TLDRs;
- Nvidia unsure China will accept downgraded H200 AI chips.
- Huawei’s Ascend chips gain market share in China’s AI sector.
- Integrators profit migrating workloads from CUDA to Huawei’s CANN.
- U.S. export rules keep Nvidia’s China strategy uncertain.
Nvidia (NVDA) stock rose just over 2% on Thursday despite CEO Jensen Huang has raising doubts over the company’s ability to sell its advanced H200 AI chips to China, even if U.S. export restrictions were relaxed.
Speaking to reporters after a meeting with former President Donald Trump and ahead of a closed-door session with the Senate Banking Committee, Huang emphasized that any downgraded version of the H200 chip would likely be unacceptable to Chinese buyers.
The U.S. has maintained export controls on high-performance AI chips since 2022, aiming to prevent sensitive technology from reaching China. While policymakers have explored legislation like the GAIN AI Act, which would prioritize U.S. customers for AI chip access before foreign sales, the measure has yet to be enacted. These ongoing regulations have created uncertainty for both Nvidia and Chinese tech companies seeking access to cutting-edge AI hardware.
Huawei Gains Ground Domestically
China’s domestic AI chip industry is quickly strengthening, reducing the potential impact of relaxed U.S. export rules. Huawei’s Ascend series, including the 910B and 910C models, has been adopted by numerous cloud and AI firms. In 2024, Huawei sold roughly 200,000 AI chips, compared to Nvidia’s more than one million H200 units.
The company’s Atlas 900 A3 SuperPoD clusters, featuring up to 384 Ascend 910C chips, deliver up to 300 petaflops, making them the largest SuperPoDs globally as of 2025. Looking ahead, Huawei’s Atlas 950 and 960 systems are projected to host over 520,000 and 1 million Ascend chips, respectively, reaching exaFLOPS and zettaFLOPS performance levels by 2026–2027.
Migration Opportunities for Integrators
As Nvidia navigates export uncertainties, opportunities are emerging for systems integrators and IT consultancies. Huawei’s CANN (Compute Architecture for Neural Networks) is compatible with some CUDA workloads, though it lags in performance and stability. This gap creates a niche for service providers to assist companies in transitioning their AI systems to CANN while Nvidia supply remains constrained.
Huawei plans to open-source CANN and associated tools by the end of 2025, accelerating adoption and giving integrators more time to develop migration expertise. Meanwhile, domestic rivals like Cambricon, Alibaba’s T-Head, and Baidu-backed Kunlun Tech are shipping CUDA-compatible chips, easing the transition and creating additional business opportunities.
Policy Uncertainty Drives Market Strategy
Nvidia’s China dilemma highlights the broader impact of policy uncertainty on global AI markets. Export restrictions, ongoing regulatory reviews, and emerging domestic alternatives force companies to balance technology leadership with geopolitical realities. Analysts note that some Chinese firms, including Tencent, have already pulled back from testing Nvidia chips in favor of domestic options.
The combination of U.S. policy, Chinese innovation, and growing competition means that Nvidia may need to rethink its approach to one of the world’s largest AI markets. While H200 chips remain a leading solution, their acceptance in China is far from guaranteed, illustrating the complex interplay between technology, trade policy, and global AI development.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
