators last week summoned Nvidia staff over alleged security risks with its less-advanced H20 chips. The action, citing calls from US lawmakers to build tracking features into the most powerful semiconductors, has yet to lead to any type of formal ban or restrictions. Either way, analysts see the move as not so much about the H20s, which Chinese state-backed entities have publicly employed for some time, but rather an easy way for Beijing to send a series of messages about the US plans: Domestic firms should be cautious, the world should be wary and Nvidia CEO Jensen Huang should influence the White House to shift course. "The recent summons of Nvidia serves as a warning for Nvidia's future products rather than a sign that the Chinese government found any loophole in H20," said George Chen, partner and co-chair of digital practice at The Asia Group, which was co-founded by former US Deputy Secretary of State Kurt Campbell. "China wants to use the Nvidia case to show China is a buyer, but it won't be a blind buyer." For now, the spat looks unlikely to blow up the wider US-China relationship. Beijing said the two sides agreed to maintain a tariff truce after talks last month in Stockholm, while Treasury Secretary Scott Bessent later said it's up to US President Donald Trump to make that call. "We're getting very close to a deal," Trump said in an interview with CNBC on Tuesday. "We're getting along with China very well." Michael Kratsios, one of the architects of a White House action plan on AI that calls for exploring chip-tracking technologies, told Bloomberg Television on Tuesday that officials are discussing the use of software or physical changes to better track restricted chips. He added that he's not had conversations "personally" with either Nvidia or Advanced Micro Devices Inc. about exploring location-tracking technology. Nvidia on Tuesday declared its opposition to any sort of backdoors, saying they enable hackers and undermine trust in US technology. "There are no back doors in Nvidia chips. No kill switches. No spyware," Nvidia said in a blog post. "That's not how trustworthy systems are built -- and never will be." The Chinese Cyberspace Administration's action suggests Beijing is drawing a broad line against any surveillance capabilities in American semiconductors, a position that may resonate around the globe, even with US allies. Trump's first administration warned governments to avoid using equipment from Huawei Technologies Co. over risks that China could use it for spying. "We started attacking Huawei because of the idea that there are secret backdoors in it, and now here the US is openly suggesting we should legally mandate backdoors in hardware that we sell. It's a huge deal," said Tom Nunlist, associate director at the Beijing-based consultancy Trivium. "What government would accept this?" The H20s have become a focal point in the broader debate over US export controls on China after American officials claimed that they allowed Beijing access to the chips as part of earlier trade talks in London. Trump's move to lift an earlier ban on their exports generated criticism from more hawkish lawmakers, who argue that the chips, while a diminished version of years-old Nvidia technology, will help China compete in AI. Commerce Secretary Howard Lutnick defended the decision, saying the US wanted to "sell the Chinese enough that their developers get addicted to the American technology stack." China's Commerce Ministry disputed the US version of events in a statement last month, saying the US "proactively" approved the sale of H20 chips and suggesting they weren't part of any wider tradeoff in return for rare-earth magnets. China views the H20s as on par with domestic offerings, even though it could still use them because local companies can't churn out enough AI chips to meet demand. "Yes, China does want the H20," said Ray Wang, a Washington-based research director focusing on semiconductors at The Futurum Group, citing significant purchases by leading tech companies such as ByteDance Ltd., Tencent Holdings Ltd. and DeepSeek before the US cut off access to the chips in mid-April. "They clearly prefer to have access to the H20." The Chinese Commerce Ministry didn't respond to faxed questions. Chinese state media has turned up the scrutiny on imported chips, with a commentary in the ruling Communist Party's flagship mouthpiece People's Daily calling devices with location tracking "infected." A Sunday editorial by China Daily dismissed the H20 as "castrated" and offered a different reason for the US policy reversal. "It is China's breakthroughs in producing its own AI chips that has prompted the US to lift its curbs on the exports of H20 chips just three months after they were banned," the newspaper said. Shares in Chinese chipmakers including Semiconductor Manufacturing International Corp. and Cambricon Technologies Corp. rose immediately after Beijing disclosed the meeting with Nvidia employees, as investors bet on homegrown alternatives. "It's a straightforward option for China to now put Nvidia on the negotiation table, either to trade for more supply security promises or to further push domestic substitution," said Tilly Zhang, a technology analyst with Gavekal Dragonomics. "Either way, it wouldn't be a loss from Beijing's point of view." While the two sides reached a truce that allowed the US to access rare-earth magnets, which are needed to make high-tech goods including smartphones and missiles, a final deal has yet to be worked out. In an interview aired over the weekend, US Trade Representative Jamieson Greer said both sides are "about halfway there" on easing China's export controls on rare earths. The desire for the chip-tracking technology stems from the US struggle to enforce export controls around the world. A proposed Chip Security Act, introduced to the House of Representatives in May, would require location-verification mechanisms on more advanced chips like Nvidia's H100 and B200, but not the H20. One possible method is "delay-based" location verification, which measures the time it takes for a signal to travel from trusted servers to target equipment to determine its location, according to analysis by Bloomberg Intelligence. Whether the US will press ahead with new mandates on chips remains an open question. Trump's desire for a deal with China means further curbs on chips are unlikely before the expected summit this fall, according to Chris Miller, professor of international history at Tufts University and author of Chip War. "The administration has many priorities and it's hard to see which is going to win out," he said. "It's very clear that the White House is going to try to balance the hawks' desire for restrictions with the broader US-China relationship."