Nvidia’s China AI-Chip Share Falls to 34% as Huawei Closes the Gap

Nvidia’s share of China’s data-centre artificial intelligence chip market fell to 34% in the first half of 2026 as US export controls restricted supplies of its processors and domestic competitors gained ground. Huawei, China’s leading domestic challenger in AI accelerators, captured 28% of the market, narrowing the gap with the US chipmaker, according to data from research firm IDC reported by Reuters on October 13, 2026. The figures highlight how geopolitical restrictions and China’s push for semiconductor self-reliance are reshaping competition in one of the world’s largest technology markets.

Nvidia shipped 678,000 AI accelerators in mainland China between January and June, a 44% decline from the same period a year earlier. Huawei shipped 572,000 units, compared with more than 300,000 a year earlier. The broader market also underwent a structural shift: shipments of non-GPU AI accelerators exceeded traditional GPU shipments for the first time in the reported period. Although Nvidia remained the largest individual supplier in IDC’s data, its declining share reflects a market increasingly shaped by export restrictions, domestic alternatives and demand for different types of AI computing hardware.

Key Takeaways

• Nvidia loses market share: Its share of China’s data-centre AI chip market fell to 34% in the first half of 2026.
• Huawei closes the gap: Huawei reached 28% market share after shipping 572,000 AI accelerators.
• Nvidia shipments fall: The company shipped 678,000 accelerators in China, down 44% year over year.
• Domestic suppliers expand: Alibaba’s T-Head, Cambricon, Baidu’s Kunlunxin and Hygon also gained positions in the market.
• Non-GPU chips overtake GPUs: Shipments of other AI accelerators rose nearly 80% to approximately 1.016 million units, exceeding GPU shipments.
• Export controls reshape competition: US restrictions have limited the advanced Nvidia products available to Chinese buyers, while Beijing has encouraged domestic alternatives.
• Market demand remains strong: IDC expects total AI accelerator shipments in China to grow 19% in 2026, even as GPU shipments are forecast to decline.

Nvidia’s Share of China’s AI Chip Market Falls

IDC data cited by Reuters shows that Nvidia’s position in China’s data-centre AI chip market weakened significantly during the first six months of 2026. Its 34% share reflects a market in which domestic suppliers are expanding their presence as access to advanced US-designed processors becomes more restricted.

The company shipped 678,000 AI accelerators during the period, down 44% from a year earlier. AI accelerators are specialised processors designed to handle demanding artificial intelligence workloads, including training models and running inference, the process through which a trained model generates responses or predictions.

The decline in Nvidia’s shipments does not necessarily mean that Chinese companies are using less AI computing power. Total accelerator shipments reached approximately 2.008 million units in the first half of 2026, according to IDC. Instead, the figures show that a larger share of this demand is being met by alternative processors, including chips designed by Chinese companies.

Nvidia remains ahead, but the lead has narrowed

Supplier	AI accelerator shipments, H1 2026	Market share
Nvidia	678,000	34%
Huawei	572,000	28%
Alibaba T-Head	Not specified in the report	7%
Cambricon	Not specified in the report	5%
Baidu Kunlunxin	Not specified in the report	4%
Hygon	Not specified in the report	3%
Other suppliers	—	Remaining market

Source: IDC data reported by Reuters on October 13, 2026. Shares are rounded. Shipment counts for individual suppliers other than Nvidia and Huawei were not specified in the cited report.

Nvidia’s lead over Huawei was only six percentage points in the first half of 2026. That gap is considerably narrower than the company’s historical position in China’s market for advanced AI processors.

However, the figures need to be interpreted within their stated scope. They refer to China’s data-centre AI chip market, not Nvidia’s global GPU business, the entire semiconductor market or all computing devices sold in China.

Huawei Expands Its Position With Domestic AI Accelerators

Huawei shipped 572,000 AI accelerators in China during the first half of 2026, giving it a 28% market share. Its shipments increased from more than 300,000 units in the same period a year earlier, according to IDC.

