Trump and the difficult balance between geoeconomics, national security, and economic freedom. The Nvidia case could be the litmus test

KEY TAKEAWAYS

• Last month, Jensen Huang, head of Nvidia, sent a clear message to the Trump administration: the company that produces the world's leading artificial intelligence (AI) chips intends to “unreservedly serve the Chinese market” in AI.

• A few days earlier, the U.S. government further banned Nvidia from even selling its latest H20 microprocessor to China, marking the first major limits the second Trump administration has imposed on the technology.

• In fact, after Biden decided on Nvidia's export restrictions in 2022, the company was forced to modify even the H100 chip so that it would not fall under the restrictions, with the result that its evolution, the H20 precisely, was sold only in China.

• The recent restrictive measures came amid rising geopolitical tensions between the United States and China, characterized by a growing trade war and Washington's fear of losing supremacy in artificial intelligence to Beijing.

• But despite all this, Nvidia chips still continue to reach Chinese AI developers through various channels, such as parallel smuggling markets or through the use of cloud of shore data centers.

• Nvidia, meanwhile, redesigns the H20 chip to circumvent limits on the export of technology between the United States and China.

• In January, the Biden administration unveiled its “AI Diffusion Framework,” but the Trump administration is reportedly planning to amend this legislation, perhaps tying access to its chips to broader trade negotiations with Beijing.

• Now a member of the U.S. Congress, he is also set to introduce a bill aimed at monitoring AI chips, including Nvidia's, to prevent their illegal introduction into China.

• The United States will not achieve dominance in artificial intelligence if its companies cannot access global markets, said ITIF, a U.S. think tank.

• Michael Frank , a member of the 2430 Group, believes that “America should embrace interdependence as a weapon against China.” The more the United States and the West try to block Chinese technological development with limiting measures, the more China will be forced to accelerate the pace of independent innovation, the expert says.

• If it is difficult for allies to buy U.S. chips, they will get them from other suppliers, including the Chinese, writes the WSJ.

• Indeed, Beijing is allocating tens of millions of dollars in subsidies to spur the growth of the city's artificial intelligence supply chain, using entirely domestic technologies, as China invests heavily in its push for technological self-sufficiency. The Yizhuang Development Zone, also known as the Beijing Economic and Technological Development Zone, said Wednesday it plans to create a nation-leading artificial intelligence industry ecosystem worth 80 billion yuan ($11 billion) by the end of this year, according to a plan shared on its official WeChat account. The ecosystem would rely exclusively on Chinese semiconductors, operating systems and open source software frameworks to ensure self-sufficiency and control, policymakers said.

• What to do. Trump's task is damn complicated. Finding a synthesis between geoeconomic interests, national security and economic freedom is not easy. The risk of sinning by excess is real. However, even businesses need to understand that the Trump Administration faces a dangerous “enemy” that uses sophisticated techniques and tactics and has been waging its own liminal war for years, that is, an incremental war, in which the spectrum of competition and confrontation with the West is so broad that the battlefield is everywhere and the war is total. And it is within this logic, control of technological assets, strategic infrastructure and major global supply chains, integrate “trans-military” and “non-military” warfare operations by Beijing. To make states more dependent on Beijing and more willing to “a new authoritarian world order with distinctive Chinese characteristics,” Liminal Warfare encompasses economic, diplomatic, legal, military, intelligence, and cyber operations put in place seamlessly. Regulations need to be adjusted and new alliances found among like-minded countries, securing their economies. It is not a matter of shutting down relations with China; however, modernization of regulatory processes becomes the main requirement to be able to defend oneself. It is necessary, in essence, to work for the regulation of the new globalization that awaits us, keeping the definition of the security perimeter of the United States at the center of political action.

  
  

Last month, Jensen Huang, head of Nvidia, sent a clear message to the Trump administration that the company producing the world's leading artificial intelligence (AI) chips intends to “serve the Chinese market” for AI without reservation.

Huang, in particular, said the loss of this market would represent a huge loss, in an industry set to grow significantly in the coming years, as the chip company faces increasingly stringent export restrictions imposed by Washington.

GettyImages

In fact, just days earlier, the U.S. government further banned Nvidia from selling even its latest H20 microprocessor to China, marking the first major limits that the second Trump administration has imposed on the technology.

After Biden had decided export restrictions on Nvidia in 2022, the company, in fact, was also forced to modify the H100 chip so that it would not be covered by the restrictions, resulting in its evolution, the H20 precisely, being sold only in China.

