OTWopentechwire
Tech Intelligence, Openly Wired
Policy

Washington Extends Chip Controls to Data Centres Serving Chinese Clients

New licensing regime targets remote compute access, adding Malaysia, Singapore, Indonesia, Thailand and Brazil to scrutiny list as administration seeks to close loopholes in semiconductor export framework.

KW
Kenji Watanabe
Hardware & Products Reporter · Tokyo
Nov 18, 2026
7 min read
Washington Extends Chip Controls to Data Centres Serving Chinese Clients
Credit: US Commerce Department / Wikimedia Commons

Licensing Requirement Targets Compute Infrastructure Globally

The Trump administration rolled out an interim rule on Monday that extends export controls beyond the sale of AI semiconductors to encompass data centres and cloud infrastructure that might provide computing resources to entities in embargoed nations. The Commerce Department will now require licences for the export, reexport, or transfer of AI chips and servers destined for data centres anywhere in the world if those facilities may supply compute capacity to companies in China, Russia, Iran, North Korea, or 18 other countries grouped under the D:5 embargo designation.

The measure represents a departure from earlier controls that focused on the point of sale. Under existing rules, chip manufacturers needed a licence to ship advanced semiconductors directly to businesses based in or owned by entities in the D:5 group. The new framework acknowledges that companies subject to those restrictions have been circumventing them by purchasing compute capacity remotely from data centres located in jurisdictions not covered by the original controls.

The interim rule describes the change as a logical extension of existing policy. It states that the restrictions will undermine the ability of companies headquartered in countries of concern to use advanced Super Intelligence semiconductors for training sophisticated AI models. The administration's concern centres on the risk that access to high-performance computing infrastructure could support military or intelligence activities that threaten US national security or foreign policy interests.

Threshold Set at 12 Million Total Processing Performance

The rule applies to data centres or networks capable of supplying more than 12 million total processing performance, an export-control metric the US government uses to quantify computing power. That threshold roughly corresponds to the maximum compute capacity of approximately 750 Nvidia H100 or H200 chips.

The scope of the rule extends to indirect access arranged through intermediaries or brokers. Exporters and data centre operators are now obligated to monitor the ownership structures and activities of their customers to ensure they are not controlled by entities from the D:5 group. This due-diligence requirement places a compliance burden on infrastructure providers that have not traditionally been subject to semiconductor export controls.

Data centre operators already providing services that fall under the new rule must obtain licences within 90 days of the rule taking effect. The deadline creates a tight window for operators in jurisdictions such as Singapore, Malaysia, and other regional hubs that have become popular locations for cloud infrastructure serving Asian markets.

Five Southeast Asian and Latin American Markets Added to Scrutiny

The Commerce Department has introduced new licensing requirements for chip exports, reexports, or transfers of advanced semiconductors to Malaysia, Singapore, Indonesia, Thailand, and Brazil. The department stated that it included these countries because they present a higher-than-normal risk of diversion, meaning chips shipped to those markets could be rerouted to end users in embargoed nations.

For these five countries, however, the licence requirements will not apply to chips destined for companies based in the United States. That carve-out suggests the administration is attempting to balance security concerns with the operational needs of US technology firms that have established data centre footprints in those markets.

The inclusion of Singapore and Malaysia is particularly significant for the regional cloud infrastructure landscape. Both countries have attracted substantial investment from hyperscale cloud providers and colocation operators over the past five years, driven by demand from enterprises across Southeast Asia. Singapore hosts data centres operated by Amazon Web Services, Microsoft Azure, Google Cloud, and a range of regional players. Malaysia has emerged as an alternative hub, offering lower power costs and land availability that Singapore lacks.

The new licensing requirements introduce uncertainty for operators in those markets. A data centre in Singapore serving customers across Asia will now need to demonstrate that none of its clients are controlled by entities in the D:5 group, a task complicated by the opacity of corporate ownership structures in some jurisdictions. The rule does not specify the level of due diligence required, leaving operators to interpret the standard themselves or wait for further guidance from the Commerce Department.

