ANALYSIS: Washington Further Tightens Restrictions on Chip Exports to China

The Guarantee That Production Capacity for U.S. Customers Will Not Be Reduced

U.S. exporters will have to demonstrate that the sale of these chips to China does not reduce the global semiconductor production capacity available to U.S. customers—a requirement explicitly designed to prioritize the protection of the domestic supply chain over any commercial considerations related to export markets. “America First” is not just a slogan here; it is a condition set forth in black and white in the regulatory text.

This requirement forces U.S. manufacturers, such as Nvidia and AMD, to carefully manage their production chains so they can prove—potentially upon request—that their sales to China do not deprive their domestic customers of supplies.

Mandatory Screening of Chinese Buyers

Chinese buyers of these chips will have to demonstrate the existence of export compliance procedures, including mechanisms for screening their own customers—a requirement that shifts part of the compliance responsibility to the Chinese entity itself rather than placing it solely on the U.S. exporter.

Placing part of the compliance burden on the Chinese buyer is a way for Washington to share the risk without relinquishing ultimate control.

Verification by an independent U.S. third party

Exported products must undergo independent testing by a U.S. third party—an additional technical requirement designed to verify that the actual characteristics of the exported chips match those declared, and that no unauthorized modifications were made prior to shipment to China. An independent test does not prove the buyer’s intent, but it does at least confirm the exact nature of the product sold.

This independent verification mechanism adds a layer of logistical complexity and additional delays to the export process, which could, in practice, slow down the pace of shipments even for applications that are ultimately approved.

What This Test Cannot Guarantee

No independent technical test, however rigorous, can guarantee the actual end use of these chips once they are delivered to China—a blind spot implicitly acknowledged by the proliferation of other administrative safeguards accompanying this revised rule.

Verifying a product at the time of shipment says nothing about what will become of it once it is in someone else’s hands, thousands of kilometers away.

"Strengthening the U.S. Technology Ecosystem"

Under Secretary of Commerce Jeffrey Kessler stated: “Export controls must evolve with technological changes, while protecting national security. Allowing the sale of the H200 to China under controlled conditions will strengthen the U.S. technology ecosystem.” Framing a sale to a strategic rival as a way to strengthen one’s own ecosystem requires a justification that deserves close scrutiny.

This wording suggests an underlying economic rationale: to maintain U.S. companies like Nvidia in a dominant position in the global chip market—including through regulated sales to China—rather than allowing that market to be entirely captured by non-U.S. competitors not subject to the same restrictions.

A Tension Between Commercial and Security Considerations

This statement illustrates a central tension in current U.S. policy toward China regarding advanced technologies: on the one hand, the desire to preserve a commercial and industrial advantage for U.S. companies; on the other, the fear that these same sales may, in the long run, strengthen China’s technological and military capabilities.

Sell to remain dominant, or refuse to sell to remain secure: this administration has, for now, chosen the first option—with conditions.

Trump’s Announcement on December 8, 2025

This July 2026 revision is a direct continuation of the announcement made by President Trump on December 8, 2025, which paved the way for the export of the Nvidia H200 to approved Chinese customers, explicitly justifying it as a means of strengthening—rather than weakening—U.S. national security. A policy that has lasted seven months is no longer a one-off move; it has become an established doctrine.

This continuity between the December 2025 announcement and the July 2026 regulatory revision indicates a stable stance by the U.S. administration on this issue, rather than an impromptu reversal under pressure from recent events.

What Changed Between December and July

Between the December 2025 policy announcement and the concrete regulatory revision in July 2026, the administration had time to refine the specific implementation details, including requirements for independent testing and buyer screening—technical details that were not necessarily specified in the initial, more general announcement.

It took just seven months to transform a policy announcement into a precise and binding regulatory text.

A "small number" of H200 chips have already been shipped, according to Bloomberg

On July 14, 2026, Bloomberg reported that a “small number” of Nvidia H200 artificial intelligence chips had already been shipped to China with U.S. license approval, though the quantity and the identities of the buyers remained undisclosed to the public. A “small number” that has not been publicly quantified leaves more questions unanswered than it answers.

