Export controls have already cost NVIDIA a measurable amount of money. In the first quarter of fiscal 2026 the company recorded a $4.5 billion charge for H20 inventory and purchase obligations it could no longer fulfil, after being told on April 9, 2025 that shipping H20 chips to China would require a licence. The following quarter's outlook reflected roughly $8 billion of lost H20 revenue from the same restrictions (source: NVIDIA Q1 fiscal 2026 results).
That is the scale of what a single rule change can do to the largest company in the AI supply chain, and it is why export-control policy sits alongside earnings as a driver of NVDA's price. For traders holding NVDAX/USDT or the NVDAUSDT Perpetual on BloFin, the value of understanding these rules is knowing which headlines carry that kind of weight and which do not.
Why US-China export controls matter for NVIDIA right now
Export control regulations shape how major American companies operate in the Chinese market, and no company feels this more acutely than NVIDIA. Since October 2022, the Bureau of Industry and Security within the Commerce Department has repeatedly tightened the rules governing which AI chips and semiconductor manufacturing tools can be shipped to China and other countries of concern. Each round has targeted NVIDIA GPUs by name or by performance threshold.
The exposure has fallen sharply as a result. At its peak in fiscal 2023, China represented roughly 19% of NVIDIA's data center revenue (source: NVIDIA fiscal 2024 annual report). By fiscal 2024 that had fallen to about 14%, and by the fourth quarter of that year NVIDIA was describing China as a mid-single-digit percentage. Today the company excludes China data center compute revenue from its forward guidance altogether.
You are not trading the policy. But NVDA and NVDAX prices move when Washington changes the rules as part of broader US technology policy, when draft rules leak, or when a licence is granted or withheld, and that transmission is fast enough to matter intraday.
Timeline: Key US export control actions targeting NVIDIA chips
The sequence matters more than any single rule, because each round was written to close the workaround the previous one created.
October 2022: BIS restricted exports of NVIDIA's A100 and H100 GPUs to China, citing national security concerns around advanced AI compute. NVIDIA responded by designing modified chips, the A800 and H800, with reduced interconnect bandwidth to fall below the new control thresholds (source: Every CRS Report). The Biden administration's October 7, 2022 rule broadened restrictions to cover advanced chips, AI compute end uses, and semiconductor manufacturing equipment.
October 2023: BIS tightened performance thresholds again, introducing metrics like performance density (compute per square millimeter of die area). This captured the A800 and H800, which had been specifically designed to comply with the 2022 rules. NVIDIA began designing a new generation of compliant chips: H20, L20, and L2.
October 2024: The Commerce Department refined performance-per-area metrics, captured additional advanced chips, and increased monitoring of indirect shipments through various hubs.
April 2025. NVIDIA was informed that even H20 exports to China would require a licence, producing the $4.5 billion charge and the roughly $8 billion revenue hit described above.
February 2026. The US government granted NVIDIA a licence to ship H200 chips to China. The terms have remained unclear enough that NVIDIA continues to assume no China data center compute revenue in its outlook.
Each of these produced noticeable volatility in NVDA, and those moves flow into NVDAX. Because the token trades continuously while NASDAQ does not, a rule announced outside market hours reaches NVDAX first, which is covered in why tokenized NVDA gaps when NASDAQ is closed.
How dependent is NVIDIA on China for AI and data-center revenue?
NVIDIA does not publish a standalone "China data-center revenue" line item. Analysts and public disclosures tell a clear story of declining exposure:
At its peak in fiscal year 2023, China represented roughly 19% of NVIDIA's data-center revenue. By fiscal year 2024, that figure dropped to about 14%. In Q4 of fiscal 2024, the company described China as contributing a mid-single-digit percentage.
The "China" exposure includes direct sales to mainland cloud providers like Alibaba Cloud, Tencent Cloud, and Baidu, plus global system integrators and OEMs shipping servers into China.
The decline in Chinese revenue has been partially offset by explosive demand from US and allied-country hyperscalers (AWS, Microsoft Azure, Google Cloud) and AI startups building large-scale training clusters.
A sudden, harsher clampdown on chip exports to China would not eliminate NVIDIA's data-center business, but it could slow growth, change guidance, and shift expectations about NVIDIA's long-term market share in AI chips. For a deeper breakdown of what drives the stock, see what moves NVIDIA stock price.
What exactly is being controlled: Advanced AI Chips and high-performance NVIDIA GPUs
The US government does not ban "all chips." Export controls focus on AI chips and NVIDIA GPUs that exceed specific compute-density thresholds. BIS measures Total Processing Performance (TPP) in tera-operations per second and performance density in TPP per square millimeter of die area, thresholds designed to limit the computing power available to China through advanced hardware. Interconnect bandwidth for large data centers (NVLink speeds, for example) is another criterion.
