China is preparing to allow some of its largest artificial intelligence companies to purchase a limited number of NVIDIA H200 AI chips, reversing months of resistance to approving imports of the US-made processors, according to a report by The Information.
The report said Chinese officials have informed companies including Alibaba, ByteDance and DeepSeek in recent weeks that they are likely to receive approval to buy H200 chips. However, the number of processors they will be allowed to import is expected to be capped at fewer than 2,00,000 units in total, less than half of what the companies had initially requested.
The move marks a shift in Beijing’s stance. While the US government had already granted export approvals to roughly 10 Chinese companies to purchase H200 chips, Chinese authorities had so far withheld domestic approval as they sought to reduce reliance on foreign semiconductors and encourage adoption of locally developed alternatives.
According to The Information, the approved H200 chips will be restricted to training AI models using public data and cannot be used to handle sensitive customer information. The policy is intended to ease a shortage of advanced AI computing capacity while maintaining safeguards around data security.
China’s leading AI developers have faced growing pressure to secure computing resources as demand for model training continues to rise. The shortage has been compounded by US export controls on advanced AI chips and Beijing’s push to support domestic chipmakers.
NVIDIA shares rose about 1% following the report.
The development comes days after reports emerged that DeepSeek and Z.ai are developing their own AI inference chips to reduce dependence on NVIDIA and Huawei.
NVIDIA CEO Jensen Huang said in May that the company’s market share in China had fallen to zero as US export controls blocked sales of advanced AI chips to the country.
- Deep-Tech and Green Finance to Drive India’s Climate Investments in 2027
- Google’s Project Suncatcher Satellite Generates First AI Response in Orbit
- DeepSeek Proposes Super-Agent Architecture to Automate AI Coding Workflows
- Arrowhead AI Launches Fletch Voice Model for Multilingual Customer Calls
- Amazon Warns Engineers Against AI Productivity Traps in Latest Memo
- Gnani AI Launches Prisma v3.0 Speech Recognition Model for 10 Languages

