Hamish Low retweeted
The Commerce Department has released a major new export control rule targeting China’s access to advanced AI compute.
It’s pretty good! Here’s what it does:
(1) Restricts Chinese access to overseas AI clusters (seeks to address one of China’s largest and fastest-growing sources of AI compute)
This is a new worldwide end-use control, which we’ve previously recommended (third option in image below).
A license is required if providing cloud AI compute to Chinese or D:5-headquartered countries, or if acting as an intermediary training models for these actors.
Technical requirements are:
- Applies to any data center or group of “linked data centers” or providing more than 12 million aggregate TPP of compute (the equivalent of around 760 H100s) of AI chips meeting the current controlled thresholds.
- “Linked data centers” are defined as sharing physical or virtual computing resources, jointly balancing workloads, or replicating data as a unified whole. So this is intended to capture multi-data center training runs.
This applies everywhere except the US. It also covers existing data centers, which have a 90-day deadline to get a license.
(2) New controls for AI chips going to Malaysia, Singapore, Indonesia, Thailand, and Brazil (seeks to address smuggling of AI chips to China via third countries)
The evidence for large-scale smuggling of AI chips to China via Malaysia and Singapore is strong, and there’s also evidence this is happening via Indonesia and Thailand. Brazil, on the other hand, is a bit of a head-scratcher.
This is a case-by-case license review policy, rather than a presumption of denial (that's the policy for chips going directly to China/D:5 countries).
BIS will evaluate each license on national security grounds.
There’s an exclusion for qualifying US-headquartered recipients, subject to reporting requirements.
(3) Formally repeals the AI Diffusion Rule
This has been a long time coming. This eliminates the country-tier system, the Validated End User authorizations (VEU) for data centers, and export controls on advanced AI model weights.
However, the replacement measures above address many of the gaps created by this, and, of course, the AI Diffusion rule wasn’t actually being enforced.
(4) Tightens controls on Chinese chip designs manufactured abroad
This modifies and strengthens the Biden-era “Foundry Due Diligence” rule, which applies to foundries and OSATs.
This seeks to prevent Chinese chip designers from circumventing export controls through overseas foundries and packaging facilities. It’s an important complement to the measures above.
—
So what are the gaps? More on this to come, but from an initial read:
(1) To determine remote-access eligibility, the rule requires due diligence on direct or indirect Chinese ownership or control. However, its licensing trigger is based around company headquarters and ultimate-parent headquarters, which could allow for some shenanigans around corporate structures.
(2) The rule doesn’t prescribe how AI compute providers must verify customers' identities, or require them to identify and verify every user accessing their compute. The rigor of these checks could vary widely across providers.
(3) The end-use restrictions don’t apply when compute is used exclusively for commercial recommender systems. So this opens a path for large-scale Chinese AI compute usage. This could be fine if verifying this kind of use is possible, but this is likely outside the current technical capabilities of many AI compute providers.
(4) The end-use restrictions don’t apply to data centers in the US. So a Chinese company can still access large amounts of AI compute via US data centers.
This is an interim final rule, meaning it would take effect upon publication in the Federal Register, while allowing 60 days for public comments.
Read the full interim final rule here: bis.gov/media/documents/rin-…
What the new BIS remote access rule does:
- Any data centre worldwide now needs a licence to give Chinese-headquartered firms access to more than 12M TPP of compute, about 760 H100s or 200 B300s. That includes chips already installed: existing providers get 90 days to apply.
- Every advanced chip export to Malaysia, Singapore, Thailand, Indonesia and Brazil now needs a licence, unless the buyer is US-headquartered and reports monthly to BIS.
- BIS says its existing rules always covered data centres whose main user is a Chinese firm, and invites self-disclosures.
- Compute used only for recommender systems is exempt, licences are case-by-case rather than presumed denied
- It finally formally rescinds the Diffusion Rule
!! BIS issues a rule around Chinese remote access bis.gov/press-release/bis-is…
This case for remote access makes sense in theory, but just doesn’t match the reality. If all the remote access was happening via US cloud providers and you saw this margin stacking, and there was clearer controls on access to and security of the chips that would be great.
