Winning the Chip War, Losing the Supply Chain?

In September 2022, National Security Advisor Jake Sullivan announced a fundamental shift in U.S. semiconductor policy. Washington would abandon the "sliding scale" approach that aimed to stay "only a couple of generations ahead" of China in advanced technologies. Instead, the United States would seek to "maintain as large of a lead as possible" in chips that power artificial intelligence and defense systems. This became known as the Sullivan Tech Doctrine: export controls engineered not merely to constrain Chinese progress, but to degrade Beijing's capabilities over time while widening America's technological advantage.

Two years later, that doctrine lies in ruins and many are not sure what has replaced it. President Trump in December 2025 announced that Nvidia can sell H200 chips to China in exchange for 25 percent of revenues. This was a major policy reversal. It signaled a fundamental reorientation from Sullivan's strategic denial policy to transactional commerce. Trump would now be treating advanced semiconductors as negotiable commodities rather than the crown jewels of U.S. national security. The Trump administration rescinded Biden's AI Diffusion Rule before it took effect in May 2025, calling it "overly complex" and "bureaucratic." Where Biden erected a three-tiered framework restricting chip exports globally—sorting countries by alliance status and capping access accordingly—Trump instead is prioritizing bilateral deals to those offering the best terms and the biggest checks for the U.S. economy.

The whiplash is striking. In July 2025, the administration approved sales of Nvidia's H20 chips to China in exchange for 15 percent of revenues. By August, Trump demanded a 10 percent equity stake in Intel as a condition for CHIPS Act funding. In December, Trump made the H200 model—chips roughly six times more powerful than the H20—available to "approved customers" in China for a 25 percent surcharge. Trump justified this move as beneficial for the U.S. economy and leveraged export controls to create new revenue streams for the Treasury.

Then just weeks later, in January 2026, Trump blocked a $2.9 million Chinese acquisition of semiconductor assets from Emcore, citing unspecified national security concerns. The juxtaposition was difficult to square. He approved billions in H200 sales, then prohibited millions in asset purchases? The likely difference is that the former deal generated revenue and maintains market share, while the other involves direct ownership of manufacturing capability. The distinction is critical, but the inconsistency highlights deeper confusion about what semiconductor policy actually aims to achieve.

One leg of the policy debate has crystallized around competing assessments of the U.S.-China technology timeline gap. David Sacks, Trump's AI and crypto czar, warned in June 2025 that China is "at most two years behind" the United States in semiconductor design, with Huawei "moving fast to catch up" and Chinese firms "adept at evading US export controls." Sacks pointed to DeepSeek's breakthrough AI model as evidence that Chinese capabilities are "only months behind" rather than years, making export restrictions futile and counterproductive. Nvidia CEO Jensen Huang called export controls "a failure," warning that restrictions "gave [Chinese companies] the spirit, the energy and the government support to accelerate their development."

Defenders of export controls argue the policies are working. While China's SMIC has manufactured 7nm chips through costly multi-patterning techniques, these chips are produced at low yields and high costs, leaving China still reliant on Dutch, Japanese, and U.S. equipment it cannot yet replicate domestically. Intel CEO Pat Gelsinger argued at Davos 2024 that China faced a "10 year gap and a sustainable 10 year gap" as a result of U.S. export control policies. The gap is not closing, Sullivan's camp argues—it is widening, and loosening controls now would squander the time gained by U.S. restrictions.

Both Sullivan and Sacks share a core assumption: maintaining American semiconductor advantage is what matters most. Sullivan pursued this through strategic denial to degrade Chinese capabilities over time. The Trump Administration has taken a similar approach to the Biden administration's "tall fences, little yards" approach under State Department's new Pax Silica alignment—a strategic partnership network announced in December 2025 bringing together Japan, South Korea, Singapore, Israel, UAE, UK, and Australia to "build a secure, prosperous, and innovation-driven silicon supply chain from critical minerals and energy inputs to advanced manufacturing, semiconductors, AI infrastructure, and logistics."

