China is engaged in a fevered, state-backed campaign to develop domestic alternatives to American AI chips, led by Vice Premier Ding Xuexiang, a confidant of President Xi Jinping, who initiated the push after Washington’s export controls threatened to cripple China’s AI ambitions. A Wall Street Journal investigation into the campaign reveals a system of intense political pressure, ambitious timelines, and a dramatic collapse in Nvidia’s market position in China that would have been unimaginable five years ago.
The Huawei Briefing That Launched the Campaign
The campaign’s origins trace to a closed-door briefing that Huawei Technologies held for Ding Xuexiang last summer. Huawei laid out a plan for rivaling Nvidia and stated that China could become self-sufficient in some critical areas of AI within three years. It was, according to the Journal, exactly what Ding wanted to hear. With Xi’s blessing, Ding launched a coordinated effort to accelerate domestic chip development and force Chinese AI companies to adopt local alternatives, regardless of whether those alternatives were yet competitive with the hardware they were replacing.
The pressure applied to Chinese tech companies has been extraordinary. According to the Journal’s reporting, Chinese AI companies were warned that anyone who resisted using local chips was a “traitor”, a characterization that carries significant weight in China’s political environment and leaves little room for commercial resistance.
This coercive dynamic has accelerated the adoption of Huawei’s Ascend chips and other domestic alternatives, even where those alternatives remain less capable than the Nvidia hardware they replace. Companies that have publicly committed to domestic chip adoption include ByteDance, Alibaba, Baidu, and Tencent. They collectively represent the largest AI compute buyers in China.
Nvidia’s Dramatic Collapse in China’s AI Chip Market
The results for Nvidia have been dramatic. The company’s share of China’s AI chip market has collapsed from 90% in 2021 to less than 60% by 2025, a decline that reflects both the success of China’s domestic chip push and the impact of US export controls that have restricted Nvidia’s ability to sell its most advanced chips to Chinese customers. China’s domestic chipmakers have seized 41% of the local AI market, a share that would have been unimaginable five years ago.
This shift is not purely the result of government pressure. As ByteDance, Alibaba, and Baidu race to buy domestic GPUs, the commercial logic of domestic substitution has become increasingly compelling. Companies that invest in mastering Huawei’s Ascend ecosystem now are building capabilities that will be valuable regardless of what happens to US export controls, and they are reducing their exposure to the geopolitical risk of depending on hardware that Washington can restrict at will.
The ban on foreign AI chips from state-funded data centers has created a guaranteed domestic market for Chinese alternatives, further accelerating the shift.
The Compute Gap and the Three-Year Horizon
Despite the progress, a significant gap remains. China’s AI computing power stood at around 14% of the US in 2025, according to the Journal’s reporting, a figure that reflects both the quality differential between Chinese and American AI chips and the sheer scale of US investment in AI infrastructure. Bridging this gap within three years, as Huawei promised Ding, is an ambitious target that will require not just better chip design but advances in manufacturing, software ecosystems, and system-level integration.
The question of whether China’s chip industry can close the performance gap fast enough to support frontier AI development, rather than just satisfying government procurement requirements, remains the central uncertainty in China’s AI strategy. Guaranteed demand does not automatically produce competitive supply.
A chip that is mandated into government data centers but cannot train competitive frontier models does not advance China’s AI ambitions in the way that Beijing is betting on. The next 18 months, as Huawei’s three-year timeline approaches its midpoint, will be a critical test of whether the campaign is delivering results that match its ambitions.
The Software Ecosystem Problem That Hardware Alone Cannot Solve
The Journal’s investigation focuses primarily on hardware, but the chip campaign faces an equally difficult software challenge. Nvidia’s dominance in AI computing is not just a function of its GPU specifications. It is a function of CUDA, the software ecosystem that makes Nvidia hardware easy to program and that has accumulated more than a decade of optimization from researchers and engineers worldwide.
Chinese alternatives like Huawei’s Ascend chips come with their own software stacks (CANN, MindSpore), but these ecosystems are far less mature, and the friction of porting AI workloads from CUDA to Ascend is a significant barrier to adoption.
Beijing is aware of this problem. The Ministry of Industry and Information Technology has funded efforts to build CUDA-compatible software layers that allow existing code to run on domestic chips with minimal modification, and several Chinese universities have launched programs to train engineers in domestic AI chip programming.
Whether these efforts can close the software ecosystem gap as quickly as the hardware gap is an open question and one that will determine whether China’s all-out chip push produces a genuinely competitive domestic AI computing stack or merely a politically compliant one.