At Huawei Connect reveals Beijing's national champion Bold Ai road map to tackle Nvidia

Huawei Technologies Co. revealed an exceptionally detailed vision for how it intends to at Nvidia Corp. gearing overwhelming lead in artificial intelligence hardware, which combines the scale with China’s state-supported driving force for technological independence. The company’s rotating chairman, Eric Xu, on Thursday has a road map to counter US opponents at its annual Huawei Connect conference in Shenzhen, a three-year roadmap designed to counteract Nvidia’s dominance in the global AI race. Huawei acknowledged that its silicon could not match the US firm in raw computer power or efficiency. Instead, it aims to compensate by utilizing the traditional strengths: large -scale grouping, advanced network and firm government support. The announcement stood out for his openness. Huawei historically mastered his hardware strategy, often forcing the analysts of the industry to break down his smartphones to determine the underlying chip technology. The firm’s latest revelations were thus a rare shift to transparency, which coincides with delicate high -level conversations between Washington and Beijing. The presentation of Huawei’s tradition of secrecy Xu’s presentation came with the kind of high drama typically associated with an Nvidia launch. He unveiled the company’s next generation of processors branded by the upgraded “Superpod” data center designs-a direct echo of Nvidia’s own terminology for large-scale computer platforms. Huawei’s new “UnifiedBus” interconnect protocol was also formally introduced, which enabled the theoretical connection of up to 15,488 Ascend chips in a single system. The strategy amounts to overwhelming competitors through numbers. Huawei insists that its self-designed interconnection system can transfer up to 62 times faster from chip-to-chip data transfers, and the upcoming NVLink144, which dwarfs the capacity of the current NVLink72 platform connecting the US company’s Blackwell GPUs and Grace CPUs. Bernstein research analysts described Huawei’s willingness to articulate his AI plans in public as a sign of confidence that it ensured reliable manufacturing capacity within China, despite being from the Taiwan Semiconductor Manufacturing Co. (TSMC) has been cut off, the world’s most advanced watering cans. “These developments indicate that Huawei has built a local semiconductor ecosystem, strong enough to withstand the supply chain failure,” Bernstein’s Qingyuan Lin wrote. Beijing pushes indigenous innovation. The daring roadmap comes because other major Chinese technological players – including Alibaba Group and Baidu – have begun to reveal the progress in designing the AI ​​chip. Such announcements have been scarce for years, as firms try to avoid the investigation into Washington. But with Chip Policy now central to US -China negotiations, Beijing has called on companies to accelerate indigenous innovation and reduce the dependence on foreign suppliers. US restrictions have already cut Huawei from advanced lithography instruments and leading memory vendors such as Sk Hynix. Nevertheless, Xu argued that Huawei could close the performance gap by connecting massive numbers of his processors to ‘Super Clusters’. The company says it has developed a self-designed memory architecture with a high bandwidth to allow chips to exchange data at unprecedented speeds, which mitigate some of the limitations of less advanced process technology. By 2028, Huawei expects its Ascend 970 chip to have a 4 terabit-per-second-interconnection speed, more than double the 1.8 TBPS NVIDIA currently offers. Xu emphasized that this skill in the network uses Huawei’s long history of advanced communication equipment for the telecommunications industry. Yet there is still significant doubt. Jefferies analysts warned that Huawei’s earlier plan to release the Ascend 910D on 5 nanometer has stopped due to poor production returns, which underline the challenges posed by limited access to advanced chipmaking equipment. “A lack of modern manufacturing tools is still the biggest obstacle in China to reduce the dependence on Nvidia,” said the bank’s Edison Lee. Performance comparison also reveals the extent of the challenge. Analysts estimate that Huawei’s upcoming Ascend 950 delivers only about 6 percent of the computer power expected from Nvidia’s next generation VR200 Superchip. Nvidia, meanwhile, still sets the pace of the industry, with even traditional opponents AMD and Intel struggling to keep up. However, clusters as China’s way forward, Huawei, insist that it can bridge the separation through scale. Xu claimed that clusters of up to a million chips can theoretically deliver computer power on an equal footing with Western opponents. He has positioned this ‘brute power’ approach as the only viable route for China to overcome restrictions in process technology. “We believe that by trusting” Superpod “and Cluster Technology, we can reach a breakthrough in the restrictions we face and provide endless computer support for our country’s AI development,” Xu told State Media last week. The Chinese government has put the semiconductor industry at the heart of its industrial policy. President Xi Jinping met with leading technology entrepreneurs, including Huawei founder Ren Zhengfei, to emphasize the need for breakthroughs in strategic sectors. Officials also promised to mobilize national resources to accelerate progress while grown after purchasing Chinese firms of certain Nvidia components. Huawei’s ambitions emphasize the shifting dynamics of the AI ​​hardware market. While Nvidia remains the undisputed leader, the combination of political pressure, government subsidies and technical ingenuity is capable of reducing Chinese businesses. The unveiled by Huawei’s roadmap, given the penchant for the secrecy of the company underline both the determination and the confidence of Beijing in a home-grown way forward. Still, a lot depends on whether Huawei can produce these chips on scale. The surprise exemption of a 7-nanometer processor for the Mate 60 Pro smartphone in 2023 showed technical ingenuity, but also exposed the boundaries of Chinese manufacturing, which is not moving beyond the node. Without further breakthroughs in lithography and manufacturing, Huawei’s AI ambitions can remain limited. For the time being, the company leans on its historical strength in networks and large -scale system integration. Whether its ‘Superpod’ groups can compensate for poorer single-chip performance will be tested in the coming years. Meanwhile, the high-profile unveiling indicates a clear message: Huawei intends to position himself as China’s national champion in the AI ​​race, regardless of the obstacles placed in his way. (With input from Bloomberg)