But very obviously that isn’t what is going on. Much of this remote access is via non-US cloud providers like Megaspeed which isn’t giving you any margin stacking apart from NVIDIA’s and sketchy OEMs like AIVRES/Inspur (itself an entity listed firm!)
Those firms operate basically like off balance sheet vehicles for Chinese hyperscalers, so remote access there is as close to just selling them the chips as you can get, without literally doing that.
Some kind of intervention or governance framework is needed here. That could well account for remote access continuing, but clearly something needs to change if we are to get to the world that @dylan522p seems to want.
The devil’s advocate case for letting China rent American compute.
“If China has access to compute outside of China, it’s paying humongous margin dollars to the US ecosystem. Nvidia makes about $7 million of profit on every megawatt deployed. The cloud provider makes another $7 to 10 million at today’s pricing.”
“If ByteDance wants to compete with Meta on video recommendation globally, and you don’t let it buy the chips but you do let it rent them, Meta potentially has a fourth of the cost of compute.”
“Application-layer companies running on open models add up to well north of $10 billion in ARR. DeepSeek makes less than Baseten and Fireworks do serving its models. America is making so much money off Chinese models.”
One of the strong arguments for why renting chips to China is better than selling them is that it gives the US optionality to control them in the future. Unfortunately when the renter is Megaspeed that case breaks down, and remote access looks more liability than optionality 🧵
Hamish Low retweeted
Whether or not Alice Huang had been recruited by the Chinese Ministry of State Security has always been a tertiary issue.
What’s quite clear is that:
1. She has been a large Nvidia customer (and a sexual relationship with her is highly problematic from a corporate policy and governance perspective for Jensen on those grounds alone)
2. She is controlling a shell company Megaspeed dedicated to smuggling compute to both US Govenment-listed Chinese AI companies and to data center build outs in China itself - all in serious violation of US law, potentially making Jensen part of a criminal conspiracy to evade US export controls.
3. She is connected to an entire web of Chinese shell companies trying to do the same.
Jensen wasn’t fired because he had an affair with an alleged MSS agent. He was fired because he:
1. Had at least one and now, quite plausibly, multiple affairs with people in his reporting chain in violation of company policy. Something that has gotten numerous CEOs fired over the years, including at Intel and McDonalds.
2. Lied to his board about his
relationship and forced the company to put out false statements to the public, putting it in legal jeopardy.
3. Was involved in a sexual relationship with a Chinese smuggler that had violated US export controls laws and provided her with billions of dollars of Nvidia chips that facilitate that criminal activity.
4. By lying and hiding that relationship from his family, the company, and the public, he put himself in a position to be blackmailed by Chinese intelligence services.
5. He very likely ordered his general counsel to interfere in an investigation by US government into his mistress’ smuggling operation.
6. He presided over stunning and extensive compliance failures to investigate and prevent his chips from being smuggled illegally into Chinese mainland and being provided to PLA affiliated institutions, all in violation of US law.
This is a shocking indictment of him and his record and the board clearly had no alternative but to fire him.
Bombshell story. The spy narrative is collapsing already.
Is there clear & convincing evidence that Alice Huang was a Chinese spy?
- AG says no. (“no conclusions reached”).
- CIA says no. (“not yet convinced”).
- DNI says no (“not entirely accurate and over-torqued”).
- Ranking Democrat on House Intelligence (incoming Chairman) says no. (“today, at least, I don’t know”). This was immediately after he was briefed.
Who argues otherwise? Anonymous officials tell CNN “it strains credulity to believe that she was not acting as an agent of the Chinese government — simply by virtue of the fact that she is a successful Chinese businesswoman working with one of America’s most powerful companies.”
This is their proof after a year of investigating and surveillance? That she was a “successful Chinese businesswoman”?
