NVIDIA ( NVDA ) looks at re-entry of China AI chip as it changes export licenses in case-by-case review
March 20, 2026 — NVIDIA is poised to return to China’s AI accelerator market after the US Bureau of Industry and Security (BIS) turned some advanced computing applications into individual license reviews for China and Macau. The policy change does not guarantee the approval or removal of controls, but it paves the way for higher compliance audited deployments – an important development for a company where Data Center revenue reached $62.3 billion in Q4 FY2026, up 93% year-over-year.
Why China Matters: Data Center Scale and the Income Gap
NVIDIA’s Data Center segment is the company’s main growth engine. In Q4 FY2026 (quarter ending January 25, 2026), Data Center revenue was $62.3 billion, compared to $35.6 billion in Q4 FY2025. For the full fiscal year 2026, NVIDIA’s total revenue was $215.9 billion, up 65% from $130.5 billion in FY2025.
| Metric | Q4 FY2026 | Q4 FY2025 | Y/Y |
| Net income | $68.1B | $39.3B | + 73% |
| Data Center Revenue | $62.3B | $35.6B | + 75% |
| GAAP gross margin | 75.0% | 73.0% | +2.0 points |
| GAAP mixed EPS | $1.76 | $0.89 | +98% |
China has historically been a relatively small but dominant player in Data Center revenue. NVIDIA’s Q1 FY2027 outlook of $78.0 billion clearly notes that the company “does not consider the Data Center fees from China in its view” – underscoring how important China’s re-licensing could be if approvals emerge.
Policy change: case-by-case licensing, not open door
In January 2026, BIS revised its license review policy for advanced computer exports to China and Macau, moving from a presumption of denial to a case-by-case review of specific applications. This does not remove controls — it means that appropriate deployments can be evaluated individually based on end-use, end-user risk, and technical parameters.
The framework is built on the technical parameters set in the BIS regulations, in particular:
– October 2023 interim final rule – strengthened advanced computer controls
– Final rule for April 2024 – updated limits and consolidated areas
Whether a particular NVIDIA chip clears the license depends on its performance limits relative to the current BIS limits. NVIDIA’s Hopper-generation products (H100, H200) are the most discussed candidates, but availability in China is primarily a question of licensing, not a question of product availability.
Competitive strength: Huawei Ascend and other domestic alternatives
China’s AI chip market has not stagnated during NVIDIA’s limited access. Huawei’s Ascend ecosystem – including the Ascend 910B and 910C accelerators – has grown in both the hardware offering and the software ecosystem (MindSpore, CANN). With a heavy burden of consideration, domestic operators have shown a willingness to use “good enough” performance with predictable supply over peak capacity and uncertain availability.
Key competitive considerations:
– CUDA lock advantage: The NVIDIA software ecosystem remains a deep groove for NVIDIA. Migrating existing training pipelines away from CUDA has significant retraining costs.
– Hypothetical exchange risk: Low-intensity workloads are easily handled. Home chips are more competitive here than training on a large scale.
– Supply certainty premium: Chinese hyperscalers and businesses place a high value on supply predictability – the structural advantage of domestically sourced chips without raw performance.
Any revenue realization for NVIDIA China depends on both license approval and whether customers have locked in other chains within a limited period.
Implications for investors: the opposite of structural risk
Side note: Even lower licensing concessions could add meaningful revenue beyond NVIDIA’s $78B Q1 FY2027 base (assuming zero China Data Center compute revenue). China was previously estimated to represent ~20-25% of Data Center revenue before restrictions – any acquisition on that basis is a feat.
Risks to watch for:
– Approval rates are uncertain – case-by-case reviews do not imply high approval rates. Policy objectives, end-user sensitivity, and the risk of disruption all play a role.
– Compliance and compliance costs – know-your-customer requirements, end-use verification, and export compliance infrastructure add cost and operational complexity.
– Risk of policy change – the export control framework could be strengthened again. BIS has shown a willingness to review restrictions and review policies based on national security assessments.
– Re-export and cross-use enforcement – NVIDIA faces exposure to legal liability if authorized shipments are transferred to restricted end-uses or end-users.
Investors should keep an eye out for BIS regulatory updates, NVIDIA’s quarterly earnings disclosures (if provided), and any comments on license approval levels in earnings calls.
