• Mon - Fri 9.00 - 19.00

Tag Archives: export controls

Chinese Firms Found a Way Around the Chip Ban. Washington Wants It Closed

The Nvidia chip export loophole lets Chinese AI firms access advanced computing power that direct export rules were meant to block. They do it by renting time on Southeast Asian data centers built with Nvidia’s top-tier chips. U.S. Commerce Department officials are now drafting a rule to close it. A draft is expected to circulate among trade groups as early as this month.

How the Nvidia chip export loophole works

Washington bars Nvidia’s most powerful chips, including the H100 and A100 series, from direct sale to China under a presumption-of-denial policy. A Chinese firm blocked from buying those chips outright has another option, though. It can instead rent remote cloud access to servers running on them, as long as those servers sit in a country the direct export ban does not cover.

Nvidia chip export loophole
Chinese Firms Found a Way Around the Chip Ban. Washington Wa

Data centers in Thailand and Singapore have become focal points for this workaround, according to CNBC. Chinese AI firms lease computing time on Nvidia-powered servers there, sidestepping the restriction that applies to physical chip shipments into China itself.

The legislative gap Washington is racing to close

A bill called the Remote Access Security Act, known as RASA, would extend U.S. export control authority to cover this kind of remote cloud access. The House passed it by a lopsided 369-to-22 vote in January, but the Senate has not yet taken it up, according to reporting reviewed by Asia Times.

Export-control lawyers say the Bureau of Industry and Security currently lacks clear statutory authority to police remote chip access until RASA becomes law. That legal gap is what has allowed the Southeast Asia workaround, sometimes referred to as the Aivres dispute, to persist.

Commerce Department’s next move

In the meantime, Commerce is drafting its own rule targeting data centers specifically in Thailand and Singapore, aiming to restrict Chinese firms’ ability to rent Nvidia-powered compute there even without new legislation. The rule could be shared with industry trade groups as soon as September, ahead of any Senate action on RASA.

The timing places the issue directly in the path of the upcoming Trump-Xi summit in Washington later this month, where the two leaders are expected to discuss whether the current arrangement continues.

Where the loophole debate heads next

Watch for two separate tracks: whether the Senate moves on RASA, and whether Commerce’s draft rule advances on its own regulatory timeline. The two could arrive in either order, and industry groups are expected to push back on any rule that raises compliance costs for cloud providers operating legally outside China.

Nvidia itself has largely stayed out of the political fight publicly, while continuing to sell chips into markets where sales remain legal under current rules.

Why enforcement has been so difficult

Export control law was built around physical shipments. A chip that never crosses a border, but is instead accessed remotely over the internet, does not fit neatly into rules written for cargo manifests and customs declarations. That gap is exactly what the Southeast Asia workaround exploits.

Data center operators in Thailand and Singapore are not necessarily breaking any law in their own jurisdictions. They lease server time to whoever pays for it, much like cloud providers everywhere. The legal question is whether the end user’s location and nationality should trigger U.S. export restrictions on that access.

Congress created the current gap somewhat inadvertently. Commerce rescinded a Biden-era know-your-customer requirement for cloud access, and export lawyers say that decision opened the door the current rule is now trying to close. Untangling that history has slowed the drafting process.

Enforcement also runs into a practical limit. Even if the U.S. tightens rules on Thailand and Singapore specifically, similar data centers could emerge in other countries not yet covered. Some analysts expect this to become a recurring pattern rather than a one-time fix.

Frequently asked questions about the chip loophole

What is the Nvidia chip export loophole?
It refers to Chinese AI firms renting remote access to Nvidia-powered servers in countries like Thailand and Singapore, avoiding the direct export ban on shipping chips into China.

What is the Remote Access Security Act?
RASA is a bill that would extend U.S. export control authority to cover remote cloud-based access to restricted chips. It passed the House 369-22 in January but awaits Senate action.

Can the Commerce Department act without RASA passing?
Commerce is drafting its own rule targeting Southeast Asian data centers, though export-control lawyers say broader enforcement authority depends on RASA becoming law.

Which countries are involved in the loophole?
Thailand and Singapore have been identified as locations where data centers provide Chinese firms with access to Nvidia-powered computing.

