Unprecedented shifts in the semiconductor landscape are unfolding as the U.S. implements stringent export controls on advanced technologies, significantly impacting China’s ambitions in artificial intelligence development. At the heart of this transformation is Huawei, where efforts to bridge the technology gap are underway as Beijing pivots towards domestic alternatives amid growing restrictions from Washington.
In an analysis of the current competition between global tech giants, Dylan Patel, founder and CEO of SemiAnalysis, remarked on the progress made by Huawei, noting that the divide between Huawei’s capabilities and those of Nvidia’s export-restricted products is narrowing. “Compared to Nvidia’s export-restricted chips, the performance gap between Huawei and the H20 is less than a full generation,” Patel noted, indicating that while Huawei’s advancements are notable, substantial hurdles remain in achieving parity with industry leaders.
The escalation of tech rivalry began with the U.S. imposing restrictions not only on chip sales to China but also on the technology essential for developing domestic AI semiconductor capabilities. These limitations encompass a wide array of segments within the semiconductor value chain, from design philosophies to manufacturing processes critical for fabricating high-performance chips. Experts concur that while Beijing has mobilized an impressive financial commitment to bolster its domestic chip industry—reportedly amounting to tens of billions—it faces formidable challenges in executing a comprehensive strategy to establish a resilient AI chip ecosystem.
Paul Triolo, a partner at DGA-Albright Stonebridge Group, underscores that U.S. export controls have paradoxically motivated Chinese companies to innovate while simultaneously impeding their ability to catch up effectively. “The restrictions have incentivized China’s industry to develop alternatives, but also present challenges that slow down the process,” Triolo explained. As firms within China such as Enflame Technology and Biren Technology enter the AI processor landscape, all eyes are on Huawei’s HiSilicon division, which is spearheading efforts to deliver competitive solutions in the global market.
Indeed, Huawei’s Ascend 910B GPU, currently the most advanced chip in mass production by the company, signifies its technological aspirations. Reports indicate that the anticipated Ascend 910C could commence mass shipments shortly; however, updates regarding its debut have yet to materialize. Patel’s analysis situates Huawei’s progress against historical metrics—while Huawei’s chips lagged two years behind Nvidia a year ago, the Ascend 910C now positions Huawei as only one year behind. This trajectory reflects notable strides in chip design capabilities, yet design alone is not enough to foster a full-fledged AI chip ecosystem.
Central to the discussion of AI chip production is fabrication, which remains a critical flaw in Huawei’s strategy. Nvidia relies on TSMC, the world’s leading semiconductor contract manufacturer, renowned for its capacity to produce advanced chips with cutting-edge technology. However, TSMC is bound by U.S. policies and has been unable to fulfill orders from companies like Huawei, which has been placed on a U.S. trade blacklist since 2019.
Consequently, semiconductor firms like Huawei are compelled to pivot towards domestic suppliers. The most prominent of these is SMIC—an entity that lags significantly behind TSMC in technology prowess and is currently capable of fabricating chips at a 7-nanometer scale, whereas TSMC has achieved processes down to 3 nanometers. Despite advances, such as purported developments in 5-nanometer technology for Huawei’s Mate 60 Pro, experts like Ray Wang illustrate that SMIC’s operational capacity still pales in comparison to TSMC’s.
A further complication arises from the procurement of advanced chipmaking equipment, primarily sourced from ASML, a Dutch company that is the sole provider of cutting-edge lithography machines. Under U.S. export regulations, ASML is restricted from exporting its most sophisticated ultraviolet (EUV) lithography systems to China, presenting a significant barrier in advancing chip production capabilities.
Without access to these essential EUV lithography tools, SMIC’s output remains stunted, forcing the company to utilize older deep ultraviolet lithography solutions. Although this has allowed for some degree of production, the yields have not met market demand thoroughly, leading to concerns about the sustainability of this workaround. Analysts have noted that attempts to develop competitive lithography technology domestically could take a decade or more, with a focus on alternative methodologies that transcend mere imitation.
Equally important in the AI chip production ecosystem is the role of memory components, particularly high-bandwidth memory (HBM)—a crucial element for data processing in AI applications. South Korean companies, including SK Hynix and Samsung, dominate the HBM market, yet recent U.S. restrictions have hindered Chinese firms’ access to these essential components. ChangXin Memory Technologies, in collaboration with Tongfu Microelectronics, is reportedly in the early phases of developing HBM solutions, but projections suggest that they remain several years behind their global counterparts.
While Huawei has managed to leverage existing HBM inventories to support the development of its Ascend 910C processor, reliance on foreign suppliers for critical components remains a precarious position for the company. Reports from SemiAnalysis highlight the significant dependence on international technology and resources for assembling AI solutions, revealing gaps that Chinese firms must address to establish true autonomy in the semiconductor arena.
Ultimately, as the landscape of global semiconductors continues to evolve under the weight of geopolitical tensions and technological competition, the implications for market dynamics, national security, and economic strategies are profound. The ability of Chinese companies, particularly leaders like Huawei, to navigate these multifaceted challenges will not only define their future prospects but also shape the contours of innovation and competitiveness in the broader tech industry. As Beijing’s efforts to bolster its domestic chip ecosystem unfold, the world will be watching closely how these developments transform the landscape of artificial intelligence and beyond.