Huawei Technologies plans to release two new AI accelerators in the Ascend series next year. According to David Wang, the rotating chairman of the board, the Ascend 960T model is scheduled for launch in the first quarter of 2027, followed by the Ascend 960PR in the third quarter.
The company is focusing on scaling computing power through clusters using its proprietary UnifiedBus interface. Currently, Huawei offers 11 components supporting this interface. Large-scale superclusters can integrate up to 1 million AI chips, while over 1,000 smaller supernodes have already been shipped to more than 370 customers.
Huawei relies on cluster scaling to bypass performance limitations caused by Western sanctions. The software ecosystem for these accelerators is supported by 5,270 active monthly developers. Due to HBM memory shortages, the company has increased prices for Ascend accelerators by up to 60 percent.
Beyond the domestic Chinese market, Huawei intends to supply its AI hardware to customers in South Korea and Egypt. The company has not disclosed specific technical specifications for the upcoming chips or the identities of its clients.