Will this come up at the Trump-Xi summit?
Yes. The two leaders are expected to discuss whether the current cloud-access arrangement continues when they meet in Washington on September 24.

For related coverage, see our reporting on the global memory chip shortage and the Anthropic-Nvidia compute deal.

Sources

Washington Is Quietly Rewriting the Rules for AI Chips — Here’s What’s Coming

The Commerce Department has confirmed new Commerce Department chip regulation is coming for artificial-intelligence hardware. Officials are staying deliberately vague on timing and targets. The signal follows a rule that already took effect January 15, 2026. That rule lets the Bureau of Industry and Security review license applications for advanced AI chips bound for China and Macau. Officials handle each application case by case. Exporters must first meet a set of supply, security and testing conditions. A further rule is expected. No formal date has been set.

What the Commerce Department has actually confirmed

A Commerce Department official confirmed that regulatory action targeting chips and artificial intelligence is coming. The confirmation stayed thin on specifics by design. Officials gave no formal rulemaking date. They also gave no defined target list.

Semiconductor wafer technology tied to Commerce Department chip regulation

How the current Commerce Department chip regulation already works

The Bureau of Industry and Security’s rule took effect January 15, 2026. It changed the export-license review policy for certain advanced computing semiconductors destined for China and Macau. Officials now evaluate license applications for specific AI chips case by case. Exporters must meet a series of supply, security and testing conditions first.

Why officials are staying vague on specifics

Commerce Department officials confirmed harsh new AI export rules are coming. At the same time, they explicitly denied reports that the administration plans to revive the Biden-era AI Diffusion Rule. That denial came from officials responding to reporting on the confirmation. Officials describe the coming approach as a new, separate framework for strategic AI accelerator export controls. They say it is not a return to the earlier rule.

Who this affects first

Chipmakers with significant AI accelerator sales bound for China face the most direct exposure. That includes firms like Nvidia and AMD, both exposed to any tightening of the current case-by-case review process. Cloud providers and enterprise buyers could also feel indirect effects. Many of them depend on those chips for AI infrastructure outside the US, so longer licensing timelines or stricter approval conditions would touch them too.

Why chip export policy keeps shifting

US policy on advanced AI chip exports has moved through several iterations in recent years. It reflects an ongoing balancing act. National-security concerns sit on one side. The commercial interests of American chipmakers sit on the other, since these firms rely heavily on international sales, including to China. Each round of rulemaking draws public comment from industry groups. They warn about losing market share to non-US competitors. National-security voices push back the other way, arguing for tighter controls. That tension will likely shape whatever the Commerce Department proposes next. Previous rounds of chip export policy have also triggered retaliatory measures and workaround strategies from affected buyers. That history makes each new rule as much a diplomatic and commercial calculation as a technical one.

What happens once a formal rule appears

Export-control rulemaking like this typically moves through a formal proposal and comment process before taking effect. Industry pushback tends to follow once specifics get published. No timeline has been set publicly for that next step. Still, the January 2026 rule offers a template for how any new case-by-case licensing framework might look.

What companies can do while they wait

Chipmakers and their customers cannot yet plan around specific new terms, since none have been published. Legal and trade-compliance teams typically use this waiting period to review how the current January 2026 licensing framework has applied to their own shipments. That groundwork often makes it easier to respond quickly once a formal proposal appears and a public comment window opens. Companies with large China-linked AI hardware businesses have historically been among the most active participants in these comment periods, given how directly the rules affect their revenue.

How markets have reacted to past chip-policy signals

Chip stocks have moved on export-policy headlines before, even without a finalized rule in hand. A confirmed signal from the Commerce Department, absent specifics, tends to produce a milder market reaction than an actual published rule. Traders generally wait for the formal proposal before repricing individual stocks, since the eventual details, not the advance warning, determine which companies gain or lose the most.

AI chip export rules: quick answers

What is the Bureau of Industry and Security?

It is the arm of the US Commerce Department responsible for administering export controls, including on advanced semiconductors.

Did the US bring back the AI Diffusion Rule?

No. Commerce Department officials denied that. They describe the coming approach as new and separate from the Biden-era rule.

Which chips are affected right now?

Advanced AI accelerators bound for China and Macau. Officials currently review them case by case under the rule that took effect January 15, 2026.