The increase reflects the growing role of Chinese-designed processors in the country’s AI infrastructure. Huawei has developed its Ascend accelerator family and related computing systems as alternatives for organisations that need to build and operate AI workloads without relying exclusively on imported processors.

Domestic technology companies and public-sector organisations are among the potential customers for these systems. Their purchasing decisions are influenced not only by raw computing performance but also by component availability, software compatibility, pricing and government policy.

Hardware is only one part of the competition

AI accelerators do not operate in isolation. They must work with servers, networking equipment, memory, storage and software frameworks to deliver useful computing capacity.

Huawei’s strategy therefore involves more than supplying individual chips. Its computing systems and software stack are designed to help customers deploy processors across larger clusters. This matters because AI training and inference increasingly require multiple accelerators to work together efficiently.

IDC said the competitive focus in China is shifting from the capabilities of a single chip to overall system performance, including how processors operate in a larger computing cluster.

That shift could create opportunities for domestic suppliers even when individual processors differ from Nvidia products in performance, compatibility or ease of use. A customer evaluating an AI infrastructure investment must consider the complete system, including whether existing software can be adapted and how reliably the system can be operated at scale.

Huawei still faces technical and software challenges

Gaining market share does not automatically establish technological parity with Nvidia across every AI workload. Different processors can perform differently depending on the task, model architecture, memory requirements and software optimisation.

Nvidia benefits from its CUDA software ecosystem, which has become widely used by AI researchers and developers. Companies that have built software and infrastructure around Nvidia hardware may face costs and technical work when switching to a different platform.

Chinese chipmakers must therefore compete on a combination of hardware, software tools, networking, availability and long-term support. The expansion of Huawei’s market share is evidence of increased adoption, but it is not, by itself, proof that every domestic alternative can replace Nvidia in every application.

US Export Controls Reshape the Chinese AI Chip Market

US restrictions on semiconductor exports have been a major factor in Nvidia’s declining share in China. Washington has limited China’s access to its most powerful AI processors, restricting which products Nvidia can legally supply to customers in the country.

The restrictions are intended to limit China’s access to advanced computing capabilities that could support military applications and other strategic technologies. They have also created commercial consequences for US chipmakers by limiting access to Chinese customers.

Nvidia CEO Jensen Huang has repeatedly criticised the restrictions, arguing that they have encouraged Chinese buyers to adopt domestic alternatives and reduced the opportunities available to American companies.

Access to advanced products remains restricted

Nvidia has developed and considered products designed to comply with changing export requirements, but the range of processors available in China is narrower than the company’s full global portfolio.

IDC did not specify which Nvidia products made up its 34% market share in the first half of 2026. It also did not clarify whether the figure included the most advanced B200 and B300 products, which are barred from export to China under the restrictions described in Reuters’ report.

This limitation matters when assessing Nvidia’s position. The market-share figure reflects shipments tracked by IDC, but it does not reveal the performance mix of all processors counted or establish how much computing capacity each supplier delivered.

Nvidia did not respond to Reuters’ request for comment on the report.

Beijing is encouraging domestic alternatives

China has also promoted domestic semiconductor development and sought to reduce reliance on foreign suppliers. Chinese companies have been required to seek approval to import compliant Nvidia chips, according to Reuters.

The policy environment creates incentives for technology companies to consider locally designed processors when building new data centres. Domestic chips may offer advantages in availability or alignment with government objectives, even if switching requires changes to software, infrastructure or workflows.

The result is a market shaped by both commercial and political considerations. Buyers must evaluate technical performance and cost while also accounting for export restrictions, import approvals and the possibility that product availability could change.

For Nvidia, these constraints create a difficult commercial trade-off: maintaining a presence in China while complying with US rules and competing against suppliers that benefit from the country’s drive for technological self-reliance.

Chinese Chipmakers Gain Beyond Huawei

Huawei is not the only domestic supplier gaining ground. IDC’s figures show that several Chinese chip designers expanded their positions in the first half of 2026.