The recent restrictive measures came against a backdrop of rising geopolitical tensions between the United States and China, characterized by a growing trade war and Washington's fear of losing supremacy in artificial intelligence to Beijing.

But despite all this, Nvidia chips equally continue to reach Chinese AI developers through various channels, such as parallel smuggling markets or through the use of cloud of shore data centers.

And it should also be added that, despite export controls, the problem of smuggling of sophisticated U.S. technology continues to persist, not only with reference to semiconductors where at stake are potential military as well as commercial applications.

In January, the Biden administration unveiled its “AI Diffusion Framework” Specifically, this rule:

• places country-specific caps, along with a licensing regime on semiconductor exports that would particularly impact GPU chips that underpin key AI applications, such as the development of large language models (LLMs);

• determines where to ship U.S. chips essential for AI;

• determines where to build the data centers that create AI, with a preference for the United States and its allies.

However, the Trump administration is reportedly planning to amend this legislation, perhaps tying access to the chips to broader trade negotiations with Beijing, but has not yet announced any changes.

A U.S. congressman is also set to introduce a bill aimed at monitoring AI chips, including Nvidia's, to prevent their illegal introduction into China.

Nvidia's appeal

China's artificial intelligence market is expected to reach $50 billion in the next two to three years, Huang said. “It would be a huge loss not to be able to manage it as an American company,” he concluded.

Selling to China would also help bring revenue back to the United States, helping to pay taxes and helping to “create a lot of jobs,” he added. “The world today is dynamic. You just have to remain flexible.”

As mentioned, the remarks come as the U.S. semiconductor giant grapples with the impact of U.S. export controls that were recently extended to its H20 AI chips tailored for China where Big Tech, including Tencent Holdings and ByteDance, currently use these chips to develop and train their AI models.

It was in early May, during a closed-door meeting with U.S. lawmakers, that Huang also expressed concern about Huawei Technologies' progress in AI.

The discussion with the U.S. House of Representatives Foreign Affairs Committee concerned Huawei's AI chips, along with the potential effects of U.S. restrictions on Nvidia chips in China.

A congressional staffer pointed out the risk of increased global demand for Huawei chips if future artificial intelligence models are optimized for that company's hardware.

Nvidia's CEO also called for a review of U.S. regulations on AI technology exports, stressing the need for U.S. companies to expand artificial intelligence globally, while noting uncertainty about the Biden administration's proposed tiered qualification system with the Export Control for Artificial Intelligence (AI) Diffusion.

How Nvidia adapted to multiple bans

In recent years, America has tried to thwart China in the race for artificial intelligence by controlling access to its advanced semiconductors.

Nvidia began selling H20 chips to China in early 2024 after its advanced A100, H100, A800 and H800 AI microprocessors were subjected to U.S. export controls to address national security concerns.

To comply with the new restrictions and not lose the burgeoning Chinese market, Nvidia decided to offer modified versions of its chips. Therefore, soon after the H2O model, custom-designed for Chinese customers in accordance with previous export regulations, was also banned from export, Huang made a surprise visit to China. Before the trip, the company had announced that these latest restrictions would cost it $5.5 billion.

Nvidia redesigns H20 chip to circumvent U.S.-China technology export limits

Nvidia will launch a technically downgraded version of its H20 artificial intelligence chip for the Chinese market within the next two months. This strategic turn comes after US officials effectively blocked the original model by requiring an export license, according to multiple industry sources.

Technical specifications for the new chip reveal significant performance compromises compared to the original design. The modified H20 will feature substantially reduced memory capacity, though some downstream customers may potentially adjust module configurations to fine-tune performance levels.

The smuggling of Nvidia chips

Despite bans, Nvidia chips continue to get into the hands of Chinese AI developers through various methods. According to an investigation, this is mainly done by renting data centers offshore, or by purchasing chips through opaque middlemen who deal directly in China.

The latter, in particular, exploit a supply chain set up to circumvent sanctions, with microprocessors shipped through third countries not covered by U.S. restrictions. In essence, goods pass through multiple jurisdictions and front companies to conceal their origin, with export documents forged and products mislabeled to evade customs.

Erich Grunewald of the Institute for AI Policy and Strategy, a San Francisco-based think tank, estimates that smuggled U.S. chips accounted for between one-tenth and one-half of China's AI model training capacity last year.

Another “gateway” to bypass U.S. restrictions is to lease data centers off shore. Johor, an area in southern Malaysia located on the border with Singapore, lends itself to this activity.