Key NVIDIA product lines affected:
Controlled: A100, H100, A800, H800, L40, L40S, RTX 4090, and newer Hopper/Blackwell-class chips (H200, B100/B200) that sit well above thresholds.
China-compliant variants: H20, L20, and L2, designed with reduced core counts, interconnect speeds, and memory bandwidth to fall below current thresholds.
Companies must reconfigure product design to comply with performance limits and regulatory requirements, and NVIDIA has done exactly that with each rule change. Nvidia's H200 chip is six times more powerful than Huawei's best chip, illustrating the scale of performance regulators are trying to keep out of China.
Export controls so far have targeted data-center and AI training chips rather than mainstream gaming GPUs. But chips like the RTX 4090 were captured under the October 2023 rules because their performance and bandwidth exceeded thresholds, so the scope can always widen.
Enforcement risk and the reality of bypassing export controls
Export controls limit direct shipments of advanced chips to Chinese entities, but enforcement is not airtight. Some buyers attempt to obtain NVIDIA chips via intermediaries in Singapore, the UAE, or Southeast Asia. In 2024, American AI chips smuggled to China numbered in the tens of thousands, according to investigative reporting and government disclosures; that matters because large numbers of chips, not just top-end performance, change the risk profile (source: CSIS).
Such workarounds increase cost, delivery time, and compliance risk. US authorities have stepped up enforcement, investigating distribution paths running through Thailand, Malaysia, and other countries. Compliance burdens increase for companies navigating multiple regulatory environments, and large distributors face growing documentation, end-user declaration, and blacklist requirements.
These measures cannot fully eliminate gray-market flows, but they reduce large-scale shipments. Exporting Nvidia chips at scale could triple China's AI computing power next year, which helps explain why enforcement matters. The key risk for NVIDIA and its investors: if Washington believes current rules are being widely bypassed, it can tighten controls further or impose more onerous reporting requirements, which could disrupt sale forecasts and spook the stock. Tech companies also face risks of substantial inventory losses due to sudden sanctions and export bans, a scenario NVDA traders should account for, especially if tighter safeguards need to happen before Washington is convinced current enforcement is insufficient.
Strategic impact on NVIDIA's China position and global market share
Export controls both hurt and help NVIDIA. They restrict immediate revenue from China, a strategic competitor in AI infrastructure, but also make it harder for Chinese rivals to access the latest US manufacturing tools, EDA software, and HBM memory, slowing domestic rivals in advanced chips.
China's domestic production of AI chips is severely constrained. Huawei's AI chip production is estimated at 805,000 units this year, and Huawei's AI chip production is expected to fall next year. China will not produce a competitive AI chip until Q4 2027 at the earliest, according to industry analysts. Beijing's push for semiconductor self-sufficiency has real implications for international tech companies, but the timeline is long.
The gap between Chinese AI developers and the rest of the AI industry is still large. Last year, 100 out of 103 AI models built by Chinese developers used U.S. hardware. Figures like Liang Wenfeng, whose DeepSeek lab trained frontier models, relied on NVIDIA GPUs. China's AI development remains limited by access to NVIDIA chips, and export controls can encourage companies to develop domestic alternatives in response to restrictions, but those alternatives lag in performance, software ecosystem (CUDA libraries, developer community), and reliability.
NVIDIA's global market share in AI accelerators remained above 70% for data-center GPUs used in training large models through 2024. A long-run decoupling from China could gradually reshape this landscape. Export controls may push Chinese companies toward alternative AI stack ecosystems and indigenous software, fragmenting the global AI hardware landscape and creating new competitive risks that NVDA traders need to monitor. If China regained access to NVIDIA chips, that compute could also support an AI Belt and Road initiative that extends influence abroad. The stakes therefore extend beyond NVIDIA's direct market share to competition across the wider world.
The national security reasoning
The Commerce Department treats advanced AI chips as dual-use technology, on the straightforward logic that the hardware which trains a commercial language model also trains a military one. The concerns build on each other rather than standing separately.
The first is capability: accelerators at this performance level support autonomous systems, cryptanalysis and strategic planning tools, not only consumer products. The second follows from it, because once an adversary holds both the hardware and a capable base model, it can fine-tune systems outside any oversight the original developer imposed. The third is scale, since compute is the input that most directly determines how quickly a national AI programme can advance, which is why the rules are written against performance rather than application. And the fourth is leverage: a country with surplus AI infrastructure can supply it to others, converting compute into diplomatic influence.
Some analysts argue for an alternative to hardware bans, in which Chinese firms rent access to AI compute hosted in the United States or allied jurisdictions rather than importing the chips. That would keep the physical hardware under domestic control while permitting commercial access. Support for the approach has varied across administrations and it remains a proposal rather than policy.
