BIS remote-access chip rule takes effect today, with a 90-day transition
The BIS remote-access AI chip rule takes effect today on its Federal Register publication, with a licence required when advanced chips serve data centers giving China-linked firms remote access to more than 12 million TPP of compute.

The News
The Bureau of Industry and Security's interim final rule, Addressing Remote Access to Advanced Semiconductors, takes effect today, Thursday, November 19, on its Federal Register publication — not 90 days later, as some early coverage suggested. The rule has two controls. First, a worldwide end-use control requiring a licence when advanced chips — ECCNs 3A090 and 4A090 and related categories — are destined for a data center giving companies headquartered in China, Macau or other Country Group D:5 destinations remote access to more than 12 million aggregate TPP of compute. Second, destination-based licence requirements for advanced-chip shipments to or within Indonesia, Malaysia, Singapore, Thailand and Brazil.
Data centers already serving covered customers have 90 days from today to obtain a licence. US-headquartered recipients can qualify for a reporting-based exclusion from the five-country licence. The same rule formally rescinds the Biden-era AI Diffusion Rule, and public comments are due 60 days after publication, on January 18, 2027.
On Nvidia's earnings call, interim chief executive Colette Kress said the company is no longer shipping licensed Hopper data center products to China. The demand now at risk is remote-access demand, not shipments: the rule targets the workaround of renting computing power in a data center abroad instead of buying the chips.
The rule closes the rental workaround: the question is no longer whether chips ship to China, but who gets to use them remotely.
Why It Matters
The effective date matters because it is a legal event, not an announcement. The rule was released on November 16 and the government and company responses were covered on November 18, but today is the day it becomes law. From today, a data center that gives a China-linked company remote access to more than 12 million TPP of advanced-chip compute needs a licence. The reporting-based exclusion for US-headquartered recipients applies only to the five-country destination licence — it does not excuse the worldwide end-use control, which no licence exception overcomes, except as noted by BIS.
The rule matters for chipmakers because it targets a workaround. Export controls used to focus on shipping chips to China. But a customer that cannot buy a chip can still rent time on one in a data center in another country. That is the remote-access gap, and it is now closed on paper. Nvidia's Kress made the shift explicit on the earnings call: the company is no longer shipping licensed Hopper data center products to China, so the demand at risk is the rental kind, not the shipment kind.
The mixed verdict reflects two offsetting facts. On the bearish side, the rule is a real constraint on demand growth in Southeast Asia, where data centers in Singapore, Malaysia and Thailand are the ones caught by the destination licences — and TSMC builds the chips, Nvidia and AMD design them, so all three feel it. On the other side, the 90-day transition, the reporting-based exclusion for US firms, and the fact that the market has been pricing export risk since the resignation week all soften the immediate blow. The rule is a headwind with a long fuse, not a shock.
Analyst's Take
Our call is Mixed. The rule is a genuine tightening — it closes the remote-access workaround that let China-linked firms rent computing power abroad instead of buying chips — but it lands on a market that has already priced in much of it. Nvidia has said it is no longer shipping licensed Hopper products to China, and the 90-day transition gives existing data centers time to apply for licences rather than shutting off overnight. The near-term cost is paperwork and delay, not a sudden revenue hole.
The longer-term effect depends on how the licences are granted. If the 12 million TPP threshold is enforced strictly and licences are slow, the rule caps a real source of demand growth for Nvidia, AMD and TSMC — the cloud and data-center operators that serve China-linked customers from Southeast Asia. If licences flow, the rule becomes a compliance cost and a paper trail. The rescission of the AI Diffusion Rule also removes one layer of uncertainty, even as it adds another.
What to watch next is the comment period and the first licence decisions. Comments are due January 18, 2027, and the industry will push hard on the threshold and the five-country list. The first test is whether data centers in Singapore, Malaysia and Thailand can keep serving covered customers under licence — or whether they start turning business away. Reader poll: Bullish or bearish on Nvidia under the new rule? Vote: Bullish, Bearish, Mixed.
Are you bullish or bearish on this story?
By The Numbers
FAQ
What is the BIS?
The BIS, or Bureau of Industry and Security, is the part of the US Commerce Department that controls exports of sensitive technology. It writes the rules on which advanced chips can be sold, shipped or accessed abroad.
What is TPP?
TPP, or total processing performance, is a measure of how much computing power a chip can deliver. The rule uses 12 million aggregate TPP as the line: above it, a data center giving remote access to a China-linked company needs a licence.
What is an interim final rule?
An interim final rule is a regulation that takes effect immediately when published, while the government still accepts public comments. The BIS rule takes effect today, November 19, and comments are due 60 days after publication, on January 18, 2027.
What was the AI Diffusion Rule?
The AI Diffusion Rule was the previous Biden-era framework for controlling AI chip exports, built around country tiers. The new BIS rule formally rescinds it and replaces it with the remote-access and destination-based controls.
👁
Source: Kurums