TLDR
- Huawei unveiled a three-year plan to challenge Nvidia’s AI chip dominance through clustering up to 15,488 Ascend processors
- The company claims its UnifiedBus interconnect can move data 62 times faster than Nvidia’s upcoming NVLink144 technology
- Huawei’s approach relies on “brute force” clustering rather than superior individual chip performance
- The announcement marks a shift from Huawei’s typically secretive approach since US sanctions began in 2020
- Individual Huawei chips still lag behind Nvidia, with next-gen Ascend 950 offering only 6% of Nvidia’s VR200 performance
Huawei Technologies has broken from its usual secretive approach to publicly announce an ambitious three-year strategy aimed at competing with Nvidia in the artificial intelligence chip market. The Chinese tech giant made this announcement at its annual Huawei Connect conference, where rotating Chairman Eric Xu outlined detailed plans to erode Nvidia’s market dominance.
The timing of Huawei’s announcement is important, coming just one day before a phone meeting between US President Donald Trump and Chinese leader Xi Jinping. This marks a departure from Huawei’s typically subdued product introductions since losing access to Taiwan Semiconductor Manufacturing Co. in 2020 due to US restrictions.
Huawei openly acknowledges that its individual chips cannot match Nvidia’s processors in raw power and speed. Instead, the company plans to rely on what it calls “brute force” – connecting massive numbers of chips together to compete with Nvidia’s performance levels.
The centerpiece of Huawei’s strategy involves clustering up to 15,488 of its Ascend-branded AI chips using a new UnifiedBus interconnect protocol. This approach allows the company to link thousands of processors into what it calls “SuperPod” systems, borrowing terminology from Nvidia’s own playbook.
Interconnect Speed Claims
Huawei claims its UnifiedBus technology can transmit data up to 62 times faster than Nvidia’s upcoming NVLink144 technology. The company says this faster data transmission between chips helps overcome the individual performance gap between its processors and Nvidia’s offerings.
By comparison, Nvidia’s current NVLink72 technology allows the company to connect 72 Blackwell graphic processing units and 36 Grace central processing units together. Huawei’s design aims to scale much larger, potentially connecting up to one million chips in what it calls “super clusters.”
The company also introduced self-designed high-bandwidth memory architecture to improve data transfer between different components within processors. This development comes despite US restrictions that severed Huawei’s ties to memory industry leaders like SK Hynix.
Huawei outlined future chip generations including the Ascend 950, 960, and 970 processors. The Ascend 970, planned for 2028, will reportedly feature a 4 terabit-per-second interconnect speed compared to Nvidia’s current 1.8 terabit-per-second offering.
Performance Gap Remains
Despite these ambitious plans, analysts point out that individual Huawei chips still lag far behind Nvidia’s offerings. Bernstein analysts noted that a single next-generation Ascend 950 can deliver only 6% of the performance of Nvidia’s upcoming VR200 superchip.
Jefferies analysts expressed skepticism about Huawei’s manufacturing capabilities, noting that the company’s plan to roll out Ascend 910D using 5-nanometer technology has not materialized due to poor production yields. The lack of advanced chipmaking equipment remains China’s biggest obstacle in reducing dependence on Nvidia chips.
Huawei has not progressed beyond 7-nanometer chip designs since introducing them in 2023 with the Mate 60 Pro smartphone. Xu did not elaborate on how the company plans to manufacture its new chips, which analysts consider a potential bottleneck.
The announcement reflects broader tensions between the US and China over semiconductor technology. Washington has implemented various restrictions to limit China’s access to advanced chip technology, while Beijing has encouraged domestic companies to develop alternatives to foreign suppliers.
Market Response
Nvidia shares declined 1.00% in premarket trading following reports of Huawei’s announcement. This came after Nvidia had gained on Monday following news of a strategic partnership with OpenAI worth up to $100 billion for AI infrastructure development.
The Chinese government has recently launched moves targeting Nvidia, including directives to stop Chinese firms from buying certain Nvidia components. Senior Chinese officials have pledged to use national resources to accelerate domestic technology breakthroughs.
Huawei’s public roadmap represents a shift toward more open communication about its AI ambitions, contrasting with years of keeping cutting-edge technology secret to avoid US scrutiny. The company plans to begin implementing its SuperPod technology as part of its three-year timeline to challenge Nvidia’s market position.
Stay Ahead of the Market with Benzinga Pro!
Want to trade like a pro? Benzinga Pro gives you the edge you need in today's fast-paced markets. Get real-time news, exclusive insights, and powerful tools trusted by professional traders:
- Breaking market-moving stories before they hit mainstream media
- Live audio squawk for hands-free market updates
- Advanced stock scanner to spot promising trades
- Expert trade ideas and on-demand support