There, Chinese companies such as ByteDance, owner of TikTok, have leased data center capacity. By resorting, then, to leasing cloud capacity, Chinese companies are able to access chips that cannot be imported into China.

SemiAnalysis, a consulting firm, estimates that nearly half of the capacity planned for Johor data centers in 2027 will integrate artificial intelligence processors such as those from Nvidia.

A lawyer advising companies in the region said it is relatively easy for Chinese companies to obtain restricted artificial intelligence chips by opening local subsidiaries.

On the other hand, several intelligence experts have long pointed out how Chinese exporters are adopting different tactics to avoid heavy U.S. tariffs, including redirecting shipments through third countries to disguise their origin. This strategy, known as “place-of-origin washing,” involves sending goods through nations such as Malaysia, Vietnam, Thailand and South Korea before re-exporting them to the United States with newly issued certificates of origin.

Authorities in countries such as Malaysia, Vietnam, and Thailand are investigating this practice and taking steps to strengthen origin controls.

Chinese exporters usually sell goods “free on board” (FOB), which means liability is transferred to buyers once the goods leave China, making enforcement difficult.

Another reported method is to mix high-cost items with cheaper goods to lower the total declared value of shipments. Middlemen are said to offer “gray area” tariff solutions for small and medium-sized enterprises.

New rules for smugglers

For this, restrictions are expected to take effect (May 15) through which the United States would go to further restrict the smugglers' network. The new rules affect not only hardware, but also cloud providers that offer access, but there is a risk that even these additional restrictions could fail.

Nvidia's defense

The company insists that it complies with U.S. export regulations. However, its supply chain is very long. Nvidia supplies processors to Google and Microsoft and to equipment manufacturers such as Dell and Supermicro, which integrate them into servers. Therefore, compliance responsibility is widespread. Cloud service providers and hardware companies should audit their customers.

In turn, Nvidia conducts periodic audits, but oversight is inconsistent, and servers, after passing initial audits, often change owners without notice. One server manufacturer executive said that properly auditing all end users is “virtually impossible.”

A historical pattern of challenges to U.S. technology export control

The struggle to prevent the smuggling of AI chips reflects a recurring challenge in U.S. export control enforcement that spans decades and involves multiple technologies.

Similar to current concerns about AI chips, many countries have historically found ways to circumvent export controls on sensitive technologies, as evidenced by recent cases of DeepSeek claiming to have used restricted Nvidia chips to develop its own AI models.

Singapore's intensified enforcement efforts, including the indictment of individuals for customs fraud related to servers potentially containing Nvidia chips, demonstrates the global nature of this challenge and the need for international cooperation.

True, the problem of smuggling of sophisticated technologies, despite export controls, has been observed over time with various technologies, but now it is manifesting itself in the area of artificial intelligence chips, where at stake are potential military applications.

An AI chip tracking bill

A U.S. congressman is preparing to introduce a bill aimed at tracking artificial intelligence chips, including those made by Nvidia, to prevent them from being smuggled into China in violation of export controls.

Representative Bill Foster, Democrat of Illinois, announced the bill, which has bipartisan support. His bill aims to mandate tracking technology for chips after sale and ensure that unlicensed ones can no longer be used in compliance with U.S. export laws.

As noted, concerns have arisen within the Trump administration about the smuggling of Nvidia's AI chips because of their importance in the development of advanced artificial intelligence systems, including those that could be used as weapons.

Completing the U.S. semiconductor reshoring strategy

The proposed Chip Traceability Act aims to complement the broader strategy to rebuild domestic semiconductor manufacturing, which began in 2022 with the CHIPS Act, which resulted in the allocation of some $30 billion in grants for 23 projects in 14 states.

The latter legislation, in fact, aims to dramatically increase U.S. manufacturing capacity for state-of-the-art logic microprocessors by addressing supply chain vulnerabilities that export controls alone cannot solve.

From this perspective, Taiwan Semiconductor Manufacturing Co.'s $65 billion commitment to three factories in Phoenix and Intel's $8.5 billion investment in four states demonstrate how policy is already reshaping global semiconductor investment patterns.

The traceability requirement would create a new technical and compliance framework around these investments, which could affect the way semiconductor companies design their chips and manage their supply chains globally.

The legislation also reflects the evolving realization that manufacturing reshoring alone is insufficient without tighter controls on where advanced chips end up after production.

Other proposals against smuggling

Some experts have proposed technical solutions, such as deactivating chips used in prohibited locations. Nvidia, however, argues that such hardware-level controls would introduce dangerous vulnerabilities and be unworkable. Instead, it suggests that software tools could transmit limited telemetry data, including location and system configuration information, to the company to confirm that the chips are where they should be.

