Huawei has unveiled new AI infrastructure designed to enhance computing power and strengthen its position in the competitive AI chip market, as it continues to navigate restrictions on access to Nvidia hardware in China. At its Huawei Connect conference in Shenzhen, the tech giant introduced SuperPoD Interconnect technology, a system capable of linking up to 15,000 graphics cards—primarily Huawei’s own Ascend AI chips—to deliver massive parallel processing capabilities. The new interconnect technology is positioned as a direct alternative to Nvidia’s NVLink, the high-speed communication framework that enables efficient data transfer between AI chips. By enabling large-scale clustering of its Ascend chips, Huawei aims to provide users with significantly increased compute power, which is essential for training and scaling advanced AI models. While Huawei’s Ascend chips currently lag behind Nvidia’s leading GPUs in raw performance, the ability to connect thousands of them into a unified system could help close the gap in large-scale AI workloads. This development is particularly significant given the recent regulatory shift in China, where domestic tech firms were banned from purchasing Nvidia’s hardware—including the RTX Pro 600D servers, which were specifically tailored for the Chinese market. The move underscores Huawei’s broader strategy to build a self-reliant AI ecosystem amid ongoing U.S. sanctions and export controls that have limited its access to advanced semiconductor technologies. By developing its own infrastructure stack, Huawei is working to reduce dependency on foreign components and strengthen its domestic AI supply chain. The company has not yet disclosed specific performance benchmarks for the SuperPoD system, but industry observers see the technology as a key step toward enabling China’s AI industry to operate independently of Western chipmakers. Huawei has also been expanding its AI software and cloud services to complement its hardware offerings. TechCrunch reached out to Huawei for further details on the SuperPoD technology and its deployment roadmap.