IDC: Nvidia's AI chip shipments in China fell 44% YoY to 678K in H1 2026, as its market share slid to 34% amid US export curbs and rising domestic competition (Reuters)

Nvidia’s shipments of data-centre AI chips in China fell to 678,000 in the first half of 2026. That was a 44% decline from the same period a year earlier. Its share of China’s market also dropped to 34%. These figures show a sharp weakening in Nvidia’s position in an important AI market.

The main forces were restricted supply and stronger local competition. US export controls limited which advanced Nvidia chips could be shipped to Chinese customers. At the same time, Chinese chipmakers gained ground. Together, those pressures reduced Nvidia’s ability to serve demand and made more room for domestic suppliers.

The immediate reality is a smaller Nvidia presence in China’s data-centre AI-chip market. The article does not identify the individual domestic companies or specify the exact export rules involved. However, the reported figures indicate that regulation and competition are affecting both Nvidia’s sales volume and its market share.

A shipment total of 678,000 sounds large, but the 44% year-over-year fall shows how much Nvidia’s supply position weakened. A 44% decline means the company shipped only 56% as many chips as in the comparable period of 2025. Using the reported figure, that implies roughly 1.21 million shipments a year earlier.

In practical terms, Nvidia shipped about 535,000 fewer chips than that estimated prior-year level. This is an approximate calculation, because the article provides the percentage and current total rather than the earlier total directly. It still illustrates the scale of the contraction clearly.

The decline matters because data-centre AI chips support large computing systems. Fewer shipments can limit Nvidia’s ability to supply Chinese customers and can create opportunities for domestic competitors. The article links the fall to US export controls and rising competition from Chinese chipmakers.

Data-centre AI chips are processors built for large server facilities rather than ordinary personal computers. They handle the repeated mathematical operations behind training AI models and producing answers from trained models. This work requires many calculations at the same time, often across large networks of connected chips.

Nvidia’s products are commonly used because its processors are designed for parallel workloads and are supported by a broad software platform. That software helps researchers and companies develop, train, and run AI systems without building every tool from scratch. The source article focuses on these chips’ market shipments, not their technical specifications.

Their importance explains why Nvidia’s reduced Chinese shipments matter. If customers cannot obtain Nvidia hardware, they may turn to domestic alternatives or redesign systems around different processors. In China, the article reports that Nvidia’s share fell to 34% as US export controls restricted supplies and local competition increased.

US export controls are government rules that can restrict companies from selling certain advanced technology to particular countries or customers. In this case, the controls limited Nvidia’s ability to supply AI chips to Chinese customers. The article directly links those restrictions to reduced supplies, but it does not identify the specific regulations, products, or licenses involved.

The mechanism is straightforward: when Nvidia cannot freely ship covered chips, Chinese data-centre operators receive fewer Nvidia products. Customers may need to delay purchases, seek alternative processors, or use chips with different capabilities. That reduces Nvidia’s sales opportunity even when demand for AI computing remains strong.

The reported result was a 44% year-over-year fall in shipments, to 678,000 chips in the first half of 2026. Nvidia’s Chinese market share also slid to 34%. The controls therefore affected not only supply volumes but also Nvidia’s competitive position as domestic Chinese chipmakers gained ground.

Nvidia’s position in China weakened during the first half of 2026. Its share of the data-centre AI-chip market fell to 34%, while shipments dropped 44% year over year to 678,000. This means Nvidia remained a major supplier, but it held a much smaller portion of the market than before.

Two forces explain the change reported in the article. US export controls restricted Nvidia’s ability to supply Chinese customers. Meanwhile, domestic Chinese chipmakers increased their competitive presence. With fewer Nvidia products available, local suppliers had more opportunity to win customers and expand their share.

The shift makes China a more contested market for Nvidia. The source does not name the domestic companies or describe their individual products, so their specific contributions cannot be measured here. Still, the figures show that supply restrictions and local competition combined to reduce Nvidia’s shipments and market share.

The supplied article says only that domestic Chinese chipmakers are gaining ground; it does not name them. Based on established industry knowledge, examples include Huawei, Cambricon, Biren Technology, Moore Threads, and Enflame. Their products generally target the same broad need: accelerating the calculations used by AI systems in data centres.

Competition can come from different chip designs. Graphics processing units, or GPUs, use many parallel computing units. Neural-processing units, or NPUs, are designed specifically for neural-network workloads. Application-specific integrated circuits, or ASICs, can be optimized for particular AI tasks. CPUs may also handle supporting work, though they are usually less specialized for massive parallel AI calculations.

The article’s measured outcome is clear even though it gives no company-by-company breakdown. Nvidia’s shipments fell to 678,000 in the first half of 2026, and its market share fell to 34%. Export controls and growing domestic competition both contributed to that change.

Advanced AI chips are strategically important because AI models require enormous amounts of computation. Countries and companies with access to powerful processors can develop models, operate services, and support data-centre capacity more effectively. The source illustrates this importance through Nvidia’s falling shipments and market share in China, where access to AI chips is being shaped by export controls and competition.

Inside a semiconductor, electrical signals move through tiny switches called transistors. Patterns of on and off states represent data and instructions. The chip combines those signals through logic circuits and arithmetic units. AI accelerators repeat matrix and vector calculations in parallel, turning stored model data into training updates or outputs from a trained model.

That hardware is connected through memory, networking, and software so many chips can work together. Faster and more efficient chips can support larger workloads. Nvidia’s 678,000 shipments and 34% Chinese market share show how access to this underlying computing capacity has become commercially and strategically significant.