A rebuilt chip industry still requires strong protections

OPINION:

The United States is reducing one of its greatest technological vulnerabilities: dependence on Taiwan for the world’s most advanced semiconductors.

Yet building more chips in America will accomplish only so much if Washington and its allies cannot prevent the most strategically important technology from reaching China’s military and intelligence ecosystem.

TSMC, the Taiwan Semiconductor Manufacturing Co. and the world’s dominant contract chipmaker, illustrates both sides of the problem. Its first Arizona fab began high-volume N4 production in late 2024, and TSMC plans to introduce still more advanced technologies there.

A Jan. 16 U.S.-Taiwan agreement added commitments by Taiwanese semiconductor and technology companies for at least $250 billion in new U.S. direct investment, backed by another $250 billion in credit guarantees.

Yet diversification is not independence: TrendForce put TSMC at 70.4% of global foundry revenue in the fourth quarter of 2025.

An equally important vulnerability is controlling where advanced chips ultimately go. Huawei has faced stringent U.S. export controls since 2019, yet a TSMC-manufactured chip ordered by Chinese chip designer Sophgo was discovered in Huawei’s Ascend 910B artificial intelligence processor.

Reuters reported that TSMC could face a penalty exceeding $1 billion in a U.S. investigation, while Lennart Heim, a researcher at Rand’s Technology and Security Policy Center, estimated that nearly 3 million chips matched the Sophgo design. TSMC says it complies with export controls and has not supplied Huawei directly since mid-September 2020.

The episode exposes a basic weakness: Semiconductor supply chains are global, whereas controls still depend heavily on individual governments. Washington needs a stronger multilateral system.

Fortunately, history offers a model.

From 1949 to 1994, the U.S. and its principal allies operated COCOM, the Coordinating Committee for Multilateral Export Controls, restricting the transfer of strategically important Western technology to the Soviet bloc. COCOM was far from airtight. Allies were not always aligned, companies wanted Eastern bloc markets, and Soviet intelligence services ran extensive technology acquisition programs.

Yet judging export controls by whether they prevent every illicit transfer misses the point. I saw evidence of this while working in the Commerce Department’s export enforcement arm near the end of the Cold War.

Around 1990, American export control officials visited Poland, Hungary and Czechoslovakia. Officials there told them that Western controls had worked — not by making advanced technology impossible to obtain but rather by making acquisitions slower, more expensive and less reliable.

Clandestinely acquired equipment also created continuing problems in obtaining spare parts, upgrades and service.

That is the lesson for trade with China. A successful export control regime need not be an embargo. Its strategic purpose is to raise the cost, time and uncertainty of acquiring the most advanced technology while permitting less sensitive trade.

In other words, Sophgo does not prove that controls are futile. It proves that enforcement and international coordination must improve.

Re-creating the Cold War COCOM wholesale would be neither possible nor desirable. Unlike in decades past, China is now deeply integrated in global manufacturing.

Instead, Washington should pursue a narrower, semiconductor-focused multilateral committee covering technologies most consequential to military power, intelligence and advanced AI.

The membership is obvious. The U.S. dominates important areas of chip design and manufacturing equipment. Taiwan has TSMC. ASML in the Netherlands supplies critical extreme-ultraviolet lithography systems. Japan and the U.S. remain indispensable in other equipment and materials, while South Korea leads in memory and semiconductor manufacturing.

No country controls the ecosystem alone. Together, they command many of its choke points.

A new arrangement should coordinate restrictions on a relatively small category: frontier AI accelerators, the most advanced manufacturing equipment, leading-edge fabrication, advanced packaging, high-bandwidth memory and associated know-how.

It should also institutionalize intelligence sharing and joint enforcement against shell companies, suspicious intermediaries and diversion networks.

One important qualification: Reducing American dependence on Taiwan must not mean hollowing out the industry central to Taiwan’s economy. The goal should be redundancy: Taiwan should remain a major advanced-semiconductor center while sufficient leading-edge capacity is built in the U.S. and other allied countries. This would prevent a blockade or attack from stopping much of the global technology economy.

Manufacturing capacity is only half of semiconductor security. The other half is preventing the most strategically significant technology from strengthening potential adversaries. Cold War experience shows that controls need not be impermeable to work. They must make illegal acquisition slower, more expensive and less reliable — and they work best when countries controlling essential technologies act together.

The need is immediate. On Aug. 10, Rep. John Moolenaar, Michigan Republican and chairman of the House Select Committee on China, called on the administration to enforce late-Biden-era rules intended to prevent another Sophgo-like diversion. Taiwan, meanwhile, added Huawei and SMIC to its export-control list in 2025.

These steps show that the foundations of coordinated action already exist. What is missing is a durable institution that aligns rules, intelligence and enforcement.

Such coordination would also reduce a recurring political problem: Unilateral American controls can push sales toward foreign competitors or generate resentment among allies asked to adopt restrictions after Washington has acted.

A standing mechanism would force governments to negotiate the narrow control list together, update it as technology changes, and share responsibility for enforcement. That is closer to COCOM’s real strength than simply copying its Cold War bureaucracy.

America and its allies are building a more resilient semiconductor manufacturing system. Now they need an export-control system to match it.

• Matt Brazil, Ph.D., served in U.S. Army intelligence, the Commerce Department’s Office of Export Enforcement and the U.S. Embassy in Beijing before entering the private sector to conduct security investigations for American companies in Asia.