Huawei moves Ascend 960DT AI chip launch to first quarter of 2027
Huawei brings forward Ascend 960DT schedule
Huawei plans to release its next-generation Ascend 960DT AI chip in the first quarter of 2027, advancing a launch that had been scheduled for the third quarter of that year. David Wang, Huawei’s rotating and acting chairman, announced the revised timetable at the company’s Huawei Connect conference.
A Huawei spokesperson said the Ascend 960 generation is launching ahead of schedule, will double performance and will advance year by year. The decision places the Chinese technology group’s chip programme in sharper competition with Nvidia as it develops domestic AI computing capabilities.
Architecture is central to the larger computing plan
The chip announcement sits within Huawei’s effort to combine hundreds of thousands, and eventually millions, of AI chips into what it describes as one giant computer. The company calls this approach the Peerium Computing Architecture.
Peerium relies on UnifiedBus, Huawei’s technology for connecting processors to memory, storage and networking hardware. Eric Xu, Huawei’s rotating chairman, said the architecture is being used to build larger AI computing systems for both model training and inference.
Huawei identified the Atlas 950 SuperPoD and SuperCluster as the first systems built on the architecture. It says an Atlas 950 SuperCluster can connect up to 256,000 accelerator cards, illustrating that its AI strategy includes the systems and interconnect layer as well as the processor itself.
System scale remains a key detail
The accelerated chip date also raises questions about the capacity of the systems Huawei is setting out around the new generation. China technology analyst Rui Ma noted that Huawei had previously said its Atlas 960 SuperPoD would scale to 15,488 Ascend 960 chips, whereas this week’s announcement referred to a 4,096-chip system.
Ma described the chip as arriving much earlier while characterising the announced SuperPoD as smaller than the earlier plan. The distinction matters because processor availability and the scale at which those processors can be connected are separate measures for organisations assessing AI infrastructure.
Huawei continues to pursue semiconductor development despite US restrictions on China’s access to advanced semiconductor technology. Ma argued that the stakes of self-sufficiency make continued Chinese development likely. For businesses tracking AI capacity, Huawei’s updated roadmap makes it important to evaluate delivery dates, performance claims, interconnect design and stated cluster size together before making infrastructure assumptions.