Undersecretary Kessler had described these shipments as “trivial” in volume—a characterization that, in the absence of specific figures made public, cannot be independently verified by this analysis.

What This Prior Activity Reveals About the Policy’s Actual Timeline

The fact that shipments had already taken place even before the publication of the July 27 regulatory revision suggests that the practical implementation of the regulated openness policy had already begun—on an individual license basis—even before the regulatory framework that governs it today was fully formalized.

Practice has sometimes preceded the rule. This is not necessarily irregular, but it is worth noting.

EUV Machines Still Out of China’s Reach

Unlike the easing of restrictions seen on certain chips such as the H200, the Dutch company ASML—the global leader in lithography machines used to manufacture the most advanced semiconductors—continues to refrain from shipping its most sophisticated extreme ultraviolet (EUV) lithography equipment to China, following several years of export restrictions. Approving a chip does not mean approving the machine that could, one day, manufacture it locally.

This distinction between finished chips and manufacturing equipment illustrates an implicit hierarchy in Western restrictions: limiting China’s ability to produce advanced chips on its own is, in the long term, more important than limiting China’s access to chips already manufactured elsewhere.

China’s Economic Significance for ASML

According to ASML’s Chief Financial Officer, Roger Dassen, as quoted by CNBC on July 17, 2026, sales to China account for approximately 20% of the company’s net revenue for 2026—a significant commercial dependency that places ASML in a delicate position between its Western regulatory obligations and its direct commercial interests in China.

Twenty percent of a global leader’s revenue is not a figure one sacrifices easily, even for reasons of allied national security.

The Risk of Indirect Benefits to China’s Military Capabilities

National security experts cited in this report fear that easing export restrictions on the H200—even with the safeguards described above in place—could ultimately benefit China’s military and artificial intelligence capabilities indirectly, despite the screening and testing mechanisms intended to prevent this risk. A regulatory safeguard does not always prevent the end use that its creators initially feared.

This concern is based in part on the acknowledged difficulty of tracking with certainty the end use of electronic components once they have left U.S. territory and been integrated into broader Chinese systems, whether civilian or military.

The Ongoing Debate Over the Actual Effectiveness of the Safeguards

To date, no independent and public assessment has confirmed or refuted the actual effectiveness of the screening and testing mechanisms put in place by this regulatory revision, leaving the debate between supporters and critics of this policy largely open and unresolved by verifiable data.

A safeguard that has not been tested under real-world conditions remains, by definition, a hypothesis rather than a certainty.

No specific official Chinese reaction to this revision has been identified

The sources consulted for this analysis contain no specific official Chinese reaction to this particular regulatory revision of July 27, 2026, although China has, in the past, regularly criticized U.S. restrictions on semiconductor exports as a whole as a form of unjustified technological containment. China’s silence on this specific text does not imply indifference; it may also indicate a calculated approach.

This specific silence could reflect a Chinese preference not to publicly comment on a measure that, despite its restrictions, nonetheless provides partial access to advanced chips that were previously more difficult to obtain legally.

China’s Parallel Strategy of Self-Sufficiency

Regardless of this partial U.S. opening, China continues to pursue its own strategy of technological self-sufficiency in semiconductors in parallel, as evidenced by the spectacular growth of more than 2,500% in profits reported by Chinese integrated circuit manufacturers for the first half of 2026.

Accepting a partial U.S. opening does not prevent Beijing from continuing, in parallel, to build its own long-term technological independence.

Access to the Chinese Market Under Strict Conditions

For Nvidia and AMD, this regulatory revision represents potential—albeit highly conditional—access to one of the world’s largest semiconductor markets for artificial intelligence; however, this access remains subject to a case-by-case licensing process for each transaction, with no guarantee of automatic approval for future applications. A market reopened under strict conditions remains a smaller market than it was before the restrictions.

