SAN FRANCISCO/SINGAPORE, Feb. 25 (GeokHub) — Chinese artificial intelligence lab DeepSeek has withheld its upcoming flagship AI model from major U.S. chipmakers, including Nvidia and Advanced Micro Devices, according to sources familiar with the matter.
The move marks a departure from standard industry practice, where AI developers typically provide pre-release access to leading chipmakers to optimize software performance on widely used hardware.
Instead, DeepSeek granted early access to domestic suppliers, including Huawei Technologies, giving them a head start of several weeks ahead of the anticipated release of its next-generation model, known as V4.
Breaking Industry Norms
Historically, AI developers collaborate closely with U.S. chip manufacturers to ensure peak efficiency on global infrastructure. DeepSeek had previously worked with Nvidia’s technical teams on earlier models.
However, for its forthcoming V4 model — expected around the Lunar New Year period — Nvidia and AMD were reportedly excluded from early optimization efforts.
Neither Nvidia nor AMD commented on the development. DeepSeek and Huawei also did not respond to requests for comment.
Geopolitical and Trade Tensions
The development comes amid heightened scrutiny over AI chip exports between the United States and China. A senior official in the administration of former President Donald Trump recently told Reuters that DeepSeek’s latest AI model had been trained using Nvidia’s advanced Blackwell chips within mainland China — potentially raising export compliance questions.
The official also suggested DeepSeek may seek to obscure technical indicators showing reliance on American AI hardware and instead publicly emphasize the use of Huawei processors.
U.S. authorities previously permitted Nvidia’s H20 and AMD’s MI308 chips — designed primarily for AI inference — to resume shipments to China, while maintaining restrictions on more advanced processors. It remains unclear whether DeepSeek has secured approval for additional U.S. chip purchases.
Despite the exclusion, analysts say the immediate financial impact on Nvidia and AMD may be limited.
Ben Bajarin, CEO of research firm Creative Strategies, noted that DeepSeek’s models often serve more as benchmarking tools rather than enterprise-deployed systems. He added that advancements in AI coding tools have reduced hardware optimization timelines from months to weeks, lessening dependency on early collaboration.
DeepSeek has emerged as a significant force in the AI landscape since early 2025. Its models have surpassed 75 million downloads on the open-source platform Hugging Face, contributing to the rapid rise of Chinese open-source AI systems competing with U.S. laboratories.
The company is among several Chinese AI firms expected to unveil new models this month, intensifying global competition in artificial intelligence development.
