Huawei AI chips projected to grow 60% to 12 billion: capturing NVIDIA China orders

Chinese tech giant Huawei estimates that in 2026, revenue from its AI chips will grow by at least 60%, reaching $12 billion, up significantly from $7.5 billion in 2025. The figure was cited by the UK’s Financial Times from Huawei internal sources and relayed by Reuters. The core driver behind Huawei’s goal is that its latest processor, Ascend 950PR, entered mass production in March 2026 and has already won most of the 2026 customers’ orders; the next-gen Ascend 950DT is expected to be launched in Q4 2026. Reuters noted that it “could not independently verify” the claims when relaying them, but the 60% growth rate aligns with Huawei’s previously disclosed pacing for its AI business.

$7.5 billion in 2025 → $12 billion in 2026: 60% YoY growth rate

Huawei’s AI chip business had an estimated revenue baseline of about $7.5 billion in 2025, with a 2026 target of $12 billion, implying annual growth of +60%. This growth rate is significantly higher than the overall growth rate of the global AI chip market (NVIDIA’s 2026 forecast is 40-50% growth, while AMD is estimated to grow 35-45%), reflecting Huawei’s special position in the China domestic market— as US export restrictions on the NVIDIA H series and AMD MI series continue, Chinese AI companies are forced to seek local alternatives, and Huawei’s Ascend line has become a de facto replacement for NVIDIA’s H100/H200.

Another signal from the FT report is that “Ascend 950PR has already captured most orders for 2026.” In the AI chip industry, locking in orders for the full year early suggests customers have a high level of confidence in supply stability—Huawei is no longer just an “emergency substitute,” but has moved into the core position of production line planning. Customers include major Chinese AI players such as DeepSeek, Alibaba Cloud, and Tencent Cloud.

Product roadmap: Ascend 950PR enters mass production in March, with 950DT taking over in Q4

Ascend 950PR is Huawei’s latest processor in the Ascend series. It entered mass production in March 2026 and uses SMIC’s 7nm process (unable to use TSMC’s 5nm or 3nm due to US sanctions). Although the process lags one generation behind NVIDIA H200’s 4nm, 950PR’s appeal in the China market mainly comes from three points: one, it can be delivered smoothly (NVIDIA is constrained by export controls); two, software-hardware integration (Huawei provides an end-to-end solution); three, pricing strategy (internal pricing is lower than NVIDIA’s similarly positioned products).

The next-generation Ascend 950DT is expected to be launched in Q4 2026, potentially adopting a more advanced process or higher-density packaging. It will be Huawei’s core product for driving a larger order scale in 2027. For the global AI compute supply structure, the rapid expansion of the Ascend line in the China market means NVIDIA’s share in China’s AI market will continue to be eroded. This trend contrasts with the US’s latest crackdown on Chinese companies “using US-made AI models.”

What to watch next: actual shipments vs. order targets, process upgrade pacing, export-control countermeasures

Achieving the $12 billion target requires that the actual shipment pacing of Ascend 950PR keeps up. Items to watch next include: first, whether SMIC can reliably supply the 7nm wafer production capacity Huawei needs; second, after the 950DT launch in Q4, whether 950PR orders lose momentum; third, whether US export controls further expand to Huawei’s chip design tools or EDA software layers, counteracting the rise of domestically produced AI chips in China. The indirect impact on the crypto industry is that “China-based AI computing” could accelerate, potentially leading to a “China version vs. Western version” split in global AI agent tooling, which would tie into the choice and acceptance of crypto AI tools (for example, DeepSeek using Polymarket vs. OpenAI using Polymarket).

This article, “Huawei’s AI chip revenue growth estimated at 60% to $12 billion: grabbing NVIDIA’s China orders,” first appeared on Lianxin ABMedia.