At this point, the only crime that we know for sure happened is the leaking of the classified memo. Who did that and why?
An investigation should be opened.
Hamish Low retweeted
Commerce is reportedly considering new regulations to restrict China's access to AI chips for data centers overseas.
This is a long-standing gap in US export controls.
However, the rule in its reported form has three massive flaws:
1. It only applies to certain countries. This makes it fairly straightforward for Chinese customers to switch to data centers in non-covered countries.
2. It only applies to future AI chip exports. This means that all existing AI data centers used by Chinese customers like DeepSeek are unaffected.
3. It only applies to chips that are at least as powerful as NVIDIA's Blackwell generation. This means that data centers in covered countries can still provide chips like NVIDIA's H200 to Chinese customers. For reference, Huawei is not planning to produce a chip as powerful as the H200 until the end of 2027 at the earliest.
If we apply these conditions to our estimates of China's remote access from different countries* in 2027, the rule would affect at most 14% of China's total end-of-year 2027 compute, or 44% of remote access.
This is almost certainly a strong ceiling, however, as it assumes that all of this compute is Blackwells or better, and that Chinese customers do not shift usage to non-covered countries.
* country-level estimates are fairly rough - they estimate shares of our existing regional data using country-level data center pipeline statistics from the World Bank
Scoop: Commerce Department is moving quickly to cut off ways for Chinese customers to access advanced chips remotely. Important scoop from @leomschwartz
We previously reported that this was one of the issues that made Jensen most irate and angry at government officials when discussed during the Biden admin.
theinformation.com/articles/…
Hamish Low retweeted
Want the definitive lowdown on what has happened with Jensen, Megaspeed, export controls, and where we go from here? A new post from my colleague Erich on the IAPS Compute Policy team's Substack goes into it all! open.substack.com/pub/erwald…
The complete timeline of how we got to the current Jensen saga
"In October 2022, the US government restricts the export of advanced AI chips to China. The ban covered NVIDIA’s A100 chip as well as its then-upcoming H100. The rule caught NVIDIA by surprise, but the company quickly designed new versions (the A800 and the H800) that just skirted the performance thresholds so it could keep selling to China. When the US government closed that gap in October 2023, NVIDIA released the further pared-down H20 for the Chinese market.
In 2023 and 2024, Megaspeed appears and starts buying billions of dollars’ worth of NVIDIA chips. It had been the Singapore arm of 7Road, a Chinese gaming company with state-backed investors, until it was split off and relaunched as Megaspeed in March 2023. Corporate records show Alice Huang as Megaspeed’s managing director from March to November 2023, and its owner at the time as Huang Le Limited, a British Virgin Islands entity bearing her Chinese name. On LinkedIn, she has reportedly described herself as Megaspeed’s CEO. In November 2023, Huang Le Limited sold Megaspeed to a Singapore company owned by Megaspeed’s current CEO, and four days later Megaspeed became an NVIDIA Cloud Partner. In December 2023, Alice Huang met with Jensen Huang and a few colleagues at a restaurant in Singapore, and in June 2024, the two were photographed at a bar in Taipei.
In 2024 and 2025, reports of large-scale AI chip smuggling into China mount. For example, in August 2024, the New York Times and The Information each reported that large quantities of banned NVIDIA chips were being smuggled into China, with brokers using shell companies and staged data centers in Southeast Asia to move single orders worth more than $100 million. Jensen Huang denied that any smuggling was happening, telling Bloomberg in May 2025: “There’s no evidence of any AI chip diversion. These are massive systems … you’re not going to be putting that in your pocket or your backpack anytime soon.” That same month, NVIDIA accused Anthropic of telling “tall tales” about chip smuggling. In June 2025, Tim Fist and I published our second report on AI chip smuggling for the Center for a New American Security, summarizing many of these cases and suggesting ways of reducing smuggling. In July, the Financial Times reported that at least $1 billion worth of banned NVIDIA chips had reached China in three months.
