During its Huawei Connect conference on Thursday, Huawei announced that it has moved up the release date for its next-generation Ascend 960DT artificial intelligence chip. The new timeline places the launch in the first quarter of 2027, a significant advance from the previously scheduled third quarter.
David Wang, Huawei’s rotating and acting chairman, revealed the updated schedule. A company spokesperson told TechCrunch that the Ascend 960 series chips are delivering double the performance of their predecessors while adhering to this accelerated rollout.
The announcement arrives just days before a scheduled meeting between U.S. President Donald Trump and Chinese President Xi Jinping on September 24 in Washington, DC. Huawei is attempting to pool hundreds of thousands of these AI chips into a single massive computing unit through its Peerium Computing Architecture. This system utilizes UnifiedBus technology to connect processors with memory, storage, and networking hardware.
Eric Xu, another rotating chairman at Huawei, stated that the company is leveraging this architecture to construct larger AI computing systems capable of handling both training and inference tasks. The Atlas 950 SuperPoD and SuperCluster are the inaugural systems built on this framework, with the SuperCluster reportedly able to link up to 256,000 accelerator cards.
Despite the optimistic timeline, some analysts have raised questions regarding the scaling of Huawei’s broader AI systems. China technology analyst Rui Ma noted on X that while the chip itself is arriving earlier than expected, the recently announced Atlas 960 SuperPoD supports only 4,096 chips, a stark contrast to the 15,488 chips promised in earlier plans.
Ma argued that U.S. restrictions on advanced semiconductor access are unlikely to halt China’s self-sufficiency efforts, describing the stakes as too high. Meanwhile, the geopolitical context remains tense; Trump has resisted calls from AI industry leaders to slow development due to safety concerns, emphasizing the need to maintain a technological lead over China. Conversely, Huawei’s Xu has argued that Chinese firms must accelerate AI progress to catch up with the U.S., suggesting that deeper advancement is necessary to properly understand and mitigate the risks associated with powerful AI systems.
Huawei keeps accelerating timelines. Let’s see if they deliver or if this is just another PR blitz before US sanctions tighten.
Double the performance of predecessors is impressive. Hope the power efficiency holds up under heavy training loads.
With Trump and Xi meeting soon, this timing feels deliberate. China is sending a clear message about tech independence.
Scaling from 4,096 to 256,000 chips sounds great on paper, but practical engineering hurdles are massive here.
Early 2027 is a bold move. I wonder if the software stack will actually be ready to support that kind of scale.