A Bloomberg Intelligence survey released on July 7 has put a precise number on a trend that chip industry analysts have been tracking for months: Chinese technology executives plan to allocate 46 percent of their AI accelerator budgets to domestic chip suppliers over the next twelve months, up from 30 percent today. The 16-percentage-point swing in a single planning cycle represents the most concrete evidence yet that US export controls are accelerating a structural shift in China’s AI hardware market, one that is moving faster than even optimistic projections for domestic chip adoption suggested two years ago.
The survey, reported by AI Weekly and confirmed by Seeking Alpha, also found that 80 percent of the executives surveyed said their total AI infrastructure spending is over budget this year. That figure captures a second dimension of the domestic chip story: Chinese companies are not simply choosing domestic chips on patriotic grounds or in response to regulatory pressure, they are doing so in an environment of severe cost pressure, where the unavailability of Nvidia’s most advanced chips has forced them to find alternatives regardless of preference.
The Domestic Chip Ecosystem Taking Shape
The Bloomberg survey does not name which domestic vendors are absorbing the shift in budget allocation, nor does it disclose its sample size or industry mix. But the broader context makes clear which companies are positioned to benefit. Huawei currently holds approximately 50 percent of China’s domestic AI chip market, which Macquarie estimated at $50 billion in a July 5 initiation report that also covered Cambricon, Hygon, MetaX, and Biren Technology. Macquarie projects the domestic market will grow from approximately $21 billion to $67 billion by 2030, a trajectory that the Bloomberg survey data suggests may be conservative.
Alibaba and Baidu have both developed proprietary AI chips for internal use, reducing their dependence on any single external supplier. The most significant new development in this space came on the same day as the Bloomberg survey: Reuters reported that DeepSeek is developing its own AI inference chip, an effort that began approximately a year ago and involves outreach to chip-design, foundry, and memory companies as external partners. If DeepSeek, which trained its R1 model on Nvidia H800 chips and adapted its V4 model for Huawei Ascend, succeeds in building its own inference silicon, it would add another major player to a domestic ecosystem that is already expanding rapidly.
(Related: China’s Domestic AI Chips Have Entered the Training Stack — and the Efficiency Gap Is Closing)
Macquarie’s Coverage Initiation and the Investment Signal
The Macquarie initiation on July 5 is worth examining alongside the Bloomberg survey data, as together they describe a market attracting both strategic capital and financial investment. Macquarie named Biren Technology as a key player in the domestic chip buildout — a designation that came just two days after Biren announced a HK$7 billion share sale to fund GPU production capacity. The bank’s projection that the Chinese domestic AI chip market will reach $67 billion by 2030 implies a compound annual growth rate that would make it one of the fastest-growing segments in the global semiconductor industry.
The survey’s finding that 80 percent of Chinese AI infrastructure projects are running over budget also has implications for the domestic chip vendors. Cost overruns at the project level create pressure to find cheaper alternatives to Nvidia’s premium-priced H100 and H200 chips — exactly the competitive window that Huawei’s Ascend series and Cambricon’s MLU series are designed to fill. As the performance gap between domestic and US chips narrows, the cost argument for domestic silicon becomes increasingly compelling even for companies that would otherwise prefer Nvidia’s ecosystem.
The Nvidia Exposure
Nvidia’s exposure to this shift is significant. The company’s China revenue has already been constrained by export controls that bar the sale of its most advanced chips to Chinese customers. The Bloomberg survey data suggests that even the chips Nvidia is permitted to sell in China, including the H20, a downgraded version of the H100 designed to comply with export control thresholds, are losing ground to domestic alternatives as Chinese companies build out their procurement planning around a domestic-first assumption.
Nvidia shares slipped approximately 1.6 percent in premarket trading on July 7 following the Reuters report on DeepSeek’s chip development, a reaction that reflects investor sensitivity to any news that accelerates the domestic chip transition in China. The Bloomberg survey adds to that picture by showing that the transition is not just a technology story but a procurement-planning reality already embedded in Chinese corporate budgets for the coming year.
The longer-term implication of the Bloomberg data is that the Chinese domestic chip market is approaching a tipping point. At 30 percent current allocation, domestic chips are a supplement to Nvidia in most Chinese AI infrastructure deployments. At 46 percent, the projected allocation for the next 12 months, domestic chips become the primary compute layer for many organizations, with Nvidia relegated to a secondary role for workloads that specifically require its capabilities. That structural shift, if it materializes as the survey suggests, would represent the most significant change in the global AI chip market since the original export controls of October 2022 first forced Chinese companies to begin building domestic alternatives.
The companies that have invested in that domestic ecosystem, Huawei, Cambricon, Biren, and the foundries and memory suppliers that support them, are positioned to benefit from a transition that is now being driven not by government mandate alone but by the procurement decisions of Chinese enterprises managing constrained budgets in a competitive AI market.