Huawei has unveiled a new artificial intelligence infrastructure designed to expand its computing power and position the company as a viable alternative to Nvidia, the U.S. chipmaker whose hardware has dominated the sector.
At the Huawei Connect 2025 conference in Shenzhen on Thursday, the company introduced the SuperPoD Interconnect, a system capable of linking up to 15,000 graphics cards, including its in-house Ascend AI processors. The technology mirrors Nvidia’s NVLink system, which enables ultra-fast communication between GPUs to handle large-scale AI training.
By clustering thousands of chips, Huawei aims to overcome the performance gap between its Ascend units and Nvidia’s highly sought-after processors. “This breakthrough means our customers will no longer be limited by individual chip performance,” a Huawei engineer told Chinese state media. “They can scale up to supercomputer levels of compute power.”
The announcement comes just 24 hours after Beijing formally barred domestic companies from purchasing Nvidia’s latest offerings, including the RTX Pro 600D servers built specifically for China. The ban reflects escalating U.S.–China tech tensions and increases the strategic importance of Huawei’s local solutions.
Analysts say the timing is no coincidence. “Huawei is stepping into the vacuum created by U.S. export restrictions,” said Kevin Xu, a technology researcher tracking semiconductor trends in Asia. “If they can deliver scalable AI compute, they stand to capture market share that would otherwise go to Nvidia.”
Although Huawei still lags in raw chip performance, clustering thousands of processors may provide enough computing power for Chinese firms to keep pace in areas such as generative AI and autonomous systems. For Beijing, the project is as much about technological self-reliance as it is about commercial competition.