Fewer limits and more innovation

The United States will not achieve dominance in artificial intelligence if its companies cannot access global markets, said ITIF, a U.S. think tank. “U.S. markets alone are simply not large enough to support globally competitive high-tech companies, which must have access to appropriately sized global markets to generate the revenues needed to reinvest in the costly effort to bring next-generation products to market,” it added.

Michael Frank, a member of the 2430 Group, believes that “America should embrace interdependence as a weapon against China.” The more the United States and the West try to block Chinese technological development with limiting measures, the more China will be forced to accelerate the pace of independent innovation, the expert says.

If it is difficult for allies to buy U.S. chips, they will get them from other suppliers, including the Chinese, writes the WSJ. If it is difficult for allies to buy U.S. chips, they will get them from other suppliers, including the Chinese. Exporting hardware is not the same as exporting capabilities. Preventing China from accessing U.S. chips may have slowed model development, but it has also sharpened Beijing's deployment capabilities in another way. China now leads the world in open source artificial intelligence, and this leadership has emerged despite constraints imposed by U.S. policymakers. What America temporarily blocks in outputs (models), it permanently enables in inputs (chips). The same pattern has already occurred in telecommunications with ZTE and Huawei semiconductors: cheaper alternatives have undermined U.S. dominance. America should know better. If artificial intelligence is devouring the world, we need to make sure it devours American hardware and resources. The United States should flood the world with American GPUs in a concerted way.

Beijing invests tens of millions of dollars to promote an artificial intelligence ecosystem without Nvidia in China

Meanwhile, Beijing is allocating tens of millions of dollars in grants to spur the growth of the city's artificial intelligence supply chain, using entirely domestic technologies, as China invests heavily in its push for technological self-sufficiency.

The Yizhuang Development Zone, also known as the Beijing Economic and Technological Development Zone, said Wednesday it plans to create a nation-leading artificial intelligence industry ecosystem worth 80 billion yuan ($11 billion) by the end of this year, according to a plan shared on its official WeChat account.

The ecosystem would rely exclusively on Chinese semiconductors, operating systems and open source software frameworks to ensure self-sufficiency and control, policymakers said.

What to do?

In recent decades, China has been conducting its own Liminal Warfare against the rest of the world, eroding production chains with the goal of making states dependent on Beijing, and this plan is succeeding.

The corollary is that the more industries weaken (semiconductors, telecommunications, critical minerals and rare earth elements, high-capacity batteries, pharmaceuticals and medical equipment), the more the national security of the states to which they belong is at risk. Without access to secure supply chains, no country is able to sustain its economy and develop weapon systems for national defense.

All of this is not to exclude China from global supply chains that intersect with those of other countries such as the U.S. and EU; however, it is incumbent to understand which circumstances create unacceptable risks and which, on the other hand, create tolerable or benign risks. In particular, one should be concerned about the so-called asymmetric interdependence that Beijing uses as a geopolitical weapon.

Such are critical supply chains, that is, those that generate risky strategic dependence that arises when limited access to a category of products can disrupt a country's economy or leave it otherwise vulnerable. This must take into account the technology, know-how, cost and time required to create alternative sources for vital industrial production. Shock preparedness is only one factor and the commercial diversification, not autarky, is the key to the solution. The goal is threefold: security, openness and prosperity of the chains.

Regulations need to be adjusted and new alliances found among like-minded countries, securing their economies. It is not a matter of shutting down relations with China; however, modernization of regulatory processes becomes the main requirement to be able to defend oneself. It is necessary, in essence, to work for the regulation of the new globalization that awaits us, keeping the definition of the security perimeter of the United States at the center of political action.

Therefore, Trump's task is damnably complicated.

Finding a synthesis between geoeconomic interests, national security and economic freedom is not easy. The risk of sinning by excess is real. However, even businesses need to understand that the Trump Administration faces a dangerous “enemy” that uses sophisticated techniques and tactics and has been waging its own liminal war for years, that is, an incremental war, in which the spectrum of competition and confrontation with the West is so broad that the battlefield is everywhere and the war is total.

And it is within this logic, control of technological assets, strategic infrastructure and major global supply chains, integrate “trans-military” and “non-military” warfare operations by Beijing. To make states more dependent on Beijing and more willing to “a new authoritarian world order with distinctive Chinese characteristics,” Liminal Warfare encompasses economic, diplomatic, legal, military, intelligence, and cyber operations put in place seamlessly.