This persistent uncertainty regarding case-by-case approval complicates long-term business planning for these companies, which cannot guarantee their shareholders stable and predictable access to the Chinese market.

Competition from Chinese Domestic Chip Manufacturers

While Washington debates the terms of access to the Chinese market for Nvidia and AMD, Chinese domestic chip manufacturers continue to develop their own alternatives—a trend accelerated precisely by years of prior U.S. restrictions that prompted China to invest heavily in its own production capacity.

Every year of restrictions gave China one more year to build its own domestic alternative.

The actual volume of exports approved under this new framework

The key issue to watch in the coming months will be the actual volume of H200 and MI325X chip exports that are actually approved under this revised regulatory framework—a figure that will help determine whether this policy represents a substantial opening or remains, in practice, largely symbolic and restrictive. The number of licenses actually approved in six months’ time will speak louder than any statement of intent made today.

If this volume data becomes public, it will help settle the current debate between those who view this revision as a significant opening for trade and those who see it as nothing more than a cosmetic adjustment to a fundamentally restrictive policy.

The U.S. Congress’s Response to This Executive Policy

The U.S. Congress, where lawmakers from both parties have in the past expressed concerns about the transfer of sensitive technologies to China, may choose to closely scrutinize the implementation of this regulatory revision, with the possibility of hearings or legislative proposals aimed at further tightening—or, conversely, codifying—this policy of controlled openness.

An executive policy remains, by its very nature, easier to modify than a law passed by Congress. This revision may not be the final word on the matter.

Allied coordination that has yet to be publicly clarified

This U.S. regulatory revision comes at a time when technological allies such as the Netherlands, through ASML, and Japan are maintaining their own export restrictions on certain semiconductor manufacturing equipment to China, with no source consulted confirming explicit and detailed coordination between Washington and these allies regarding the specific relaxation of the H200 restrictions. A U.S. policy that moves forward on its own, without allies acting in parallel, creates its own areas of friction.

This lack of documented explicit coordination could, over time, create tensions among Western allies if their respective approaches to the export of sensitive technologies to China continue to diverge noticeably.

Allied companies caught between two national priorities

Companies like ASML, for which nearly one-fifth of revenue depends on the Chinese market, find themselves caught between the regulatory requirements of their own national government and the sometimes contradictory signals sent by the United States regarding the level of openness deemed acceptable toward China in this strategic sector.

Navigating between two governments with differing approaches has become, for these companies, an almost daily challenge.

A gradual escalation over the past several years

U.S. restrictions on semiconductor exports to China have gradually tightened over the past several years, before undergoing—during the period from December 2025 to July 2026—a partial and controlled easing for certain specific chips such as the H200, which represents a notable shift from the previous trajectory of continuous tightening.

A trajectory that tightened for years, then partially eased over the course of a few months, warrants examination as a shift in policy, not merely a technical adjustment.

What This Shift Reveals About the Internal U.S. Debate

This shift suggests the existence of an active internal debate within the U.S. administration between proponents of a hardline security stance—favoring maximum restrictions—and proponents of a more trade-oriented approach, concerned with preserving the dominant position of U.S. companies in the global chip market in the face of international competition.

This internal debate, rarely made public in detail, can nonetheless be read between the lines through the regulatory hesitations observed between January and July 2026. A debate that is never fully revealed always ends up giving itself away in the timing of decisions.

A republished regulation, additional safeguards, and chips that had already left the country even before the framework was fully formalized: this is what is documented in the July 27, 2026, revision regarding the export of U.S. semiconductors to China. What is clear are the specific terms of this revision and the quote attributed to Jeffrey Kessler. What remains uncertain is the actual effectiveness of these safeguards in the face of an end use that cannot be traced with certainty once the border has been crossed. Export controls have never prevented technology from traveling; at best, they have merely slowed its rate of spread. The question is not whether these chips will have a strategic use in China, but when that use will become apparent. The next export figures, once published, will reveal whether Washington has opened a narrow door or a wider one than announced.

By Maxime Marquette, columnist