Nvidia's China AI chip share is forecast to collapse from 40% to 8% as Huawei scales

A forecast that redraws the AI chip map

Nvidia held roughly 40% of China's AI chip market as recently as 2025, neck-and-neck with Huawei. By the end of 2026, Bernstein, the global equity research and brokerage firm, projects that figure will collapse to around 8%, while Huawei's share rises to approximately 50%, according to reporting by Fast Company and the Associated Press. For enterprise technology buyers and supply chain leaders evaluating AI infrastructure vendors, that is not an investor story. It is a sourcing risk signal.

The shift is the compounded result of two forces working in tandem. U.S. export controls, first imposed in 2019 and progressively tightened, blocked China from buying Nvidia's most advanced AI chips and the ASML extreme ultraviolet lithography machines needed to manufacture comparable silicon domestically. In response, Beijing redirected procurement toward domestic chipmakers, with Huawei emerging as the primary beneficiary. By the time the Trump administration negotiated a limited reprieve allowing sales of Nvidia's H200 chips to China, the market had already pivoted.

Nvidia CEO Jensen Huang, speaking with the Associated Press, said the company previously held roughly 95% market share in China before export controls effectively shut it out. He framed the dilemma plainly: protecting national security and aggressively competing in global markets are goals that must be pursued simultaneously, not sequentially.

Huawei's Ascend 950 closes the performance gap

Hardware parity is part of what is accelerating the shift. Huawei's Ascend 950 series chips are now viewed by industry analysts as roughly comparable to Nvidia's H200, which has long been considered one of Nvidia's most capable commercial products, according to Fast Company's reporting. Huawei also announced last September that it was deploying large-scale AI computing clusters combining thousands of chips, a capability that previously set companies like Nvidia and AMD apart in global markets.

Antonia Hmaidi of the Mercator Institute for China Studies, who focuses on semiconductors, told the Associated Press that Nvidia has definitively lost significant ground to Huawei, which now leads domestically. He Hui, director of semiconductor research at Omdia, offered a sharper read: China now believes in its own supply self-sufficiency. That confidence, once it sets in at the procurement level inside Chinese enterprises and state institutions, is difficult to reverse even if export restrictions ease further.

There is a caveat that enterprise architects outside China should note. Nvidia's lead in raw chip performance remains intact at the frontier. Fast Company and the Associated Press report that cutting-edge AI development inside China, including training large language models such as DeepSeek, still depends on Nvidia hardware. Several documented cases of chips being smuggled into China to circumvent export controls further illustrate that demand for Nvidia's technology has not disappeared. It has been constrained, not eliminated. Rui Ma, founder of Tech Buzz China, noted to the Associated Press that AI chip demand continues to exceed available supply inside China.

Siemens and Nvidia bet on agentic AI for manufacturing supply chains

Outside China, Nvidia's enterprise partnerships are accelerating. Siemens announced a strategic collaboration with Nvidia to introduce self-verifying agentic AI workflows, targeting high-tech manufacturing pipelines and component production optimization, according to Supply Chain Digital. The announcement positions the two companies at the center of a broader shift toward end-to-end supply chain orchestration, where AI systems can predict disruptions and respond automatically rather than waiting for human intervention.

Supply chain orchestration has become a competitive pressure point for industrial manufacturers. Supply Chain Digital noted that manual processes and fragmented digital systems are increasingly inadequate for the complexity and volatility now standard in global supply chains. Agentic AI, where software agents can execute multi-step decisions and verify their own outputs, represents the next architecture layer above conventional automation and rule-based planning tools.

The Siemens-Nvidia collaboration arrives as other major supply chain technology providers work through their own AI integration roadmaps. For procurement and operations leaders evaluating platform investments, the question is no longer whether agentic AI will reach manufacturing infrastructure, but which vendor stacks will be the standard for that capability within the next product cycle.

What the bifurcation means for enterprise buyers

The two developments, Nvidia's rapid loss of share in China and its deepening role in Western industrial AI, describe the same underlying dynamic from opposite angles. The global AI chip market is splitting into distinct supply ecosystems. Enterprise buyers with operations in both markets face a genuine infrastructure complexity problem: the vendor stack that works in North America or Europe may be unavailable or politically untenable in China, and vice versa.

For CIOs and supply chain technology leaders, vendor concentration risk has moved from a theoretical concern to a documented outcome. Bernstein's forecast of an 8% Nvidia share in China by year-end is a specific, citable benchmark, not a directional guess. Organizations that assumed a single global AI chip strategy could serve all geographies are now navigating a more fragmented reality, one where Huawei's Ascend clusters and Nvidia's GPU platforms are not interchangeable.

Nvidia's next disclosed earnings figures and any further changes to U.S. export licensing will be the concrete markers to watch. Huawei has not publicly confirmed full-year production volumes for the Ascend 950, and capacity constraints on domestic Chinese chip manufacturing remain a variable. How quickly that gap closes will determine whether Bernstein's 8% floor holds or drops further.