Huawei Accelerates AI Chip Roadmap, Targets 2027 Launch for Ascend 960 Series

Huawei Technologies will launch its flagship Ascend 960DT artificial-intelligence chip in the first quarter of 2027. [3][4] Rotating chairman David Wang announced the timeline at the company’s annual summit in Shanghai, marking a faster rollout than the originally planned late 2027 release. [3] Another variant of the new series, called the 960PR, will be delivered in the third quarter of 2027. [3][7]

The Ascend 960 is reported to double the compute performance of its predecessor. [6] The company intends to release a new generation of Ascend chips every year, with the 970 and 980 models expected to follow in 2028 and 2029. [14]

To build larger computing systems, Huawei is relying on its proprietary interconnect technology marketed as UnifiedBus. [10] The technology helps connect large numbers of chips to share information and work together more efficiently. [7] Huawei has developed 11 semiconductors based on UnifiedBus technology for use in its large systems. [7]

Huawei has already shipped more than 1,000 smaller linked systems to over 370 customers. [7] Looking ahead, the company's SuperPod connectivity tech will soon allow as many as 100,000 Ascend chips to operate in a single cluster. [4] For its largest linked systems, which Huawei calls superclusters, the UnifiedBus technology can support up to 1 million AI processors. [7][15]

Domestic demand for Huawei's current generation of hardware remains high among Chinese AI model developers. [8] The Ascend 950 family includes the 950PR aimed at inference workloads and the 950DT optimized for training and decoding tasks. [10] DeepSeek is planning to deploy at least 160,000 Huawei Ascend 950DT chips at a data center it is building in Inner Mongolia. [4]

However, rotating chairman Eric Xu noted that Huawei cannot produce enough artificial-intelligence computing equipment to meet demand in China. [8][14] Because of this capacity constraint, Huawei is limiting overseas sales and does not have a plan to expand into the international market in a fully-fledged way. [14] Huawei supplies some countries where demand is particularly strong, but overall international volumes remain limited. [14]

The push for domestic computing infrastructure comes as Washington imposes export controls that restrict China's access to certain advanced computing chips and semiconductor manufacturing equipment. [7] Huawei has faced separate United States trade restrictions since May 2019. [8] If Chinese technology companies cannot freely buy the most advanced foreign chips, they can seek more computing power by joining larger numbers of locally available chips. [7] This approach only works well if the chips can exchange information quickly enough. [7]

Eric Xu explicitly linked Huawei's chip push to China's effort to reduce its reliance on foreign technology. [8] Chinese companies and policymakers have placed greater emphasis on building domestic alternatives to foreign hardware. [20]

Huawei’s accelerated roadmap signals a strategic shift toward system-level architecture to counter individual chip performance gaps. By focusing on UnifiedBus and massive SuperPod clusters, Huawei is attempting to match the computing capacity of systems utilizing advanced Nvidia chips. While Nvidia dominates the global market, Huawei is steadily locking in domestic hyperscalers like DeepSeek who face strict hardware embargoes. The sheer scale of Huawei's proposed million-processor superclusters indicates an ambition to run the world's most compute-hungry frontier models entirely on Chinese silicon. What the sources don't address: whether Huawei's manufacturing partners can achieve sufficient production yields to fulfill the massive domestic hardware demand they are currently failing to satisfy.