The project is led by Under Secretary of State for Economic Affairs Jacob Helberg with the explicit aim of reducing U.S. dependencies and protecting U.S. technology and infrastructure from "undue access or control by countries of concern." In short, it aims to keep China out of U.S. and allied supply chains. The initiative aims to develop an insulated ecosystem spanning critical minerals, energy infrastructure, advanced manufacturing, and AI deployment within a network of core partners. Members are expected to build "trusted technology ecosystems" and coordinate responses to China's "overcapacity and dumping." The major difference from Sullivan's strategic denial approach is President Trump's willingness to permit sales of older generation chips to Chinese companies.

Both Sullivan's export controls and Trump's Pax Silica rest on the same premise: U.S. control of semiconductors and their supply chains creates dependencies for both partners and adversaries that give Washington long-term leverage. Sullivan believed widening the technology gap through export controls would keep allies locked into American technology and adversaries locked out.

Trump's Pax Silica believes coordination with allies while maintaining tiered technological advantages achieves similar structural outcomes—long-term dependence on U.S. innovation—while generating revenue for American firms and binding U.S. allies into long-term supply chain commitments. The core difference is that the Biden administration weaponized the Foreign Direct Product Rule to freeze China's tech capabilities in place. Trump monetized access for older chips with the aim of keeping Chinese companies dependent on older U.S. technology while restricting them from the newest U.S. innovation.

There is no clear "right" answer on which policy works best. Trump's Pax Silica may have a slight edge in that it also addresses the other end of the supply chain—deliberate efforts to reduce U.S. dependence on China's monopoly in rare earth minerals processing. But both Sullivan's export controls and Trump's Pax Silica share the same underlying assumption: that semiconductor advantages create lasting strategic leverage. This rests on treating chips as the binding constraint in technological competition. Chips remain the center of gravity for U.S. technology strategy, and that focus is justified. But semiconductor advantages mean little if the materials enabling every chip generation remain under Beijing's control.

China holds astounding advantages at the base of the technology supply chain and has actively sought to restrict U.S. access in retaliation to U.S. export controls. This matters because while the United States maintains leads in advanced semiconductor design and cutting-edge fabrication, the U.S. will have to focus both on innovating across the supply chain to preserve its existing lead in chip manufacturing as well as catching up to China's mineral and rare earth monopoly.

According to the International Energy Agency, China is the leading refiner for 19 of 20 strategic minerals, controlling 70 percent average market share. This concentration has intensified in recent years despite pledges to diversify supply chains. China possesses near-monopoly conditions for materials like gallium (98 percent of global production) and germanium (68 percent). But the numbers understate the problem. China's dominance extends across the supply chain from extraction to processing expertise, including accumulated industrial knowledge, workforce capability, and operational experience which, presently, no other country can match. Mineral experts I have spoken with argue these advantages will take years to be replicated in the West.

AI experts in Washington I interviewed see it differently. They argue U.S. critical mineral gaps are easier to bridge than China's semiconductor gaps. Alvin Camba's recent analysis in War on the Rocks offers a useful framework here. He explains that rare earth leverage "burns" while semiconductor leverage compounds. In other words, Beijing's mineral restrictions are most potent at the moment of imposition, but they create a cadence of impacts which can have a jumpstarting effect. First, the price spikes, which then mobilize Western governments to unlock capital and accelerate projects which drive diversification.

But closing those gaps would still require sustained alignment of capital investment, regulatory reform, streamlined permit processes, research infrastructure, and long-term procurement guarantees. The prospects for that alignment in the United States seem grim following Trump's targeting and defunding key universities critical for innovation in the sector and ongoing targeting of core allies necessary to support western reindustrialization of critical and rare earth element processing.

