Huawei plans Q1 2027 launch of new AI chip as it takes on Nvidia

Huawei said at its Huawei Connect conference that it has moved up the launch of its next-generation Ascend 960DT AI chip to the first quarter of 2027, earlier than a previously planned third-quarter 2027 release. The company reiterated plans to scale AI systems using its Peerium Computing Architecture and highlighted existing Atlas systems that link large numbers of accelerator cards.

By AI Newsroom· Reviewed by Pranav, Founder & Editor-in-ChiefPublished about 11 hours agoUpdated about 2 hours ago0 views
Huawei plans Q1 2027 launch of new AI chip as it takes on Nvidia

Why It Matters

The accelerated timeline signals Huawei's push to compete with dominant AI-chip suppliers such as Nvidia and to advance China's domestic AI hardware capabilities amid ongoing U.S. export restrictions. Progress or setbacks in Huawei's chip and system scale-up could affect the broader global competition over high-performance AI infrastructure.

Key Facts

  • New launch timing: Ascend 960DT expected in Q1 2027
  • Previous timetable: Previously planned for Q3 2027
  • Announcement venue: Huawei Connect conference
  • Company spokespeople: David Wang announced the update; Huawei spokesperson confirmed to TechCrunch
  • Architecture name: Peerium Computing Architecture

Huawei told attendees at its Huawei Connect event that it will bring forward the debut of its next-generation Ascend 960DT AI chip to the first quarter of 2027, rather than the third quarter of 2027 the company had previously signaled. A Huawei spokesperson confirmed the revised schedule to TechCrunch, and the updated timeline was announced on stage by David Wang, Huawei's rotating and acting chairman. The firm said the Ascend 960 series is arriving ahead of schedule and described year-over-year performance improvements. Huawei is using what it calls the Peerium Computing Architecture to link processors, memory, storage and networking via its UnifiedBus technology, with the goal of creating much larger AI computing systems for both training and inference workloads. Huawei already markets systems built on that architecture, including the Atlas 950 SuperPoD and SuperCluster. The company has said an Atlas 950 SuperCluster can connect as many as 256,000 accelerator cards. However, an analyst flagged a discrepancy between earlier scaling claims and this week's announcement: Huawei had previously indicated an Atlas 960 SuperPoD could scale to 15,488 Ascend 960 chips, while current materials referenced a SuperPoD with 4,096 chips. The accelerated chip timeline comes against a backdrop of U.S. restrictions on China's access to advanced semiconductor technology and heightened geopolitical attention. The announcement arrived a week before a scheduled meeting between U.S. President Donald Trump and Chinese President Xi Jinping on September 24 in Washington, DC. Industry observers cited by the source said the restrictions are unlikely to stop China from pursuing indigenous semiconductor development, and Huawei executives have framed faster domestic AI progress as necessary to catch up with the U.S.

Keep Reading