From 2023 to 2025, NVIDIA tries to discredit critics at Washington think tanks. In 2023, according to The Information, NVIDIA tried to use its influence, including funding, to sideline Greg Allen, a supporter of export controls at the Center for Strategic and International Studies. (NVIDIA said at the time that it had never taken issue with his work.) In 2025, according to Sam Hammond of the Foundation for American Innovation, he posted that NVIDIA was “arming China’s military” and NVIDIA emailed his boss.
In September 2024, a Commerce Department official visits a Megaspeed data center in Malaysia. The New York Times would later report that US officials “were concerned that NVIDIA technology on site was still in boxes”. In early 2025, Alice Huang and Kang Wei, her business partner, reportedly pitched investors on a data center project, saying they had “no problem securing GPUs” and citing Alibaba as a client. A year after the visit, US officials reportedly still had no conclusive evidence that Megaspeed had smuggled chips to China.
In 2025, after heavy lobbying by Jensen Huang and NVIDIA, the US government loosens the AI chip controls on China. NVIDIA originally proposed selling the B30A, a version of its latest-generation Blackwell chips, but after internal discussions, in December 2025, the US government allowed it to sell H200s. The H200 is from the earlier Hopper generation but still a substantial upgrade over what was previously available in China. Back in July, at a press conference in Beijing, Jensen Huang had said: “I hope to get more advanced chips into China than the H20. … Export control[s] are things that are outside of our control and they can be quite disruptive to our business. It is our job only to inform the governments of the nature and the unintended consequences of the policies that they make.”
In summer 2025, Singapore authorities detain Alice Huang. According to the New York Times, they barred her from leaving the country and questioned her as part of an investigation into false import declarations, though the investigation into her was later closed for lack of evidence. NVIDIA’s general counsel then reportedly called Under Secretary for Industry and Security Jeffrey Kessler, who oversees export controls at the Commerce Department, to ask whether his agency was involved in the detention.
In October 2025, the Megaspeed case becomes public through a New York Times report. The report said that Commerce was investigating whether Megaspeed, which had bought “nearly $2 billion” worth of NVIDIA products, was helping Chinese companies evade US export controls. Most of these purchases went through Aivres, the US subsidiary of a Chinese server builder called Inspur, which is on the US Entity List for acquiring US technology in support of China’s military. NVIDIA said there was no evidence that its chips had been smuggled into China, and that Megaspeed had not raised any red flags. NVIDIA also said that Megaspeed was owned and operated by a company headquartered outside China, with no Chinese shareholders, and that it was “running a small commercial cloud”. At the time, Singapore police said that “investigations are ongoing into Megaspeed for suspected breaches of our domestic laws”. (Megaspeed says this investigation ended in 2026 with no action against it, but according to the New York Times, Singapore authorities have said they are still investigating the company.)
In late 2025, the Commerce Department pauses its investigation into Megaspeed. A few months earlier, NVIDIA’s general counsel had emailed agents working on the case “seeking information about the inquiry” and said their boss (Kessler) had given him permission to reach out. The agents declined to meet with him. Commerce officials then disclosed the episode to the Justice Department, which asked Commerce to pause its work, reportedly to avoid compromising the case. (The Justice Department’s own criminal investigation reportedly continues. According to Bloomberg, Commerce “has not made any progress since then, despite the agency having enough evidence to move forward”.)
In late 2025, at the US government’s request, NVIDIA audits Megaspeed and fails to account for all the chips. NVIDIA promised in December 2025 to carry out follow-up visits, but has since declined to say whether it has carried them out. NVIDIA said at the time that it had asked the US government whether it needed to cut off Megaspeed and was told no. (NVIDIA now says it had already blocked sales to Megaspeed in June 2025.) In November 2025, a month after the Commerce investigation became public, Megaspeed imported at least 17,544 more of NVIDIA’s most advanced chips, worth about $787 million, according to trade records analyzed by Bloomberg.