When will the new Commerce Department chip regulation take effect?

No date has been set publicly as of this report.

Does this affect Nvidia specifically?

Yes, indirectly. Nvidia sells substantial AI hardware bound for China, which puts it directly in the path of any further tightening.

Where can I read the current export rule in detail?

Legal analyses of the January 2026 rule break down the current case-by-case review conditions. Firms that track export-control policy publish these regularly.

Related coverage on Tamara News

Nvidia H200 Chips Head Back to China Under Revised Export Rules

Nvidia has secured a path to keep selling advanced AI processors to China, reopening a market that had been effectively closed under earlier export restrictions. Reports this month confirmed Nvidia H200 chips China shipments are moving forward, with Chinese technology firms including ByteDance and Tencent reportedly receiving around 10,000 H200 units each in recent weeks under the revised US export framework.

How Nvidia H200 chips China sales became possible

Nvidia H200 chips China

The shift follows a January 2026 rule from the US Commerce Department’s Bureau of Industry and Security that moved export license reviews for certain advanced semiconductors from a presumption of denial to case-by-case approval, provided chips meet specific supply, security and testing conditions, according to policy analysis from the Council on Foreign Relations. That policy reversal opened the door for Nvidia to resume large-scale H200 shipments after months of restricted access.

Separately, reports emerged in August that Nvidia was preparing a China-specific chip built on licensed technology, though the company disputed the report, with a spokesperson saying it has “no China-specific LPU product in our roadmap” at this time, according to GuruFocus.

Why the export policy keeps shifting

Washington has swung between tightening and loosening AI chip export rules for more than a year, reflecting a tension between national security concerns and the commercial interests of US chipmakers who don’t want to cede the Chinese market to domestic rivals. A separate rule tightening tranche was drafted and then withdrawn by Commerce in March 2026 before it took effect, underscoring how unsettled the policy remains even as chips actually begin moving.

What it means for the broader AI supply chain

The renewed China access matters beyond Nvidia’s balance sheet. AI chip demand has been a major driver of the debt-financed data center buildout we detailed in our coverage of AI infrastructure debt and chipmaker bond markets, and any change in where chips can legally ship affects both revenue projections and regulatory risk for the sector. It also intersects with broader efforts to regulate AI systems, including the transparency rules that took effect under the EU AI Act this month.

Where Nvidia’s China access goes from here

Congress and China hawks within the administration are expected to keep pushing back against the case-by-case approval framework, meaning the policy could tighten again with little notice. Nvidia, for its part, has said it will continue complying with whatever rules are in force rather than banking on any particular policy outcome. Investors are watching third-quarter earnings guidance for signs of how much revenue the resumed China shipments are actually contributing.

Rival chipmakers are watching closely too, since any framework that works for Nvidia could set a template other companies try to use for their own China-bound products. Analysts caution that the back-and-forth nature of the policy this year, tightening in some months and loosening in others, makes it difficult for chipmakers to plan production and inventory with confidence, and some firms have said they are deliberately keeping China-bound supply chains flexible enough to redirect volume elsewhere if the rules shift again. Semiconductor equipment makers, memory suppliers and cloud providers that depend on predictable chip allocation are all factoring this policy volatility into their own planning, with several citing export-rule uncertainty as a specific risk factor in recent investor disclosures.

Nvidia chip exports: frequently asked questions

What changed to allow Nvidia H200 chips China shipments?
A January 2026 Commerce Department rule shifted export license reviews for qualifying advanced chips from presumed denial to case-by-case approval.

Which Chinese companies are reportedly receiving H200 chips?
Reports indicate ByteDance and Tencent have each received approximately 10,000 H200 units in recent weeks.

Is Nvidia building a China-specific AI chip?
Reports suggested this, but Nvidia has publicly denied having a China-specific LPU product on its roadmap.

Could the export rules tighten again?
Yes. Commerce has already drafted and withdrawn one tightening rule in 2026, and policy pressure from China hawks in Congress continues.

Why does this matter for the AI industry beyond Nvidia?
Chip export access affects revenue forecasts across the AI supply chain, including data center financing tied to chipmaker performance.

Are these chips subject to any conditions?
Yes, approvals require chips to meet specific supply, security and testing conditions set by the Bureau of Industry and Security.