Context: Megaspeed Scandal and Offshore Compute Rush

The timing of the rule follows closely on the resignation of Nvidia co-founder Jensen Huang amid a scandal involving an alleged Chinese spy linked to one of the chipmaker's customers. The US government has reportedly sought to place that customer, Megaspeed, on a special entity list, believing the company has been purchasing thousands of Nvidia chips to provide compute capacity to Chinese businesses in order to bypass existing export-control restrictions.

Last week, reports indicated that Chinese AI companies are rushing to secure computing capacity overseas amid fears of tighter restrictions from the United States. The pattern described in those reports aligns with the administration's rationale for the new rule: that companies subject to chip export controls have been acquiring compute access indirectly, rendering the original controls less effective.

The interim rule rescinds the Biden administration's AI Diffusion Rule from January 2025, which sought to impose country-level limits on exports of advanced AI chips and restrict transfers of certain powerful AI models. In May, the Trump administration announced it would discard that framework, arguing that it imposed unnecessary restrictions on US companies and the country's allies. The new rule narrows the focus to compute access that could be used for training or running advanced AI models, rather than applying broad country-level caps.

Implications for Regional Cloud Infrastructure

The rule's impact will be felt most acutely by data centre operators in markets that serve a diverse customer base spanning multiple jurisdictions. Operators in Singapore, for example, often host workloads for clients in China, India, Australia, and Southeast Asian countries simultaneously. The new licensing requirements will force those operators to segment their customer base and potentially refuse service to clients they cannot verify as compliant.

The 90-day compliance window is short relative to the complexity of the task. Operators must review existing customer contracts, assess ownership structures, and apply for licences where necessary. For facilities that have already installed thousands of AI chips, the prospect of having to halt service to certain customers or face penalties creates operational and financial risk.

The rule does not apply to commercial recommendation systems, such as those used in e-commerce or social media. That exclusion suggests the administration is targeting compute capacity used for training large-scale AI models, rather than inference workloads that power consumer applications. The distinction may provide some relief to operators serving e-commerce platforms and social media companies, though the boundary between training and inference is not always clear-cut.

Broader Shift in Export Control Philosophy

The interim rule reflects a broader shift in US export control philosophy, moving from a product-centric approach to one that considers the entire supply chain and access pathways. Earlier controls focused on preventing the physical shipment of chips to embargoed countries. The new framework acknowledges that in an era of cloud computing and remote access, physical location of hardware is less important than who can use it.

This shift has precedent in other areas of export control, such as software and encryption technology, where the focus has long been on access rather than physical transfer. Applying the same logic to semiconductors and data centre infrastructure represents an expansion of that principle into the hardware domain.

The rule's effectiveness will depend on enforcement. The Commerce Department has limited resources to monitor data centre operations worldwide, and the complexity of corporate ownership structures in many jurisdictions makes verification difficult. The rule places the burden of compliance on exporters and operators, but it remains to be seen how rigorously the department will audit those entities or penalise non-compliance.

Forward-Looking Considerations

The interim rule is expected to be followed by a period of public comment and potential revision. Industry groups representing chip manufacturers and data centre operators are likely to raise concerns about the compliance burden and the risk of unintended consequences, such as driving cloud infrastructure investment away from US-allied markets.

The rule's inclusion of Malaysia, Singapore, Indonesia, Thailand, and Brazil signals that the administration views those markets as potential diversion points. Whether that assessment is accurate will depend on how effectively those countries enforce their own export control and anti-diversion measures. The rule does not impose sanctions on those countries, but it does place them under heightened scrutiny, which could affect their attractiveness as data centre hubs.

For Chinese AI companies, the rule closes one avenue for accessing advanced compute capacity but is unlikely to eliminate all workarounds. The history of export controls suggests that determined actors find alternative pathways, whether through shell companies, third-party brokers, or other means. The administration's challenge will be to stay ahead of those adaptations without imposing costs on legitimate commerce that outweigh the security benefits.

Read next
Policy

China's Autonomous Driving Sector Builds Resilience Against Chip Supply Shocks

Wei Zhang · 7 min
Policy

Washington Moves to Restrict AI Chip Sales Amid Nvidia Leadership Crisis

Sofia M. Reyes · 5 min
Policy

Helen Toner on OpenAI, Board Accountability and the Limits of Trust in AI Safety

Daniel R. Whitfield · 5 min
Spot something wrong? Email corrections@opentechwire.com. We log every correction publicly.