Technology Restrictions Have Become a Central Instrument of Economic Statecraft

The United States is intensifying its technology confrontation with China, but the situation has become harder to read. What was once a clear bipartisan consensus around restricting China's access to advanced semiconductors has fractured, replaced by a triangular standoff between Congress, the White House, and the chip industry that is reshaping US strategy in real time.

For years, the core logic was containment: progressively tightening export controls on high-end chips and the equipment used to fabricate them, while simultaneously investing in domestic capacity through the CHIPS and Science Act of 2022. That law channeled tens of billions of dollars into manufacturing incentives and research initiatives, from TSMC's Arizona fab to Samsung's Texas expansion, Intel's Ohio investment, and new capacity from GlobalFoundries and Micron. The ambition was not merely defensive. It reflected a recognition that technological leadership, once ceded, is extraordinarily difficult to reclaim, and that the architecture of global chip production had become a strategic vulnerability as much as an economic asset.

That consensus is now under stress in ways that deserve serious scrutiny.

The transactional turn

In December 2025, the Trump administration reversed the Biden-era presumption of denial for Nvidia's H200 chip to China. The Commerce Department's Bureau of Industry and Security formally shifted its licensing posture to case-by-case review in January 2026, subject to strict conditions: third-party testing in the US before export, a volume cap limiting China-bound shipments to 50 percent of domestic US sales, and a 25 percent tariff attached to each shipment with revenue flowing to the Treasury. President Donald Trump framed the arrangement explicitly in transactional terms: "We're allowing them to do it, but the United States is getting 25% of the chips, in terms of the dollar value."

The commercial stakes are significant. Chinese technology companies reportedly placed orders for more than 2 million H200 chips for 2026, and Nvidia CEO Jensen Huang has previously estimated the China market could be worth $50 billion annually. The policy shift represents a clear break from the Biden administration's approach and has alarmed national security officials on both sides of the aisle. Former Deputy National Security Adviser Matt Pottinger warned Congress that H200 sales to China would "supercharge Beijing's military modernization," enhancing capabilities across nuclear weapons design, cyber warfare, autonomous drones, and AI surveillance.

Beijing has created its own complications. China's government has reportedly indicated it will approve H200 purchases only under exceptional circumstances, and concerns about Nvidia's hardware tracking capabilities have led to friction, illustrating that even when Washington opens a door, Beijing may choose not to walk through it for reasons of domestic industrial policy.

Congress pushes back

The legislative response has been swift, bipartisan, and in open conflict with the White House. The central vehicle is the AI Overwatch Act, introduced by Rep. Brian Mast (R-Fla.) and advanced by the House Foreign Affairs Committee in January 2026. The bill would give both chambers 30 days to review and potentially block any export license for advanced AI chips to foreign adversaries, modeled on existing congressional oversight of arms sales. It would also impose a mandatory denial requirement for chips more powerful than the H200, including Nvidia's Blackwell architecture, and would close what critics call the threshold-gaming loophole: any chip first marketed after January 1, 2026 that approaches the performance thresholds for covered chips would face automatic denial, preventing the kind of incremental workarounds Nvidia used with its China-targeted H20.

The bill has put Mast in direct conflict with Nvidia and with the White House. David Sacks, the administration's AI and crypto czar, pushed back publicly, while Mast accused Nvidia of "fighting to sell millions of advanced AI chips to Chinese military companies like Alibaba and Tencent." The GAIN AI Act, which would have required chipmakers to give US customers first access before any export, passed the Senate as part of the National Defense Authorization Act but was stripped from the final bill in the face of White House resistance. A separate measure, the SAFE Chips Act, seeks to lock in existing chip export restrictions for 30 months.

Congress has also moved against a different kind of leakage. The Remote Access Security Act, which passed the House in January 2026 by 369 to 22, closes what lawmakers call the "cloud loophole": foreign companies can effectively import controlled chip capacity by renting time on advanced GPUs hosted in data centers in third countries, without a single chip ever crossing a border. One widely cited case involved a Shanghai-based startup accessing Nvidia Blackwell systems through an arrangement routed through Indonesia. The bill would bring remote access to controlled hardware under the same export control framework as physical shipments.

The equipment frontier

The most recent development, introduced in the House this week with a Senate companion expected before the end of April, extends the battleground from finished chips to the machines used to make them. The Multilateral Alignment of Technology Controls on Hardware (MATCH) Act would transition from the current model of entity-specific restrictions to a countrywide ban on exports of advanced semiconductor manufacturing equipment. The immediate targets are ASML's deep ultraviolet (DUV) lithography systems and Tokyo Electron's etching and deposition tools, equipment that remains legally exportable to most Chinese fabs even as the most advanced EUV systems have been blocked since 2019.

