Who the AI Wall Actually Stops
Washington spent the year turning AI from a product it sold into a privilege it rations — and by November the wall was stopping its friends, while China held the gate from the other side.
In Hong Kong this month, roughly 600,000 Nvidia H200 chips sit stranded in port — chips Washington licensed for sale to China in April with a 25 percent cut of the proceeds going to the U.S. Treasury [1]. They were not stopped by the American wall. Beijing is withholding the import approvals deliberately, to force its own processors into use and cut its dependence on American technology [2]. Fourteen months after Washington began turning AI from a product it sold into a privilege it rations, the question is who the rationing actually stops. A year and a half ago the regime ran the other way. In May 2025 the administration scrapped the Biden-era diffusion rule so American firms could supply allies freely, and the boom in nations building their own AI capacity that followed was the product of that open door: Humain's $10 billion AMD deal, 18,000 Nvidia GB300s headed to Saudi Arabia [3]. The same administration then spent the next year dismantling what it had built. The reversal began with the friends, not the adversary. In September 2025, Commerce revoked the license-free "validated end-user" status — the standing exemption that let a firm ship to its own operations in China without asking Washington each time — from Samsung, SK Hynix, TSMC, and Intel. Their licenses would maintain existing operations but never authorize upgrades [4]. China's Ministry of Commerce registered its own reading.
The U.S. move, driven by self-interest, has turned export controls into a tool, and will cause a serious negative impact on the stability of global semiconductor industrial and supply chains. — Ministry of Commerce
By June the rationing had a name and a doctrine. Washington launched Pax Silica with 34 other nations [5], and at the launch Deputy Secretary of State Christopher Landau defined the mission.
Pax Silica exists to keep these technologies and the future growth of all of us in trusted hands. — Christopher Landau
The same month, Anthropic suspended its advanced Mythos and Fable models in India on a U.S. government directive [5]. Trust narrowed from nations to vetted persons: Indian tycoons Mukesh Ambani and Gautam Adani were designated "trusted custodians of Western technology" as they pledged $200 billion for AI infrastructure [6]. By August the plumbing existed — a $50 million Panama platform running AI-assisted tracking of chips and critical minerals for the two dozen "trusted suppliers" [7]. In April 2026 Nvidia secured administration approval to sell H200 chips to China — a licensed deal carrying a 25 percent government cut of the sales, production resuming in late March [1]. The allies had been pushed onto case-by-case licenses that never authorize upgrades; the adversary got a priced lane. The record names the reason that lane stayed open: the administration avoided policies targeting China to keep critical rare-earth shipments flowing ahead of an agreement expiring January 10 [8]. Enforcement moved inside private firms. Nvidia now requires Southeast Asian cloud providers to register with its Fleet Intelligence tracking system or lose partner status and preferred chip access, has removed half its pre-approved Asian buyers, and instructs providers to keep at least 51 percent of services for customers outside China [9]. Jensen Huang described the reach of the program.
We've studied all their deployments of our GPUs and counted every one. — Jensen Huang
Congress is legislating the same job into the defense authorization, folding three chip-tracking bills into the FY2027 measure [8]. And the labs themselves asked for the custody.
my view here is it has always been very strange that this technology is being built by a private company. — Dario Amodei
Sam Altman and Elon Musk endorsed the proposal [10]. China is building the mirror image from the opposite end. Its draft rules would require licenses to export the core technology of "globally leading" models while explicitly exempting the interfaces that let outsiders call up those models — the exact inverse of Washington's new licensing of remote computing access for companies in China and Macau [11][12] — and Beijing announced them in its own vocabulary.
the new controls were necessary to guard against abuse and malicious use and to ensure that A.I. always remains under human control. — Ministry of Commerce
The diversion through partner territory was real and enormous. Singapore's Megaspeed International is under investigation on both sides of the Pacific, and U.S. prosecutors accused Super Micro associates of using falsified paperwork to divert roughly $2.5 billion in Nvidia-powered servers to China [9]. American chips have been barred from China since August 2022 and have leaked anyway for five years, at every level — from shell companies in Singapore to the Nvidia chief executive's own phone [13][14]. Jensen Huang made the case that the wall costs American AI while it leaks anyway.
China's AI moves on with or without U.S. chips. It has to compute to train and deploy advanced models. The question is not whether China will have AI, it already does. The question is whether one of the world's largest AI markets will run on American platforms. — Jensen Huang
Beijing has now enforced the wall from the other side. The block on the 600,000 H200s is stated industrial policy, and the buyers rerouted around the stranded chips — Alibaba, ByteDance, and Tencent ordered more than three million export-compliant RTX Pro 5500 cards worth over $35 billion instead [2]. The adversary's access now turns on Beijing's choice, not Washington's. For the friends, the edge is Washington's discretion — which is why, in that same June, Indian officials were left seeking assurance that trusted partners' access would not be abruptly cut off [5].
- 1. Jensen Huang Urges U.S. to End China Chip Bans
- 2. China Blocks 600,000 Nvidia H200 Chips in Hong Kong
- 3. Trump Policy Shift Drives Global Sovereign AI Surge
- 4. US Revokes Chip Equipment Waivers for Samsung, SK Hynix, TSMC
- 5. US and 34 Nations Launch Pax Silica AI Initiative
- 6. US Restricts AI Chip Exports to Southeast Asian Nations
- 7. US Launches Pax Silica Program to Fast-Track AI Trade
- 8. Lawmakers Push Chip Restrictions After Nvidia CEO Spy Link
- 9. Nvidia Mandates GPU Tracking Amid US Congressional Scrutiny
- 10. Anthropic CEO Proposes Government AI Control and China Restrictions
- 11. US Tightens AI Chip Controls Over Chinese Model Distillation
- 12. China Proposes AI Export Controls While Approving DeepSeek, MiniMax
- 13. U.S. AI Chips Smuggled Into China Despite Export Controls
- 14. Nvidia CEO Jensen Huang Resigns Amid Espionage Scandal