Alibaba’s T-Head held 7% of China’s AI accelerator market, followed by Cambricon at 5%, Baidu’s Kunlunxin at 4% and Hygon at 3%. Nine of the 11 Chinese chipmakers tracked by IDC increased their shipments compared with a year earlier.

These companies operate in a competitive market with different products, customer relationships and technical approaches. Their combined growth suggests that China’s shift toward domestic AI hardware is broader than a single company replacing Nvidia.

Alibaba, Baidu and Cambricon add to domestic supply

Alibaba’s T-Head develops processors for computing workloads connected to the company’s technology and cloud operations. Baidu’s Kunlunxin focuses on AI processors, while Cambricon specialises in AI chip design. Hygon also supplies processors used in China’s computing ecosystem.

Their expansion gives Chinese customers more options when planning AI infrastructure. It may also support the development of software tools, engineering expertise and supply-chain capabilities within the domestic semiconductor sector.

However, market share does not provide a complete measure of commercial success. Revenue, margins, processor performance, software maturity and the ability to supply large deployments are also important. IDC’s shipment figures indicate adoption, but they do not establish that all suppliers are equally profitable or technologically interchangeable.

Non-GPU AI Accelerators Overtake Traditional GPUs

One of the most significant developments in IDC’s data is the changing composition of China’s AI accelerator market.

Total AI accelerator shipments reached 2.008 million units during the first half of 2026. Shipments of traditional graphics processing units, or GPUs, declined 26% to 991,000 units. By contrast, shipments of other AI accelerators rose nearly 80% to approximately 1.016 million units, overtaking GPUs by volume.

The category includes processors designed for specific AI tasks, particularly inference. Unlike general-purpose GPUs, some specialised accelerators are optimised for defined workloads and may offer different trade-offs in performance, power consumption and cost.

Why specialised processors are gaining ground

AI computing includes several distinct tasks. Training a large model requires substantial computing resources to process data and adjust model parameters. Inference involves running a trained model to produce outputs for users or applications.

The hardware requirements can differ between these tasks. A processor designed for a particular inference workload may be attractive to a company running a large number of repeated operations, even if it is not suitable for every training job.

Specialised processors can also be integrated into systems designed around specific applications. Government agencies, internet companies and telecommunications providers may choose hardware based on their workload requirements, available software and deployment constraints.

IDC reported that Huawei, T-Head, Kunlunxin and Cambricon together accounted for 95% of China’s non-GPU accelerator segment. Their position suggests that domestic companies have become particularly influential in this part of the market.

Shipment volume does not equal computing performance

The fact that non-GPU accelerators surpassed GPUs by shipment volume does not establish that they delivered more total computing power. A shipment count treats each unit as one device, even though processors can differ substantially in performance and capabilities.

Likewise, a higher number of specialised chips does not prove that they can replace high-end GPUs for every large-model training workload. Comparisons require information about the specific chips, their configurations, software support and the tasks they perform.

The change nevertheless indicates that China’s AI infrastructure market is becoming more diverse. Buyers are adopting different types of processors rather than relying exclusively on conventional GPUs.

China’s AI Computing Market Is Still Growing

Despite restrictions on advanced GPU shipments, IDC expects China’s total AI accelerator shipments to grow 19% in 2026. The research firm forecasts a 9% decline in GPU shipments for the full year.

The projections suggest that demand for AI computing is continuing to expand even as the mix of hardware changes. Chinese technology companies are investing in models, applications and data-centre infrastructure, creating demand for processors capable of handling training and inference.

The figures are forecasts, not final annual shipment results. Actual outcomes will depend on component availability, production capacity, demand from data-centre operators and the pace at which companies deploy new systems.

Domestic production must keep pace with demand

Expanding domestic chip supply requires more than designing a processor. Suppliers must secure manufacturing capacity, package the chips, provide supporting components and build software that enables customers to use them efficiently.

