Three Tracks, One Clock

On June 22, China's Ministry of Commerce placed 10 US entities on its export control list and banned procurement from 46 American defense firms ^[1]. The targets included MP Materials and USA Rare Earth, the only US rare-earth miners, plus Lockheed Martin, Raytheon, General Dynamics, and Boeing Defense. No American chip designer appeared on either list. That omission is the latest data point in a pattern visible across a dozen separate Chinese moves this year. China's response to US semiconductor controls runs on three tracks at once, each in a different domain. Circumvention persists where China still needs American chips. Substitution accelerates where it can build its own. Retaliation stays on the defense-minerals track, where China holds near-monopoly leverage. Chip-specific countermeasures remain absent, consistent with China's position as net chip importer. **Circumvention.** At least seven Chinese military-linked universities, including Beihang and Northwestern Polytechnical, both on US blacklists, sought to lease Nvidia H200 chips through remote compute access, exploiting a framework that restricts physical shipments but not compute leases ^[2].

Just like it would be nonsensical for the American military to use Chinese technology, it makes no sense for the Chinese military to depend on American technology. — Nvidia

The US closed one channel on May 31. The Bureau of Industry and Security clarified that licensing requirements apply to all businesses headquartered or parented in China regardless of subsidiary location, targeting chips flowing through countries like Malaysia ^[3]. Another gap remains at TSMC, where the US has accused ASML of illegally shipping EUV technology to China. The Netherlands declined to act, calling its existing rules already among the strictest in the world ^[4]. China has separately developed its own prototype EUV machine using former ASML engineers ^[4]. **Substitution.** China certified nine domestic AI chips under a national security framework, the first time AI chips were included as a standalone category ^[5]. Self-sufficiency rose from 10% in 2021 to 41% in 2026, with Morgan Stanley projecting 86% by 2030 ^[3].

10% → 41% → 86% AI chip self-sufficiency, 2021–2030 (projected) — Morgan Stanley ties the trajectory to US export controls forcing domestic development ^[3]

The build extends past components into architecture and ecosystem. Huawei's Tau Scaling Law targets 1.4nm-equivalent performance by 2031 without EUV lithography, optimizing system-wide signal delays instead of shrinking transistors ^[6]. He Tingbo, Huawei's chip chief, promised a major reveal before winter 2026 ^[6]. Alibaba delivered 560,000 Zhenwu-series chips to more than 400 customers across 20 industries, with the Chinese government barring firms from buying Nvidia chips ^[7]. A $295 billion AI data-center network led by the NDRC mandates 80% domestic sourcing for core technology including AI chips ^[8]. DeepSeek optimized its V4 model for Huawei Ascend chips, integrating labs, state financing, and domestic hardware into a self-sustaining stack ^[9].

It would be a "horrible outcome" if AI models around the world eventually ran best on non-American hardware. — Jensen Huang

That warning comes from the CEO most exposed to China's substitution drive. **Retaliation.** China imposed export controls on rare earths in April 2025, affecting US aerospace, defense, and semiconductor manufacturing ^[10]. Scandium, vital for chip production, is among the controlled minerals. Rare-earth export value surged 45% in the first five months of 2026 despite only a 2.2% volume increase, as global buyers paid premiums to secure supply ^[11]. China processes roughly 90% of the world's rare earths ^[10]. On June 22, MOFCOM's sanctions landed on the only US rare-earth miners and 46 defense firms ^[1].

any relevant export activities currently underway must cease immediately — Ministry of Commerce

On June 4, MOFCOM spokesperson He Yongqian condemned US chip export controls as discriminatory and accused them of disrupting global semiconductor supply chains ^[12]. No chip-specific countermeasures followed. The verbal condemnation was the entire chip-domain response. Meanwhile, the retaliation infrastructure keeps expanding, but not into chips. A June 1 State Council decree authorizes Beijing to adopt retaliatory countermeasures against countries that restrict Chinese investors ^[13][14]. China ordered firms to defy US Iran sanctions ahead of the May Trump-Xi summit ^[15] and blocked the EU's Nuctech anti-subsidy probe under new regulations shielding Chinese firms from foreign interference ^[16]. These are legal instruments that could be repurposed against US chip firms. None has been. The pattern points to a structural constraint rather than a tactical choice. China retaliates where it has monopoly leverage and substitutes where it needs to escape dependence. It does not retaliate where doing so would cut off its own supply. Self-sufficiency stands at 41%, meaning 59% of China's AI chip needs are still imported ^[3].

Each side retaliates in the domain where it holds leverage: the US controls what China needs (advanced chips), China controls what the US needs (minerals). The asymmetry is the story ^[10][3].

The May Trump-Xi summit captured the imbalance. The US cleared roughly 10 Chinese firms to purchase up to 75,000 Nvidia H200 chips each under a new licensing regime, but no chips shipped ^[17]. Jensen Huang joined the US delegation and argued for selling regulated chips to China. China did not lift its rare-earth export restrictions, which remained 50% below pre-restriction levels ^[17].

So, we’re going to put in U.S. best practices, U.S. values on this, and then roll those out to the world. — Scott Bessent

The June 8 Pentagon expansion of the 1260H blacklist to 188 Chinese firms, adding Alibaba, Baidu, Tencent, and BYD, came after that summit, prompting MOFCOM to threaten countermeasures ^[18]. Those countermeasures materialized June 22, on defense firms and rare-earth miners, not chip designers ^[1]. The State Council decree sits ready, an enabling instrument that becomes costless to deploy only when substitution makes US chips expendable, projected at 86% self-sufficiency by 2030 ^[13][3]. The race underneath is temporal. The US is racing to enforce controls before substitution makes them irrelevant. China is racing to substitute before enforcement closes every gap. Taiwan is tightening, considering stricter AI chip export bans and criminalizing smuggling of advanced hardware ^[19]. The allied front is not eroding. But it is racing against a clock set by China's build-out. The transformation from one-sided enforcement into mutual attrition depends not on China deciding to retaliate in chips. It depends on China no longer needing to import them. The legal framework is in place. The substitution curve is climbing. The question is which arrives first: complete US closure of every gap, or Chinese self-sufficiency that makes the gaps irrelevant.