Key Notes
- IDC estimates put Nvidia at 34% and Huawei at 28% of China’s first-half AI-accelerator shipments.
- Nvidia’s unit shipments declined 44% from a year earlier as market-access restrictions constrained its business.
- Huawei is expanding its hardware and software ecosystem while Nvidia continues growing globally.
Nvidia Corp. (NASDAQ: NVDA) held 34% of China’s data-center AI-chip market in the first half of 2026, while Huawei reached 28%, according to IDC estimates reviewed by Reuters. The figures show domestic suppliers gaining ground as export restrictions limit Nvidia’s access to Chinese customers.
Reuters reported on October 13 that Nvidia shipped 678,000 AI accelerators in China from January through June, down 44% from a year earlier. Huawei shipped 572,000, compared with more than 300,000 in the same period of 2025.
The results put Huawei roughly six percentage points behind Nvidia by shipment share. They measure China’s AI-accelerator market, rather than worldwide semiconductor sales, and do not establish that the two companies’ products offer equivalent performance.
Market Access Is Reshaping the Competition
Nvidia’s own regulatory filing describes the commercial obstacles behind its Chinese business. For the quarter ended July 26, the company said U.S. and Chinese restrictions prevented it from offering a competitive data-center product for broad distribution with approval from both governments.
The company disclosed that U.S. licenses issued from February allowed small quantities of H200 chips to reach specific Chinese customers. Chinese restrictions nevertheless limited sales, and the shipments Nvidia completed under that program represented less than 1% of its latest quarterly Data Center revenue.
Nvidia also warned that restricted access was helping rivals build larger customer and developer ecosystems. That is a longer-term concern for shareholders: a lost hardware sale can also give another supplier an opportunity to establish its software, technical support and commercial relationships.
The filing describes Nvidia’s ability to compete at the end of its fiscal quarter. IDC’s estimate covers shipments during the first half of the calendar year. Those are different measures, so the reported 34% share should not be treated as a measure of Nvidia’s current ability to accept new Chinese orders.
Shipment Share Does Not Measure Revenue or Performance
The IDC figures reported by Reuters are based on units shipped. A processor count does not capture differences in selling prices, memory, networking or the amount of computing work each device can perform.
Reuters also noted that IDC did not identify which Nvidia products were included in its estimate. That limits conclusions about the mix of hardware behind the market-share number.
For investors, the useful distinction is between commercial adoption and technical leadership. Higher domestic shipments indicate that Chinese buyers are deploying alternatives, but a comparison of training speed, energy consumption or total operating cost would require separate evidence for specific systems and workloads.
Huawei Is Competing Across Hardware and Software
Huawei’s September 17 announcement illustrates the breadth of its approach. At HUAWEI CONNECT 2026, the company introduced the Atlas 960E SuperPoD and described a design combining Ascend processors, optical interconnects and its UnifiedBus architecture.
Huawei says a single Atlas 960E system can scale to 4,096 neural processing units. It also brought forward its roadmap for Ascend 960DT availability to the first quarter of 2027 and Ascend 960PR to the third quarter. These are company specifications and future availability targets, rather than independently verified comparisons with Nvidia’s latest systems.
Software support is another part of the strategy. Huawei said its Ascend ecosystem supports more than 90 leading third-party open-source projects and is supported as a PyTorch accelerator backend. Its CANN software platform has moved toward sustained, community-driven open-source development.
The direction builds on Huawei’s earlier plan to challenge Nvidia through integrated systems, as we previously reported. For customers considering a switch, the ability to run and maintain their models matters alongside the specifications of the processor itself.
Nvidia’s Global Growth Provides a Different Picture
Pressure in China has coincided with strong growth across Nvidia’s wider business. The company’s fiscal second-quarter results showed revenue of $96.2 billion for the period ended July 26, up 106% from a year earlier. Data Center revenue reached $89 billion, an increase of 117%.
Nvidia also forecast third-quarter revenue of $108 billion, plus or minus 2%, without assuming any Data Center compute revenue from China. That was guidance issued with its August results, rather than a new forecast accompanying the IDC report.
As we reported after those earnings, the global AI-infrastructure buildout remains central to Nvidia’s business. The latest Chinese shipment estimates address its competitive position in one market and should not be read as evidence that its worldwide revenue has contracted.
The issue for shareholders is how those two trends develop together: continued growth outside China and a rival ecosystem gaining scale inside it. Future market-share reports, changes in licensing and evidence of customer adoption will help show whether China remains primarily a restricted sales opportunity or becomes a source of stronger competition elsewhere.
