President Donald Trump has ordered an investigation into national security risks posed by the import of semiconductors, the equipment used to build them, and the products that use them.

A key goal of the probe is to assess the potential for building out a domestic semiconductor manufacturing capability to meet U.S. demand for advanced computer chips and understand the role of foreign manufacturing capabilities in supporting that goal.

The United States is a significant producer of semiconductors and currently manufactures about 12 percent of the world’s chips. But it relies heavily on imports from Taiwan and South Korea for advanced chips, such as those used to create artificial intelligence (AI).

What’s more, the United States has steadily lost its share of global semiconductor manufacturing over the years, as American companies have turned more to design and research while offshoring chip manufacture.

That has placed the United States in a precarious position, in which it has become overwhelmingly reliant on just two international partners for its supply of the advanced semiconductors required to fuel innovative technologies, from consumer goods to defense articles.

The investigation will also inform the administration on how best to curb exports of American-designed chips to communist China, whose market share of semiconductor manufacturing and related products is rapidly rising.

Premium Picks

Trump Admin Launches National Security Probe Into Pharma, Semiconductor Imports

US Tariffs May Drive China’s Economy Into a Depression, Experts Say

China Ramped Up Fentanyl Precursor Production to Target US, Expert Says

Semiconductors are the foundation of virtually all modern electronic devices, from pickup trucks to cell phones, refrigerators, and hypersonic missiles.

Often made out of silicon and germanium, they are crucial for building electronic components because of their unique ability to conduct electricity under certain conditions while not conducting electricity under other conditions.

It is this capability that allows logic operations to occur in computers. At the most foundational level, the zeros and ones of binary code refer to varying degrees of voltage in a digital circuit. Semiconductors’ ability to conduct or not conduct electricity makes them an excellent material for transistors, providing switches that allow computer chips to select between the two voltage options, thereby forming the underlying logic of all non-quantum computers.

This means that many major technological advancements depend in turn on advances in semiconductor manufacturing nodes, unique processes that distinguish one generation of semiconductors from another.

The smaller a semiconductor is, the more transistors can be built, and therefore the more processes the final computer chip will be able to run.

Broadly, chips built on the older and larger processes of 28 nanometers (nm) or more are considered mature technologies, whereas those at 14 nm or less are considered advanced. At present, 5 nm and 3 nm are the most advanced nodes used in mass production.

Innovative new technologies such as AI require semiconductors that are smaller, lighter, and capable of processing more information more quickly. That’s because they involve incredibly intense computing that requires massive calculating power and the ability to simultaneously access multiple datasets.

Minute differences in semiconductor architecture can also have profound effects on the user end of a technology.

These effects can be reflected in minor issues of usability, such as how quickly a chatbot can respond to a user’s inquiry, or in major issues of safety and reliability, such as how quickly a self-driving car can decide whether an apparent object in its view is a mere shadow or a human being.

This processing power also has immense implications for research that could have national security implications, such as crunching and extrapolating data to develop new medicines or weapons.

In 1990, some 37 percent of the world’s semiconductor manufacturing took place in the United States, according to a report by J.P. Morgan. In 2025, the United States is projected to produce just around 11 percent.

That decrease correlated directly with rapid increases in chip manufacturing in China, South Korea, and Taiwan, which took on more manufacturing work as American tech companies focused more on research and development.

Now, the vast majority of chips produced in both China and the United States are mature or otherwise non-advanced models, while advanced chips are overwhelmingly produced in Taiwan.

According to J.P. Morgan, Taiwan Semiconductor Manufacturing Company, commonly known as TSMC, manufactures more than 90 percent of the world’s advanced AI logic chips.

For comparison, only two fabs in the United States currently manufacture chips on the 5 nm node. One is owned by Samsung and the other by Taiwan’s TSMC. Three more U.S. facilities are in production that will be capable of producing state-of-the-art chips, all owned by TSMC. Other facilities may be in progress and develop similar chips, but they have not yet announced the exact nodes they will produce.

Those semiconductor manufacturers owned by U.S. entities also trail their international competition in terms of revenue volume.

In the last quarter of fiscal year 2024, TSMC’s total revenue hit $26.8 billion. The total revenue of GlobalFoundries, the United States’ largest chip maker, was $1.8 billion.

The semiconductor supply chain is more complex than the manufacturing output or revenue metrics would suggest, however. That’s because of the specialization semiconductor companies have taken on in the past several decades. During that time, many companies stopped simultaneously designing and building semiconductors, instead choosing to specialize in one side or the other.