Scaling production can be particularly challenging for advanced semiconductor products because manufacturing equipment, materials and technical expertise are distributed across an international supply chain.

China’s domestic chipmakers have made progress in increasing shipments, but the pace and sustainability of that expansion remain important questions. A growing market share does not eliminate constraints related to manufacturing yields, advanced packaging, memory supply or software development.

For buyers, the availability of sufficient quantities of reliable processors may matter as much as a chip’s theoretical performance. For suppliers, the challenge is to turn early adoption into a consistent supply of products that can support large deployments.

What the Market Shift Means for Nvidia

China remains commercially important to Nvidia because it has a large technology sector and substantial demand for AI computing. Yet export controls and domestic competition have made the market more difficult to serve.

The company faces a two-sided challenge. It must comply with US restrictions while offering products that remain attractive to Chinese customers. At the same time, local suppliers are gaining experience, developing software ecosystems and building relationships with customers who may prefer domestic systems.

Software compatibility remains an advantage

Nvidia’s CUDA ecosystem is one of its major competitive strengths. Developers have built tools and applications around its software, making it easier to deploy workloads on compatible hardware.

Moving to another processor can require software adaptation, testing and performance optimisation. These costs can make customers reluctant to switch even when alternative chips are available.

However, restrictions on access to Nvidia products can change the calculation. If a buyer cannot obtain a desired processor or faces uncertainty about future supply, the incentive to invest in a domestic alternative may increase.

The long-term outcome will depend partly on whether Chinese suppliers can make their hardware easier to use and support a broad range of AI applications. It will also depend on the availability and performance of Nvidia’s permitted products.

Nvidia’s China position should not be confused with its global business

The 34% figure applies to China’s data-centre AI chip market in the first half of 2026. It does not mean Nvidia has lost 66% of its global market or that its worldwide AI business is shrinking by the same amount.

The company’s performance in other markets, its product pipeline and demand for AI infrastructure outside China are separate factors. Any assessment of Nvidia’s global business requires financial results and market data covering those broader activities.

The China figures are nevertheless significant because they show how government restrictions can alter market access and encourage competitors to build alternatives in a strategically important industry.

The Bigger Picture

The competition between Nvidia and Huawei reflects a wider contest over the infrastructure that will power the next generation of artificial intelligence. Advanced processors are central to AI model development and deployment, but access to them increasingly depends on geopolitics, supply chains, software ecosystems and domestic industrial policy. China’s market is demonstrating that restrictions on imported hardware can accelerate demand for local alternatives, particularly when customers have strong incentives to maintain access to computing capacity.

The data also shows that competition is not simply a contest between two chipmakers. Alibaba, Cambricon, Baidu and Hygon are expanding alongside Huawei, while non-GPU accelerators are gaining ground against conventional GPUs by unit shipments. This broadening supplier base may give Chinese customers more options, but it does not establish that all domestic products can match Nvidia across every workload. System performance, software compatibility, manufacturing capacity and reliability will determine how far the shift can go.

Looking Ahead

China’s AI accelerator market is expected to grow in 2026 even as restrictions limit the supply of advanced GPUs. IDC’s forecast of 19% growth in total accelerator shipments, alongside a projected 9% decline in GPU shipments, points to a changing hardware mix rather than a collapse in demand. The key indicators to watch will be domestic production volumes, adoption of specialised accelerators, software improvements and the availability of Nvidia products that comply with export rules. Subsequent IDC data will help establish whether Huawei and other domestic suppliers can sustain their gains.

For Nvidia, preserving its position in China will depend on the products it can legally supply and the continued value of its software ecosystem. For Huawei and other Chinese chipmakers, the challenge is to convert rising shipments into dependable, scalable computing systems that meet the needs of large AI deployments. The first-half figures show that the competitive gap has narrowed, but the longer-term outcome remains uncertain. Market share, processor performance and total computing capacity will need to be assessed separately as the industry develops.