In March 2026, the US government charges Super Micro co-founder Wally Liaw and two others in the largest AI chip smuggling case to date. Prosecutors alleged that the three had sold a middleman company in Southeast Asia at least about $2.5 billion worth of AI servers since 2024, most of which were smuggled to China. (Liaw has pleaded not guilty. One of the other two, Ting-Wei “Willy” Sun, reportedly pleaded guilty in October 2026.) Bloomberg later reported that some of the servers allegedly went to Alibaba (which Alibaba denies).
In spring 2026, Secretary of State Marco Rubio oversees the writing of a classified memo alleging that Jensen Huang is intimately involved with Alice Huang and that she is a Chinese spy. According to later reports, US intelligence had concluded that China’s Ministry of State Security had recruited her.
On October 1, 2026, Bloomberg reports that US officials believe China has smuggled in the equivalent of at least 115,000 restricted NVIDIA chips. (This should be viewed as a lower bound; the trade data discussed above implies several times more.) NVIDIA acknowledged that smuggling happens but, as mentioned above, called it “a drop in the bucket”.
On October 29, 2026, the New York Times reports that US intelligence agencies have concluded that Alice Huang is a Chinese intelligence asset who developed a close relationship with Jensen Huang. Jensen Huang said in response that his interactions with Alice Huang “have been related to Megaspeed, and it was all completely normal and not unusual”. An NVIDIA spokesperson likewise said that the interactions “are not unusual or unexpected” and that NVIDIA is “confident that if the US government believes that any person is a threat to American industry, it would let us know”.
On November 5, 2026, the Republican chairmen of three House committees ask Jensen Huang to appear at a closed-door briefing.14 The next day, the New York Times reported that Alice Huang had been recruited by the Ministry of State Security “to learn more about American AI chips and potentially influence the company to push for its most advanced chips to be made available to China”. In response, NVIDIA said that its customers do not ask it to press the US government to change its policies on chip sales to China.
On November 7, 2026, the New York Times reports that the Commerce Department has granted H200 licenses to apparent Chinese shell companies. At least two of the licensed firms had no physical presence at their listed addresses and appeared to be linked to companies on the Entity List, and at least two others had ties to Chinese military suppliers. Commerce appears to have approved “roughly two dozen Chinese firms to buy or handle NVIDIA’s H200 chips”.
On November 8, 2026, Jensen Huang acknowledges the relationship with the alleged Chinese spy, and NVIDIA’s board opens an investigation. In a message to employees, he wrote: “I was unfaithful to my wife. My actions hurt her and our children. There is no excuse for what I did. I am deeply sorry. … I love our country and our company. I am committed to protecting NVIDIA and supporting America’s national security. I am proud and grateful to be an American.” The board said it had “no reason to believe that either national security or NVIDIA secrets were compromised”. NVIDIA said that outside counsel will conduct the investigation, and that Jensen Huang will continue to lead the company in the meantime."
Read the full piece by my colleague Erich at The Substrate!
1
4
541
"People often ask me how big a deal AI chip smuggling is and whether it means the export controls aren’t working. What I usually tell them is that the export controls are working, but that they’re not as effective as they could be. According to IFP, the top five US AI companies have on average about 16 times as much compute as the top five Chinese ones, and China produces only around 1-2% as much compute domestically each year as the US and its allies do.
So thanks to the export controls, the US has far more compute than China does. But because of smuggling and remote access, Chinese companies get far more compute than they otherwise would—something like 80% more, if IFP’s estimates are correct."
Export controls have had a really large counterfactual effect, but they could do even more with some basic improvements to their enforcement and scope. Stricter enforcement against smuggling, taking action on Chinese remote access, and dealing with HBM smuggling via a basic whitelist approach, could all have an outsized effect in extending the already commanding US lead on compute.
Erich tells you everything you want to know about the Jensen saga, what the data actually tells us about smuggling and remote access, and what lies ahead for Nvidia, Megaspeed and export controls.
1
92