Advanced Micro Devices, commonly known as AMD, Nvidia, and Qualcomm are what are now known as “fabless” manufacturers: companies that research, design, and sell semiconductors but do not fabricate them.

TSMC and GlobalFoundries are known as “pure play” foundries, that manufacture chips for others but produce nothing under their own brand.

Much of the modern semiconductor landscape is built this way, and the overall arrangement is unlikely to change, even as U.S. export controls seek to push manufacturing back to the United States.

U.S. leaders have long worried that giving China access to top-tier semiconductors could boost the communist country’s global power because advanced chips used in AI can also be used for military applications, including autonomous weapons, cyberwarfare systems, and surveillance infrastructure.

In essence, the United States is trying to preserve its waning edge in AI research, quantum computing, space exploration, and defense by limiting China’s access to the technologies needed to advance in these areas.

This process began in 2018 during Trump’s first term, when he signed the National Defense Authorization Act for the next fiscal year. The act prohibited the government’s use and procurement of equipment from two Chinese companies, Huawei and ZTE, because of security concerns.

In June 2020, the United States moved to designate Huawei and ZTE as threats to national security because of their close ties to the Chinese Communist Party (CCP) and its military, and expanded export controls to China’s leading semiconductor manufacturer, Semiconductor Manufacturing International Corporation, commonly known as SMIC.

Under those rules, doing business with the designated companies required a special license from the U.S. government, and effectively banned many advanced semiconductors from being sold directly to China’s largest chipmaker.

In 2022, the Biden administration added more advanced and high-performance chips and related commodities that contain such chips to the list of restricted items under the claim that such technologies could be used for both civilian and military purposes.

U.S. companies such as Nvidia and AMD were at that time blocked from selling high-end AI chips to China, and forced to create watered-down designs for the Chinese market.

In 2023, the Biden administration then moved to close what it saw as loopholes in the previous list by further restricting some of the watered-down chips that Nvidia had created for China and adding tighter controls on semiconductor manufacturing equipment, limiting China’s ability to build its own fabs.

Then, in 2024, the United States began putting diplomatic pressure on its allies, incentivizing leaders in the Netherlands, Japan, South Korea, and Taiwan to limit their own exports of chipmaking tools and technologies to China.

The goal was simple enough: choke off China’s access to the global supply chain for advanced semiconductors, as well as the tools, software, and expertise required to create them.

Now, Trump’s new investigation into the semiconductor supply chain could yield further restrictions on international access to semiconductors, and his administration has expressed a desire to replace Biden-era restrictions with new, streamlined rules for chips.

In one headline-catching incident in late April, smugglers went so far as to hide dozens of Nvidia processors among a shipment of live lobsters, but the chips were intercepted by Hong Kong authorities at the Hong Kong-Zhuhai-Macao Bridge Hong Kong Port.

Intermediaries throughout Southeast Asia have also facilitated the diversion of technology to prohibited entities, making it easier for China to acquire advanced semiconductor technology despite restrictions.

The CCP is also pouring billions of dollars into its own semiconductor ecosystem, including a push for SMIC to rapidly build out advanced fabs to create modern chips domestically.

To that end, the regime has launched the China Integrated Circuit Industry Investment Fund to boost local chip design and manufacturing. First spun up in 2014, the fund underwent a phase shift in 2024 and is now investing tens of billions of dollars in China’s burgeoning AI chip industry.

Chinese companies are also developing novel workarounds to sidestep the relatively limited access they now have to the most advanced chips. Many are now developing AI models that work on massive numbers of lower-end chips by making their models smaller, faster, and less dependent on super-high-end hardware.

A key example of this is DeepSeek, a Chinese AI company that threw the semiconductor industry into disarray toward the end of 2024 when it released its own ChatGPT competitor, which the company claimed was built in only two years and without access to the advanced chips that American companies said were required to produce such technology.

Another example is SMIC’s recent creation of an advanced chip using older manufacturing methods that were previously thought incapable of creating the highest-tier semiconductors.

Semiconductor industry consensus previously held that extreme ultraviolet lithography technology was required to produce state-of-the-art chips with tiny 5 nm or even 3 nm nodes. But SMIC has now domestically produced its own 5 nm chip without access to the heavily regulated technology that Washington and its allies believed it would need.

The new Chinese method is slower and more error-prone, but it works and provides another example of how quickly the communist nation is adapting to the many restrictions placed on it.

Whether the United States can hold back Chinese development of advanced semiconductors for much longer is an open question, as is China’s ability to quickly seize a greater share of the international semiconductor market.