How NVIDIA is managing the exposure
NVIDIA CEO Jensen Huang has described China as an important market while acknowledging that the company's market share for AI-training GPUs in China fell from roughly 95 percent to zero after export controls took effect (source: Tom's Hardware). Jensen Huang has stated repeatedly that NVIDIA will comply fully with all U.S. export controls while working to support Chinese customers within legal limits.
NVIDIA has invested in designing compliant versions of its advanced chips for China, adjusting performance to meet U.S. thresholds while still delivering usable AI compute to local cloud and internet companies. The company is also diversifying demand not just across regions but through partnerships with companies around the world, including non-Chinese hyperscalers, sovereign AI projects in Europe and the Middle East (including deals in Taipei and other allied capitals), and U.S. enterprises, reducing reliance on China over time.
Chinese regulators may retaliate against foreign companies through various means, including antitrust investigations, which adds another layer of risk for Intel and other foreign chipmakers. Investor perception of how effectively NVIDIA navigates this environment can move NVDA's valuation multiples and create short-term volatility.
What this means for trading NVDAX on BloFin
Export-control risk reaches a BloFin position through three channels, and they compound rather than operate separately.
The first is headline sensitivity. A new rule, a leaked draft or a license decision reprices NVDA quickly, and because NVDAX tracks the stock's economic value, that repricing arrives in the token at the same time. When the news lands outside NASDAQ hours it arrives in the token first.
The second is earnings language. NVIDIA's guidance is where policy becomes a number, and the phrasing around China demand, export controls and geopolitical risk in each quarterly report is what analysts rebuild their models on. The $8 billion revenue adjustment did not arrive as a headline; it arrived as a line in an outlook.
The third is leverage. If you hold the NVDAUSDT perpetual rather than spot, a policy-driven gap is amplified by whatever multiple you are running, and a position sized comfortably for an ordinary week can be uncomfortable across a rule change. Leverage and liquidation covers the arithmetic, and spot and perpetual contracts sets out what each wrapper does differently.
None of this is a signal on its own. Export-control scenarios are one input into a view on NVIDIA, and they work best combined with the demand picture rather than traded as isolated headlines. What each wrapper gives you is set out in tokenized NVDA versus real NVDA stock.
Frequently asked questions
Can export controls cut NVIDIA off from China entirely?
In practice they have come close for the highest-performance products. NVIDIA's share of AI training GPUs in China has fallen from roughly 95% to zero, and the company now excludes China data center compute revenue from its forward guidance. What remains is lower-performance silicon and software sold under existing rules, plus whatever the February 2026 H200 license permits. The rules have tightened in stages rather than as a single ban, and each revision has narrowed what is permissible rather than closing the market outright.
Do the controls affect gaming GPUs as well as data center chips?
They mainly target data center and AI training silicon, but the boundary is a number rather than a product category. The RTX 4090, a consumer gaming card, was captured by the October 2023 rules because its performance exceeded the threshold that applied. Any future product that crosses whatever line is current at the time will be caught the same way, which means no category is permanently exempt and the scope can widen without a new policy being announced.
Why does NVDA sometimes fall sharply on a single policy headline?
Because the valuation is built on expectations of sustained dominance in AI compute, and a policy shift changes the expectations rather than the current quarter. Analysts respond by trimming revenue models, raising the risk premium, or both. On a high forward multiple, small changes to growth assumptions produce large changes in price. The April 2025 H20 decision is the clearest example: a licensing requirement became a $4.5 billion charge and roughly $8 billion of forgone revenue within two quarters.
How much of NVIDIA's revenue is still at risk from China?
Less than most headlines imply, because most of it has already gone. China peaked at roughly 19% of data center revenue in fiscal 2023, fell to about 14% the following year, and is now excluded from guidance entirely. The residual risk is less about losing what remains and more about retaliation, about Chinese domestic alternatives maturing over several years, and about the precedent that rules can change between one quarter and the next.
Should I buy or sell NVDAX because of export controls?
BloFin does not provide personalized financial advice, and this article is not a recommendation. Export-control risk is one input among several, and it matters most as context for interpreting news rather than as a trigger on its own. Your position should reflect your own research, risk tolerance and time horizon, and if you are trading with leverage it should reflect the possibility of a gap arriving while you are not watching.
Researched and written by the BloFin Academy editorial team with AI-assisted drafting. Primary sources include NVIDIA's first-quarter fiscal 2026 results, NVIDIA's fiscal 2024 annual report, the Congressional Research Service, the Lawrence Berkeley National Laboratory export control office, and the Center for Strategic and International Studies. All facts independently verified against cited documentation current as of August 2026.
Nothing in this article constitutes financial, investment, or trading advice. Cryptocurrency and tokenized stocks such as NVDAX are volatile. Do your own research and only trade with capital you can afford to lose.