Meanwhile, China is increasingly confident in demonstrating its leverage over mineral supply chains. Beginning in mid-2023, Beijing implemented export restrictions on strategic minerals. In August 2023, China introduced licensing requirements for key rare earths, notably gallium and germanium. By December 2024, China banned all exports of these materials to the United States, along with antimony and superhard materials. In April 2025, China added seven medium and heavy rare earth elements—samarium, gadolinium, terbium, dysprosium, lutetium, scandium, and yttrium—to export control lists. By October 2025, China expanded restrictions to cover rare earth production equipment and related technologies. Following Trump-Xi discussions in late October 2025, China suspended many of these restrictions for one year beginning November 2025, though the core licensing system for rare earths implemented in April 2025 remains in force.

Beijing showed that it too can use strategic denial in its competition with Washington. Chinese officials created immediate pressure on Western mineral-heavy industries without fabricating a single advanced semiconductor. Then, Chinese authorities strategically dialed restrictions back as bargaining leverage. However, Beijing may have overplayed its hand. Camba argues correctly that China's export controls accelerated international coordination through the Quad, the U.S.–Australia Critical Minerals Partnership, and the G7 Minerals Security Partnership to pool stockpiles, share data, and align finances.

Refining capacity and supply chain control compound over decades of expertise, methods, and experience. These advantages are unlikely to be overcome quickly through investment alone. The U.S. let China build this lead when it reduced its strategic stockpile and offshored materials processing decades ago.

This partly explains Beijing's confidence. Deng Xiaoping famously said in the 1980s, "The Middle East has oil. China has rare earths." China doesn't necessarily need to match American chip design today to exercise leverage over global technology supply chains tomorrow. Beijing's grip on mineral processing gives it options to restrict, delay, or reshape markets in ways that compound rather than depreciate.

Pax Silica lays the foundation for a new semiconductor order, but Trump's actual implementation reveals a deeply transactional logic beneath the institutional framework. The official structure suggests membership determines access—Pax Silica signatories get current-generation chips, non-members receive degraded technology or nothing at all. Yet Saudi Arabia receives advanced Blackwell chips despite not joining the coalition, while the UAE's G42 maintains deep Chinese research ties even after formally divesting from Chinese firms to secure U.S. chip access. The pattern shows chip distribution operating through deal-making rather than exclusively via institutional commitment.

The UAE is a unique case. The UAE, for now, is the only GCC country part of Pax Silica. The UAE had deep AI partnerships with Chinese firms in 2023, then U.S. policymakers used these ties to justify an attempted block of Emirati AI giant G42 from accessing U.S. advanced chips. Eventually, G42 agreed to divest from Chinese technology and severed ties with ByteDance and Huawei as a precondition for chip access. Another Abu Dhabi investment vehicle—Lunate— took over management of G42's Chinese-focused fund.

The UAE's calculus favors U.S. partnership for the time being because the U.S. has promised and slowly delivered on providing technology Beijing cannot match. In November 2025, the Commerce Department approved shipment of approximately 35,000 high-end Nvidia chips to Saudi Arabia's Humain and the UAE's G42. Gulf states are receiving Blackwell-generation chips—more advanced than the H200s approved for China. In raw silicon terms, the UAE is leapfrogging China entirely.

Reported plans dwarf even these initial approvals. The Trump administration is reportedly considering deals that would allow more than one million advanced AI chips to be exported to the UAE alone—enough to build one of the world's largest AI training clusters. The UAE has announced construction of a five-gigawatt AI campus in Abu Dhabi, the largest AI infrastructure project globally, and OpenAI is weighing opening a data center there. In neighboring Saudi Arabia, Humain received commitments for 18,000 Blackwell chips in a first phase, with Amazon investing more than $5 billion alongside Humain to develop an "AI zone" with AWS infrastructure.

Yet chips alone do not fully make an AI ecosystem, and this edge only matters if the UAE can translate hardware access into indigenous capabilities. U.S. pressure on the UAE to divest from China is unlikely to untangle those ties. Both countries maintain deep ties at the personnel and institutional level, creating channels likely to outlast corporate partnerships.

The lesson for Pax Silica is that dependencies built on technology advantages remain contingent on maintaining those advantages. If U.S. terms become too restrictive, Chinese alternatives remain available. If China does not catch up to U.S. capabilities, the UAE will still be in an advantageous position.

