English
The United States and China have entered a decisive new phase in the technology cold war. On January 13, 2026, the U.S. Bureau of Industry and Security (BIS) abandoned its blanket presumption of denial for certain advanced AI semiconductors bound for China, replacing it with a case-by-case licensing framework for NVIDIA's H200 and AMD's MI325X graphics processing units. The policy shift, effective immediately, imposes a 25% tariff under Section 232 and a 50% volume cap on eligible shipments relative to U.S. domestic sales. Beijing retaliated within days with its most stringent export controls on rare earths and permanent magnets, leveraging its near-monopoly on global processing to threaten U.S. defense supply chains. This calibrated decoupling is fragmenting global semiconductor supply chains, forcing allied nations to pick sides, and accelerating the creation of two parallel AI ecosystems that could reshape global innovation dynamics for years.
What Is the BIS Case-by-Case Licensing Framework?
The BIS rule, announced on January 13, 2026, revises the license review policy for exports of advanced semiconductors to China and Macau. Under the new framework, exporters of NVIDIA H200 and AMD MI325X chips—and similar devices with Total Processing Performance (TPP) below 21,000 and DRAM bandwidth under 6,500 GB/s—may apply for case-by-case review. To qualify, exporters must certify that U.S. domestic demand is fully met, that China-bound shipments do not exceed 50% of U.S. volumes, that products undergo independent third-party testing in the United States, and that rigorous Know Your Customer (KYC) procedures are in place. The rule follows President Trump's December 8, 2025 announcement allowing such shipments under strict conditions. BIS Under Secretary Jeffrey Kessler stated that the policy aims to "evolve export controls with technology changes while protecting national security."
The BIS export control framework also requires exporters to identify end users, prevent unauthorized access, and maintain detailed records. Reexports from other countries and in-country transfers remain under a presumption of denial, effectively limiting the policy to direct U.S.-to-China shipments. The rule introduces a 25% tariff on foreign-produced chips re-exported to third countries, protecting domestic AI infrastructure investment.
China's Rare Earth Retaliation: A Strategic Countermove
Beijing responded to the BIS rule by tightening its already stringent export controls on rare earth elements and permanent magnets. China's Ministry of Commerce (MOFCOM) announced that its October 2025 Announcement No. 61—which applies the foreign direct product rule (FDPR) to rare earths for the first time—remains in full effect despite a November 2025 suspension of certain provisions. Under these rules, foreign firms must obtain Chinese government approval to export magnets containing even trace amounts (0.1%) of Chinese-origin heavy rare earths or using Chinese technology. Starting December 1, 2025, companies affiliated with foreign militaries are largely denied export licenses, and military-use requests are automatically rejected.
China controls roughly 70% of global rare earth mining, 90% of processing, and 93% of permanent magnet manufacturing, according to the International Energy Agency. The rare earth supply chain dependency is particularly acute for U.S. defense systems, including the F-35 fighter jet, submarines, Tomahawk missiles, and radar systems. Export controls have triggered sixfold price spikes outside China, with licensing approval rates below 25% for European firms. Over 80% of European companies depend on Chinese supply chains for critical minerals essential to defense, EVs, and renewable energy.
Supply Chain Fragmentation and the Rise of Parallel Ecosystems
The calibrated decoupling is fragmenting global semiconductor supply chains. The geography of chip manufacturing has shifted, with the U.S. now commanding approximately 22% of advanced chip capacity—up from 12% in 2020—driven by $52 billion in CHIPS Act disbursements catalyzing $450 billion in private investment. However, supply chain fragmentation adds 25–35% to landed costs for advanced chips. Procurement professionals are adopting multi-region sourcing mandates, "just-in-case" inventory buffers of 6–12 months, and a new "friendly shoring" pattern concentrating manufacturing among allied nations including Japan, South Korea, Taiwan, the Netherlands, Germany, and the United States.