The bill addresses a critical compliance gap. Dutch sales of advanced lithography equipment to China doubled from 2022 to 2023, and again from 2023 to 2024. Once equipment enters China, enforcement becomes difficult: verification visits require approval from Chinese authorities and can take weeks. Allied companies, including ASML and Tokyo Electron, continue to service installed machines inside Chinese fabs, enabling repairs and incremental upgrades that entity-level controls cannot reach. The MATCH Act would also invoke the foreign direct product rule to compel allies to align their own restrictions, reflecting concern that US leverage over European and Japanese equipment suppliers is eroding.

This matters because semiconductor manufacturing equipment is perhaps the most genuine chokepoint in the entire technology competition. China's leading foundry, SMIC, has pushed to 7-nanometer production using multiple DUV exposure techniques in the absence of EUV, and has reportedly produced 5-nanometer chips using DUV workarounds, though at low yields and high cost. TSMC remains far ahead. But Beijing is funding parallel workarounds across nanoimprint lithography, advanced multi-patterning, and long-shot EUV prototypes. The assumption that export restrictions can freeze a capability gap in place has repeatedly proved optimistic.

The strategic incoherence problem

What emerges from this legislative landscape is not a strategy but a set of competing strategies operating simultaneously. The administration is loosening controls on finished chips while Congress attempts to tighten them. The executive branch withdrew a draft rule in February 2026 that would have further restricted global AI chip sales, signaling discomfort with the kind of multilateral controls the Biden team had been negotiating with Japan and the Netherlands, while Congress pushes legislation designed to force exactly that multilateral alignment. The bipartisan letter to Commerce and State Departments in February 2026 was direct: "Entity-specific controls, while valuable, cannot substitute for countrywide restrictions on the most critical chokepoint tools."

The pattern extends beyond any single decision. The administration formally rescinded the Biden-era AI Diffusion Rule in May 2025, promising a simpler replacement, yet no replacement materialized. A draft rule circulated internally in February 2026 was pulled from regulatory review within weeks, with a former official noting the withdrawal likely reflected unresolved divisions within the administration itself over how to achieve AI dominance. The result is a policy vacuum: the Biden framework is gone, the promised successor does not exist, and the ad hoc arrangements filling the gap, including the H200 revenue-sharing deal, carry no statutory basis.

The incoherence carries real costs. Industry faces a state of permanent uncertainty: export policy under the Trump administration has oscillated between restriction and relaxation on the H20 alone multiple times. Chinese technology companies have responded by accelerating investment in domestic alternatives, offshoring compute to data centers in Singapore and Malaysia, and negotiating directly with Beijing over whether to purchase US hardware at all. Each policy reversal reduces the credibility of the US as a reliable supplier and accelerates the self-sufficiency dynamic that export controls were designed to prevent.

China's response to the pressure has been consistent and deliberate. Since 2017, the New Generation Artificial Intelligence Development Plan has guided a whole-of-state effort to make China a global AI innovation center by 2030. The newly approved 15th Five-Year Plan reinforces that trajectory by elevating AI to a core pillar of national competitiveness, calling for breakthroughs in foundational research, accelerated self-reliance in strategic technologies, and the expansion of the "AI+" initiative across manufacturing, public services, and governance. Firms like Huawei and Cambricon are engineering purpose-built accelerators for training and inference workloads that do not need to match Nvidia's general-purpose performance to be strategically decisive in specific domains, including defense, surveillance, and industrial automation. China's advantage in applied AI is further amplified by deployment scale: vast datasets and rapid rollout across commerce, finance, healthcare, logistics, and energy create feedback loops that aggregate compute metrics alone cannot capture.

The talent dimension deserves equal attention. China graduates more STEM PhDs annually than the US, and while emigration historically benefited American institutions, tightened visa policies and improving domestic research conditions are beginning to reverse that flow. The competition for the world's most skilled AI researchers is, in many respects, as consequential as the competition for chips.

What a coherent strategy would require

A purely defensive posture built around export controls and tariffs is necessary but insufficient. Containment slows diffusion; it does not reverse capability gaps that already exist, and it provides no substitute for sustained domestic investment. The current moment calls for something harder: a clear articulation of what the US is actually trying to achieve. Preventing China from accessing the most advanced chips indefinitely is probably not achievable over a long time horizon, as the history of technology diffusion argues against it. The more tractable goal is maintaining a meaningful lead long enough for American firms, institutions, and alliances to consolidate structural advantages that are less dependent on any single technology chokepoint.

That requires combining selective, well-coordinated export controls, including genuine multilateral alignment with Japan and the Netherlands on manufacturing equipment, with a broader program of domestic renewal: consistent long-horizon federal R&D funding, private-sector investment in foundational technologies, and the institutional conditions that make the US an attractive destination for global talent. It also requires that Congress and the executive branch develop a shared framework for technology competition rather than continuing to fight each other through competing legislation and administrative reversals.

The US retains significant structural advantages: deep capital markets, world-class research universities, a dynamic private sector, and a broad alliance network. But these advantages are not self-sustaining. They are the product of choices made over decades, and they are eroding under the pressure of geopolitical symbolism substituting for coherent industrial policy. Winning this competition will require strategies that go beyond tariffs and export controls, and that begin, as a minimum, with a government capable of maintaining a consistent position.