The Trump administration has adopted what Alasdair Phillips-Robins describes as the "big announcement, slow follow-up" tactic. He explains that after promising millions of chips to Saudi Arabia and the UAE in May 2025, it took six months for first approvals, with Commerce signing off on just 35,000 chips—far less than publicly discussed figures. The delay came as Commerce Secretary Howard Lutnick pushed Emirati leaders to follow through on their promises of U.S. investment before signing off on exports, with national security concerns also holding up approvals.

The pull factor for GCC countries is also already delivering major wins for Gulf AI ecosystems. The Gulf states' "triple advantage"—cheap electricity, massive desalination capacity for data center cooling, and strategic geography serving four billion users with sub-100-millisecond latency—creates structural pull for U.S. technology firms seeking to escape domestic power constraints and permitting bottlenecks. Projected U.S. AI data center demand of 130 gigawatts by 2030 far exceeds available domestic generation capacity under development. These power constraints make the Gulf's advantages a potential structural necessity for U.S. AI ambitions which Washington cannot replicate domestically.

This muddies U.S. leverage. The United States can condition chip exports on Gulf states ending joint military exercises with China. Washington can demand verifiable evidence that the UAE is complying with its demands for limiting Chinese access to critical U.S. technology. But the real question lies in how the U.S. responds if officials do not find the UAE is operating within U.S. requirements—or how the UAE responds if U.S. domestic political constraints threaten to undermine the original terms of cooperation.

The GCC's commitment to trillions in AI spending built on the back of American chips is a strong enough economic incentive that Trump will take a risk and monitor compliance in real time.

Finally, we get to Trump selling H200s to China. What does it mean that Trump can approve selling advanced chips to China for the right price? It means semiconductor advantages work differently than most strategic assets. Trump can negotiate chip access deal by deal. He cannot, however, negotiate away mineral dependencies that took decades to build. The Administration's frantic efforts to secure critical mineral stockpiles—threatening Greenland invasions, leveraging Ukraine and DRC peace negotiations for resource access, rushing new refining plants—reflects recognition of this asymmetry.

China's mineral dominance operates by different rules entirely. Beijing's lead in processing and refining has compounded over decades and continues to appreciate. Chinese strategists accept near-term chip performance gaps to maintain long-term control over materials enabling all chip generations. When Trump offered H200 chips, Beijing appeared to reject them, with officials indicating they "want semiconductor independence" and prefer to "prop up Huawei." From China's view, mineral dominance provides leverage over the entire technology stack because every semiconductor, battery, and data center requires processed material inputs.

Both Biden and Trump treated semiconductors as strategic assets comparable to oil in the twentieth century. Control over oil supplies has shaped international order for generations. But what if semiconductors behave more like steel in the nineteenth century? Steel was a critical input, but its technological frontier advanced and capabilities diffused to new countries, which created new centers of innovation and expertise. Britain dominated steel production in 1850. By 1900, American and German mills had surpassed it through newer processes and scale of production. Semiconductors could follow this pattern—each new generation of chip creates opportunities for competitors to leapfrog through new innovation, process optimization, or simply outspending incumbents.

But the same story could also be true for China's lead in critical mineral refining. One major investor in the mining sector I interviewed argued that once supply chains realign and processing and refining technology diffuses to other parts of the world, the mineral race will slow and so will dependence on China. This is perhaps the emergence of the economic dimensions of multipolarity as global supply chains adjust to reduce reliance on current problematic chokepoints, wherever they may exist.

Washington has to both manage its chip lead and bridge the mineral gap. That is not a small feat. Pax Silica may succeed on its own terms, but Trump's transactional technology policies also have downsides. If chips are a source of American power, and that dominance can be accessible to other countries for a price, then semiconductor access operates as a service to be purchased, not a strategic asset to be protected. China's rejection of H200 chips suggests Beijing may wholly reject the premise of any form of dependency on western technology moving forward. For now, China is accepting performance gaps today to maintain structural control over critical minerals tomorrow—with a belief that they will eventually overcome the performance gaps.