Two parallel AI ecosystems are emerging. The U.S.-led Chip 4 Alliance coordinates export controls, while China's domestic ecosystem—led by SMIC (which has achieved 5nm manufacturing), Huawei's HiSilicon, and Cambricon—remains two to three generations behind but is advancing rapidly. China has mandated that all AI models use domestically hosted compute infrastructure and register algorithms with the Cyberspace Administration of China, creating a walled garden. The World Economic Forum ranks geoeconomic confrontation as the top global risk for 2026–2028. Full decoupling could cost U.S. firms $77 billion in lost semiconductor sales in the first year, while Chinese R&D investment targets $50 billion by 2030.
Allied Nations Forced to Pick Sides
The trilateral U.S.-Japan-Netherlands framework has been central to controlling semiconductor manufacturing equipment exports. Japan's 23-item equipment controls, enacted in July 2023, cover six major process steps and have significantly impacted Japanese equipment makers like Tokyo Electron and Nikon. The proposed MATCH Act (Multinational Adjustment of Technology in Critical Hardware Act), introduced in the U.S. House in April 2026, would extend controls to older-generation DUV lithography tools, further pressuring allies. South Korea faces particular exposure, with Samsung and SK Hynix operating major facilities in China. Both companies received annual licenses in January 2026 replacing their expired Validated End User (VEU) status, introducing annual uncertainty. The Japan-South Korea-US techno alliance has recalibrated into a pragmatic partnership focused on critical mineral supply chains, AI, and quantum computing, but faces structural vulnerabilities from U.S. trade protectionism and unresolved historical grievances.
Expert Perspectives
Analysts at the Center for Strategic and International Studies (CSIS) warn that China's rare earth controls represent a fundamental shift from treating rare earths as market commodities to strategic assets under national security oversight. A multi-institutional analysis argues that China is weaponizing control—not scarcity—through temporary, reversible restrictions that maintain pricing power and extract strategic concessions while discouraging Western alternative investment. Rebuilding independent rare earth supply chains could take 20–30 years, far exceeding the current geopolitical window. Western nations face a narrow 12–18 month window to act decisively or accept prolonged structural dependency.
Industry observers note that despite the BIS rule allowing limited sales, China has blocked H200 imports at customs and warned domestic companies against purchasing the chips, favoring domestic alternatives like Huawei's Ascend series. This suggests that even where U.S. policy permits exports, Chinese industrial policy may restrict imports, deepening the bifurcation.
FAQ
What is the BIS case-by-case licensing framework for AI chips?
The BIS case-by-case licensing framework, effective January 13, 2026, allows exports of NVIDIA H200 and AMD MI325X chips to China under strict conditions: a 25% tariff, 50% volume cap relative to U.S. sales, independent third-party testing, and KYC compliance. It replaces the previous blanket presumption of denial.
How did China retaliate against the new U.S. chip rules?
China tightened export controls on rare earths and permanent magnets, applying the foreign direct product rule for the first time. Foreign firms must obtain Chinese approval to export magnets containing Chinese-origin rare earths, and military-related requests are automatically denied. This threatens U.S. defense supply chains.
What are the TPP and DRAM bandwidth limits for eligible chips?
Eligible chips must have Total Processing Performance (TPP) below 21,000 and DRAM bandwidth under 6,500 GB/s. The NVIDIA H200 (TPP: 15,832) and AMD MI325X (TPP: 20,800) fall within these thresholds.
How are allied nations like Japan and South Korea affected?
Japan and the Netherlands coordinate equipment export controls with the U.S. under a trilateral framework. South Korea's Samsung and SK Hynix face annual license uncertainty for their China facilities. The MATCH Act proposes extending controls to older DUV tools, pressuring allies further.
What is the economic impact of the AI chip decoupling?
Full decoupling could cost U.S. firms $77 billion in lost semiconductor sales in the first year. Supply chain fragmentation adds 25–35% to advanced chip costs. China targets $50 billion in R&D investment by 2030 to build domestic alternatives.
Conclusion: A New Era of Managed Competition
The January 2026 BIS rule and China's rare earth countermeasures mark the most consequential geopolitical development in AI supply chains so far this year. The calibrated decoupling is not a full break but a managed competition that creates permanent structural dependencies and parallel ecosystems. Companies must navigate a fragmented regulatory landscape, maintain dual supply chains, and invest in compliance intelligence. The outcome will shape not only the semiconductor industry but also the broader balance of technological power between the world's two largest economies for the next decade.
Follow Discussion