to leave a comment.
More than 1,000 SuperPods already deployed... "New product performance is 3-4 times that of existing products" also observed
Chinese telecommunications equipment maker Huawei announced that it would bring forward the release of its next-generation AI chips earlier than originally planned, citing the computing field as the core of its future artificial intelligence (AI) strategy.
According to reports including the Gwanchangpan Daily and Bloomberg on the 17th (local time), Huawei's Rotating Chairman Wang Tao announced at the '2026 Huawei Connect Conference' held in Shanghai that "the Ascend 960 chip has been prepared ahead of schedule and achieved double the performance (compared to existing products)."
He further explained that the launch plan for Ascend 960, originally scheduled for the fourth quarter of next year, has been moved up, with the 960DT for AI model training to be released in the first quarter of next year and the 960PR for inference in the third quarter of next year, respectively.
Ascend 970 is scheduled to be announced in 2028, and Ascend 980 in 2029.
Previously, at this event last year, Huawei unveiled its AI chip development blueprint for the next three years, stating that it would "release new Ascend products once a year and double their computing power with each release."
Bloomberg assessed that this announcement reflects Huawei's confidence, noting that Huawei has been increasing disclosures related to data center hardware development since last year. This is a departure from its past practice of maintaining secrecy about production plans.
China views US sanctions in the advanced semiconductor sector as a chokehold on its nation and emphasizes its determination to resolve it through self-reliance and self-strengthening, with Huawei seen as the vanguard of this effort.
This year, Huawei introduced the 'Tau Law' to overcome the limitations of 'Moore's Law' and is focusing on the AI computing supernode system 'Atlas 950 SuperPod' to overcome the performance limits of individual chips by linking them together.
This technology links thousands of AI chips into a single computing pool. China, while at a disadvantage in individual chip aspects due to US restrictions, plans to overcome this through its superiority in chip-linking systems.
Wang, the Rotating Chairman, stated today, "More than 1,000 Atlas SuperPods have already been deployed, and there are over 370 customers. The Atlas 950 SuperPod has already begun large-scale commercialization."
He added, "The new Atlas 960 SuperPod will be the first in the industry to use Near-Package Optics (NPO), a technology that enhances data transfer speed and energy efficiency. Among these, the Atlas 960 liquid-cooled SuperPod is planned for release in the third quarter of next year."
While the existing Atlas 950 SuperPod connects up to 8,192 Neural Processing Units (NPUs), the Atlas 960 SuperPod can connect 15,488 chips, and its training and inference performance are estimated to have improved by approximately 3 and 4 times, respectively, according to the South China Morning Post (SCMP).
Wang, the Rotating Chairman, viewed 'Unified Bus' technology, which makes information sharing between chips more efficient, as a key link in next-generation large-scale AI systems.
He also stated, "The core of Huawei's AI strategy is computing," and "through 'SuperPod Plus (+) clusters,' we will establish China's robust computing power foundation and create new options for the world."
Previously, Guo Ping, Chairman of Huawei's Supervisory Board, the company's highest oversight body, also stated at a recent new employee seminar that Huawei is focusing on connectivity and computing, adding, "Huawei's goal in this area is to become Nvidia."
Newsletter
Get key news delivered to your email every morning
to leave a comment.