MANILA, Philippines — Nov. 14, 2026. The US agency responsible for licensing advanced artificial-intelligence chip exports is facing a difficult policy test as Nvidia’s own financial disclosures show how restrictions can disrupt sales without guaranteeing access to the Chinese market. The Bureau of Industry and Security (BIS) changed its approach to selected chip exports in January, but Nvidia’s August filing shows that a licence alone does not secure a customer or eliminate commercial losses.
That makes the central issue broader than a dispute involving one company. Washington must administer national-security controls while checking buyers and supply chains, and manufacturers must decide whether a sale can comply with both US rules and restrictions imposed by destination markets.
Nvidia Chip Export Controls: What BIS Actually Changed
On Jan. 13, 2026, BIS announced that applications to export Nvidia H200 chips, AMD MI325X chips and comparable products to China would be considered case by case, subject to stated safeguards. The change was not a blanket authorisation for every customer, chip or transaction.
BIS specified conditions including screening procedures at the Chinese purchaser, independent third-party testing in the United States and evidence that a proposed shipment would not reduce global semiconductor capacity available to US customers.
Those requirements matter because licensing is a decision about particular products and transactions. A publicly announced policy of reviewing applications does not itself approve a shipment, identify its final user or establish that a purchaser has passed due diligence.
The agency’s published description of its Office of National Security Controls says engineers and analysts evaluate licensing requests, technical classifications and national-security questions. That work requires technical judgement as well as legal compliance, even before possible enforcement questions arise.
Nvidia’s Filings Show the Cost of Restricted Access
Nvidia’s second-quarter fiscal 2027 results reported US$96.2 billion in revenue, including US$89.0 billion from Data Center, for the quarter ended July 26, 2026. The company also excluded China Data Center compute revenue from its forecast for the following quarter.
That is an important distinction: strong worldwide sales do not mean every regional market remains accessible. The company’s regulatory filing reported a US$4.5 billion charge in fiscal 2026 connected to H20 inventory and purchase commitments after licensing requirements affected demand.
The same filing described a further US$0.4 billion charge in the first half of fiscal 2027 linked to H200 inventory and purchase obligations. Nvidia said approved H200 shipments represented less than 1% of Data Center revenue in its most recent reported quarter.
These are company-reported financial effects, not a calculated total loss to the US economy or a measure of all chips reaching China. They also do not establish that a particular customer or Nvidia itself violated export-control law.
Why Approval and Delivery Are Two Different Questions
The company’s filing says US authorities issued licences covering limited H200 shipments to specified China-based customers beginning in February 2026, but restrictions in China limited what could actually be sold.
This is why a change in the US licensing policy cannot automatically be reported as a reopening of the Chinese AI market. The exporter must meet US requirements; it also needs a willing customer, a viable commercial transaction and permission to complete the transaction at the destination.
Nvidia described a US inspection process for H200 products and a related 25% tariff arising when they were brought into the United States for that process. Its filing stated that it had not been able to pass that cost on to customers.
For the company’s shareholders, that introduces a mixture of policy, pricing and inventory risks. For the public, it shows why headlines about either a ban or an approval may describe only one stage of a longer commercial process.
WWC has previously reported on a Chinese university’s procurement of Nvidia H200 processors. The existence of a procurement record, however, must be distinguished from verified evidence of a specific unlawful export route.
Export Enforcement Extends Beyond a Shipment’s Address
BIS’s May 31, 2026 guidance says certain advanced-computing licence requirements apply to entities headquartered in restricted jurisdictions, or owned by a parent headquartered there, even if the immediate recipient operates elsewhere.
That is important for multinational data centres and companies buying through intermediaries. A recipient’s street address may not answer the legal question of who ultimately controls or uses the technology.
BIS also published diversion-prevention guidance identifying indicators that warrant closer examination in transactions involving advanced computing chips. Such guidance assists compliance decisions; it does not mean every reseller, data-centre operator or cloud customer is suspected of wrongdoing.
WWC’s earlier coverage of questions surrounding an Nvidia-related smuggling inquiry illustrates why allegations about an investigation require a separate evidentiary standard from the existence of export-control rules.
The public government records examined for this article do not establish that agency officials improperly halted a specific investigation, that particular Nvidia chips were diverted illegally, or that any named official committed misconduct. Those claims would require original investigative records, formal findings or attributable government statements.
Why Agency Oversight Matters to the AI Supply Chain
Analysis: A well-functioning licensing system must be clear enough for legitimate exporters to follow and rigorous enough to identify high-risk transactions. Both objectives become harder when semiconductor capabilities, intermediaries and AI computing services evolve quickly.
A licence review is not the same as a criminal investigation. The former tests a proposed export against applicable regulations; the latter requires evidence supporting a potential violation. Combining the two can produce misleading claims about what regulators have actually decided.
The agency’s public regulations also protect the confidentiality of certain licensing information. That limits what outsiders can independently verify about individual applications. Transparency about general policy, enforcement procedures and published decisions is therefore especially important.
WWC has separately covered questions involving Nvidia’s leadership and China-related allegations. Those contested claims should not be substituted for a formal BIS enforcement conclusion, nor should they determine how ordinary export applicants are assessed.
For Philippine companies using US-designed computing equipment, the practical lesson is to verify applicable export classifications, counterparties and end-use restrictions rather than assume a product purchased through a third country is automatically unrestricted.
What Happens Next
The measurable next steps are official BIS rule changes, published enforcement outcomes and Nvidia’s subsequent disclosures on licensed shipments, costs and China-related revenue. Any congressional finding or agency investigation into internal conduct should be reported only when its original record can be examined.
The evidence available now establishes a changing export policy and material commercial consequences. It does not independently establish the reported November crisis inside BIS or wrongdoing by agency officials.
Frequently Asked Questions
What does the Bureau of Industry and Security do?
BIS is part of the US Commerce Department and administers export controls, including licence reviews for specified advanced computing technologies.
Are Nvidia H200 exports to China completely banned?
Not under the January 2026 policy described by BIS. Certain applications receive case-by-case review subject to detailed requirements; individual transactions still need the required authorisations.
Did Nvidia record losses related to chip restrictions?
Nvidia disclosed a US$4.5 billion H20-related charge in fiscal 2026 and a US$0.4 billion H200-related charge in the first half of fiscal 2027.
Has an official investigation proved wrongdoing at BIS?
No such finding was established in the primary records inspected for this article. Reported accusations should not be treated as adjudicated facts.
