Huawei unveiled its roadmap for the following technology of Ascend AI chips through the current Huawei Join occasion in Shanghai, China
In sum – what to know:
Doubled chip output – Huawei’s plan requires 600,000 Ascend 910C items subsequent yr and as much as 1.6 million dies throughout fashions.
Know-how hole narrowing – Huawei’s 7nm Ascend chips nonetheless path Nvidia’s 4nm GPUs, however advances in packaging and interconnects level to regular progress.
Chinese language vendor Huawei Applied sciences is making ready to considerably broaden manufacturing of its most superior AI processors, in line with a Bloomberg report. The Shenzhen-based firm goals to fabricate about 600,000 items of its Ascend 910C chip in 2026, roughly twice the anticipated output for this yr, the report mentioned, citing individuals conversant in the plans.
Together with different fashions within the Ascend line, Huawei might distribute as much as 1.6 million dies subsequent yr, the report added.
The manufacturing push underscores Huawei’s rising position in China’s effort to cut back reliance on overseas know-how. Partnering with Semiconductor Manufacturing Worldwide Corp. (SMIC), Huawei is utilizing an enhanced 7-nanometer course of to make the chips, whereas rivals like Nvidia depend on extra superior 4nm nodes from TSMC, in line with the report.
Chinese language cloud and web gamers together with Alibaba, Tencent, and DeepSeek are anticipated to be among the many principal prospects, the report added.
For now, analysts count on Huawei’s chips to stay primarily used for inference reasonably than coaching the biggest AI fashions.
Huawei unveiled its roadmap for the following technology of Ascend AI chips through the current Huawei Join occasion in Shanghai, China, outlining plans for 3 new chip collection over the following three years.
The roadmap begins with the Ascend 950 household, which incorporates two chips: the Ascend 950PR, optimized for prefill inference and suggestion, and the Ascend 950DT, designed for decoding and coaching. Each chips share the Ascend 950 Die and ship enhancements in a number of dimensions, the corporate mentioned.
The Ascend 950PR will hit the market within the first quarter of 2026, out there in card and SuperPoD server codecs.
The Ascend 950DT, set to be launched within the final quarter of 2026, targets coaching and decoding workloads with greater calls for for reminiscence and bandwidth. It integrates HiZQ 2.0 HBM, providing 144 GB reminiscence, 4 TB/s reminiscence entry bandwidth, and a 2 TB/s interconnect.
Huawei can even introduce the Ascend 960 chip within the final quarter of 2027. In contrast with the Ascend 950 chip, it’s going to double compute energy, reminiscence, and interconnect capability. The chip additionally introduces HiF4, a proprietary 4-bit precision format designed to ship better accuracy than commonplace FP4 options, promising a significant increase to inference throughput.
The roadmap culminates within the Ascend 970, scheduled for the fourth 2028. Whereas nonetheless below growth, the corporate mentioned the goal is to double FP4 and FP8 compute energy in contrast with the 960, double interconnect bandwidth, and improve reminiscence bandwidth by at the least 1.5 occasions.
“Typically, we’ll observe a 1-year launch cycle and double compute with every launch. All through this course of, we’ll hold evolving our Ascend chips, making them simpler to make use of, supporting extra knowledge codecs, and rising their bandwidth. The aim is to remain on prime of ever-growing demand for AI compute,” Huawei’s rotating chairman Eric Xu mentioned.
“Gaps nonetheless exist between Huawei and Nvidia by way of particular person chips and ecosystems, and we’re working to shut these gaps. However we have now the potential to construct SuperPoDs and SuperClusters – that are the supply of our confidence,” Xu lately instructed Chinese language media.
