The Chinese artificial intelligence startup DeepSeek has quietly excluded major American chipmakers Nvidia and AMD from early access to its newest AI model, a move that carries significant implications for U.S.-China technology competition and the semiconductor industry’s already fraught relationship with Beijing’s most prominent AI lab.
According to a report by The Information, DeepSeek denied early access to its latest model to both Nvidia and AMD, two companies that have long served as the backbone of global AI compute infrastructure. The decision marks a notable shift in how China’s leading AI developers are managing relationships with American technology firms, particularly as Washington continues to tighten export controls on advanced semiconductors destined for Chinese entities.
A Deliberate Distancing From American Chip Giants
The exclusion is especially striking given the deep interdependence between AI model developers and the hardware companies whose chips power their training runs. Nvidia, in particular, has been the dominant supplier of GPUs used in AI training worldwide, and its chips have historically been sought after by Chinese AI labs — sometimes through gray-market channels after U.S. export restrictions limited direct sales of the most advanced processors.
DeepSeek’s decision to cut Nvidia and AMD out of early access suggests the company is actively distancing itself from American semiconductor firms, possibly to avoid drawing further scrutiny from Chinese regulators or to signal alignment with Beijing’s push for technological self-sufficiency. It may also reflect a strategic calculation: by withholding early model access, DeepSeek limits the ability of American chipmakers to optimize their hardware and software stacks for DeepSeek’s architecture, potentially giving Chinese chip designers a relative advantage.
DeepSeek’s Rapid Ascent Has Already Rattled Markets
DeepSeek burst onto the global AI stage in January 2025 when it released its R1 reasoning model, which demonstrated performance competitive with OpenAI’s offerings while reportedly requiring a fraction of the compute resources. The release sent shockwaves through financial markets, temporarily wiping nearly $600 billion off Nvidia’s market capitalization in a single trading session — the largest single-day loss for any U.S. company in history at the time.
The startup, backed by the Chinese quantitative hedge fund High-Flyer, has since become a focal point in debates over whether U.S. export controls on advanced chips are actually working. DeepSeek’s engineers have demonstrated an ability to achieve remarkable model performance using older-generation Nvidia chips and innovative training techniques that reduce the total amount of compute needed. This has led some analysts and policymakers to question whether restricting chip sales to China is merely accelerating Chinese innovation in efficiency rather than slowing AI progress.
Export Controls and the Widening Rift
The Biden administration imposed sweeping restrictions on the sale of advanced AI chips to China beginning in October 2022, with subsequent rounds of tightening in 2023 and 2024. The Trump administration has continued and in some cases expanded these controls. Nvidia has been forced to develop China-specific chips with reduced capabilities — such as the H20 — to comply with the rules, though even these downgraded processors have faced additional restrictions.
In April 2025, the U.S. government required new licenses for the export of Nvidia’s H20 chips to China, a move Nvidia said would result in $5.5 billion in charges. AMD similarly disclosed hundreds of millions in expected losses from tightened export rules. Both companies have publicly warned that overly aggressive restrictions risk ceding the Chinese market to domestic competitors without meaningfully slowing China’s AI development.
What DeepSeek’s Move Means for the Chip Industry
DeepSeek’s exclusion of Nvidia and AMD from early model access adds a new dimension to this dynamic. Historically, chipmakers have relied on close collaboration with leading model developers to ensure their hardware is well-suited to the latest AI workloads. Early access to new models allows chip companies to run benchmarks, optimize drivers, and develop software tools that make their products more attractive to the broader market.
By shutting Nvidia and AMD out of this process, DeepSeek is effectively reducing the information advantage that American chipmakers have enjoyed. If DeepSeek’s models become widely adopted — and there are signs they are gaining traction, particularly among developers in Asia — this could create an opening for Chinese chip companies like Huawei’s HiSilicon, which has been developing its Ascend series of AI processors as a domestic alternative to Nvidia’s products.
Huawei and the Push for Domestic Chips
Huawei has been aggressively positioning its Ascend 910B and newer Ascend 910C processors as viable alternatives for AI training and inference workloads within China. While independent benchmarks suggest these chips still lag behind Nvidia’s most advanced offerings in raw performance, the gap has been narrowing. If DeepSeek optimizes its models to run efficiently on Huawei hardware — and early reports suggest some Chinese AI labs are already doing so — it could significantly boost the commercial viability of China’s domestic chip industry.
This scenario represents one of the outcomes U.S. policymakers have long feared: that export controls, rather than containing Chinese AI capabilities, could catalyze the development of a parallel technology supply chain that eventually competes with American firms on the global stage. Jensen Huang, Nvidia’s CEO, has repeatedly warned of this risk, telling investors and policymakers that “if they can’t buy from us, they’ll build their own — and then we’ll have a competitor we wouldn’t have otherwise created.”
The Broader Geopolitical Context
The timing of DeepSeek’s decision is also notable given the current state of U.S.-China trade relations. The two countries have been engaged in an escalating tariff war, with the Trump administration imposing duties as high as 145% on certain Chinese goods and China retaliating with its own levies. While a temporary 90-day pause in some tariff escalations was announced in May 2025, the underlying tensions remain unresolved, and the technology sector has become one of the primary battlegrounds.
China’s government has made clear that AI is a strategic priority. Beijing has directed significant state funding toward domestic AI champions and has encouraged Chinese companies to reduce their dependence on American technology. DeepSeek’s decision to exclude Nvidia and AMD from early access fits neatly within this broader policy framework, whether or not it was directly encouraged by government officials.
Industry Reactions and What Comes Next
Neither Nvidia nor AMD has publicly commented on being excluded from early access to DeepSeek’s latest model. However, the move is likely to intensify debate within both companies — and across the semiconductor industry — about how to maintain relevance in a market that is increasingly bifurcating along geopolitical lines.
For Nvidia, which derived roughly 17% of its revenue from China before the latest round of export restrictions, the loss of access to DeepSeek’s models is more than a symbolic blow. It represents a tangible erosion of the feedback loop between hardware and software development that has been central to Nvidia’s dominance. If Chinese AI developers increasingly optimize for domestic hardware, Nvidia’s CUDA software platform — long considered one of its most powerful competitive advantages — could see its influence diminish in the world’s second-largest AI market.
A New Phase in AI’s Great Power Competition
AMD faces similar, if somewhat less acute, challenges. The company has been working to expand its presence in the AI accelerator market with its MI300 series of chips, but it has a smaller footprint in China than Nvidia and less to lose in absolute terms. Still, being shut out of early access to one of the world’s most talked-about AI models is not the kind of signal any chipmaker wants to receive.
The DeepSeek episode underscores a fundamental tension at the heart of U.S. technology policy toward China. Washington wants to slow Chinese AI development by restricting access to the most advanced chips. But the restrictions are also pushing Chinese AI companies to develop workarounds, build domestic alternatives, and — as DeepSeek’s latest move suggests — actively decouple from the American technology supply chain. The question policymakers must now grapple with is whether this decoupling serves U.S. strategic interests or whether it is creating the very competitive threat it was designed to prevent.
For investors and industry executives watching the AI chip wars unfold, DeepSeek’s decision to exclude Nvidia and AMD is a data point that deserves close attention. It suggests that the walls going up between the U.S. and Chinese technology sectors are being built from both sides — and that the long-term consequences for the global semiconductor industry are only beginning to come into focus.