Chips & compute
The US just made the chip ban follow the owner, not the address
Spin up a subsidiary in Singapore? Doesn't matter anymore. The rule asks who your parent is — and Commerce quietly admitted it failed to ask for a year.
The answer
On 1 June 2026 the US tied AI-chip controls to Chinese parentage, killing the overseas-subsidiary workaround.
On 1 June 2026, US Commerce / BIS guidance spelled out that advanced-AI-chip licensing applies to any business headquartered in or parented by a Chinese company — full stop, wherever it sits. For years the workaround was to stand up an overseas unit in Malaysia, Singapore or the UAE and route the chips through it. According to US officials, that workaround was used. At scale. TrendForce estimates hundreds of thousands of Nvidia and AMD chip servers may have reached Chinese-controlled entities through those subsidiaries. The door is now shut — the question is how much walked through it before anyone noticed.
Why this one's different — and why it's also an admission
Geography-based bans are leaky by design — companies are mobile, addresses are cheap. An ownership-based test is a different animal: you can move an office overnight, but you can't quietly swap out who your parent company is without it showing up in filings. That's why this clarification matters more than the headline length suggests. But don't miss the embedded admission. The guidance notes that firms which already bought export-controlled chips without a licence may continue operating them until further notice. That's not a loophole; that's Commerce telling you, in passive-voice bureaucratese, that it knows unlicensed stock is already out there and isn't making them pull it. Former State Department official Chris McGuire was more direct, telling Al Jazeera: 'Chinese companies have been buying these chips, very likely at scale.'
This guidance, which notes both that these sales require a license and that firms could continue to operate already-purchased components, suggests that American firms had illicitly sold these components to Chinese firms.
Put the before and after side by side:
| Before (pre-1 June) | After (1 June 2026) | |
|---|---|---|
| Enforcement test | Shipping address / buyer location | Ultimate parent company headquarters |
| Overseas subsidiary | Possible grey area if 3rd-country registered | Covered — licence required regardless of address |
| Chips in scope | Nvidia Blackwell (China-bound) | Nvidia Blackwell + Rubin, AMD MI350X — any owner with Chinese parent |
| Already-purchased stock | Ambiguous | May keep operating 'until further notice' |
| TSMC due diligence | Not required | Still not required — remaining gap |
The bottom two rows are the story the press release isn't telling.
Who's sweating — and who isn't
Nvidia said it was already compliant — its spokesperson told Al Jazeera that 'licences are required to ship controlled products to PRC-headquartered companies, consistent with our existing approach.' Make of that what you will: either Nvidia genuinely had ownership-based due diligence running before the rule existed, or it's getting ahead of the narrative — probably some of both. AMD is named in the guidance too, so the heat isn't Nvidia-specific. The real compliance pain falls on distributors and cloud resellers with large indirect sales channels, who now have to trace their buyers' parent trees. That's an ops and legal headache, and the guidance offers no new enforcement machinery to help them do it.
Nvidia said it had already been operating in keeping with the clarified rules: 'The guidance reaffirms that NVIDIA's sales and vetting process is correct — consistent with our existing approach, licences are required to ship controlled products to PRC-headquartered companies.'
The gap that's still open
McGuire flagged it directly in a post on X: 'This statement does NOT say that BIS will enforce the parts of US regulations requiring TSMC to do enhanced due diligence on AI chip orders. This is a massive loophole that still needs to be closed.' TSMC manufactures the chips Nvidia designs. If downstream export controls tighten without corresponding upstream production due diligence, the control chain has a weak link at the foundry level — Chinese firms could, in theory, use the same third-country cut-outs to commission chip production at TSMC, not just buy finished accelerators. Watch for follow-on BIS rulemaking — and for allied governments to face increasing pressure to align their own frameworks. This guidance is meaningful and structurally durable. It is also not the last word.
Frequently asked questions
What's the actual change on 1 June 2026?
Will this really stop workarounds?
How did the loophole get this big?
What does 'continue operating until further notice' in the guidance mean?
What's left to fix?
Sources
- US says ban on AI chip shipments applies to Chinese firms outside China — Al Jazeera, 1 June 2026
- U.S. takes step to halt Nvidia AI chip shipments to Chinese firms outside China — CNBC, 31 May 2026
- U.S. Moves to Block AI Chip Exports to Overseas Chinese Units as Loophole May Have Fueled Large Shipments — TrendForce, 1 June 2026
- Commerce Department Admits Failure To Enforce AI Export Controls on China — Foundation for Defense of Democracies, 2 June 2026