Chinese regulators have granted conditional approval for leading AI startup DeepSeek to purchase Nvidia Corp.’s H200 artificial intelligence chips, according to two sources familiar with the matter reported by Reuters.
The clearance, which follows similar permissions for tech giants ByteDance, Alibaba, and Tencent to acquire over 400,000 H200 units collectively, comes with stipulations still under finalization by China’s National Development and Reform Commission (NDRC), the nation’s top economic planning authority.
Approvals for all four companies were issued by China’s Ministry of Industry and Information Technology and Ministry of Commerce, but the exact conditions—potentially including usage restrictions, reporting requirements, or preferences for domestic alternatives—remain in flux.
Neither the ministries nor the NDRC responded to requests for comment, and DeepSeek, headquartered in Hangzhou, did not reply to inquiries.
Nvidia CEO Jensen Huang, addressing reporters in Taipei on January 29, stated that his company had not yet received confirmation of the approvals and believed Chinese authorities were still finalizing licensing details.
Nvidia did not comment specifically on DeepSeek’s clearance, underscoring the opaque and protracted nature of cross-border AI hardware deals amid U.S. export controls.
The H200, Nvidia’s second-most-powerful AI processor, cleared for export to China, has symbolized the escalating U.S.-China tech rivalry. While the U.S. government under the Trump administration formally authorized H200 sales earlier in January 2026, Beijing had delayed imports, reportedly urging firms to prioritize homegrown alternatives to foster self-reliance in semiconductors.
Nvidia has faced significant revenue headwinds from these curbs, estimating $8 billion in lost sales from prior restrictions.
DeepSeek’s involvement adds a layer of complexity, given its rapid ascent and ties to U.S. scrutiny. Founded in 2024 and backed by investors including High-Flyer Capital Management, the startup disrupted the global AI landscape in early 2025 with models like DeepSeek-V3, which achieved frontier-level performance at a fraction of the compute cost of U.S. counterparts such as OpenAI’s offerings.
V3’s efficiency—requiring just 2.788 million GPU hours on Nvidia’s restricted H800 chips—stemmed from what U.S. lawmakers allege was direct technical assistance from Nvidia, including co-design of algorithms, frameworks, and hardware.
On January 28, House Select Committee on China Chairman John Moolenaar (R-Mich.) sent a letter to U.S. Commerce Secretary Howard Lutnick, citing internal Nvidia records to claim the company aided DeepSeek in honing models later linked to Chinese military applications. Moolenaar urged strict enforcement of U.S. conditions for H200 shipments, warning that any DeepSeek purchases could violate export controls prohibiting sales to entities assisting the People’s Liberation Army.
U.S. officials believe DeepSeek has supported China’s military and evaded restrictions, potentially inviting bipartisan congressional backlash.
Huang’s recent diplomatic engagements in China—including a January 2026 visit where he emphasized Nvidia’s commitment to the market—may have influenced the approvals.
Analysts suggest these conditional clearances represent a partial easing, not a full reopening, as Beijing weighs AI competitiveness against self-sufficiency goals. Others speculated on DeepSeek’s V4 model benefiting from the chips, amid broader discussions of U.S.-China tech decoupling.
Nvidia shares dipped modestly in premarket trading, reflecting ongoing uncertainties. However, the approval is expected to test U.S. resolve on enforcement, as NDRC finalizes terms, potentially escalating tensions if military links are substantiated.







