Key points
- Huawei openly acknowledges its silicon’s inferiority to Nvidia’s in raw power.
- Huawei unveils a three-year plan to challenge Nvidia’s AI market leadership, emphasizing scale and networking.
- The company showcased its “SuperPod” design, capable of linking thousands of its Ascend AI chips.
- Huawei claims significantly faster data transmission speeds between chips compared to Nvidia’s technology.
Huawei, China’s tech giant, has publicly declared its ambition to challenge Nvidia’s dominance in the burgeoning artificial intelligence (AI) market. This bold move, marked by an unusually high-profile announcement at its annual Huawei Connect conference, signals a shift in the company’s typically reserved approach.
Acknowledging the performance gap between its own silicon and Nvidia’s superior chips, Huawei is betting on its strengths in networking and sheer scale to close the gap. Rotating Chairman Eric Xu detailed a three-year vision, emphasizing a “brute force” strategy to overcome technological limitations imposed by US sanctions.
The company’s strategy centers around its “SuperPod” design, a data center platform inspired by Nvidia’s own technology. This system allows for the connection of up to 15,488 Ascend-branded AI chips using Huawei’s newly unveiled UnifiedBus interconnect protocol.
This protocol boasts significantly faster data transmission speeds than Nvidia’s current and upcoming technologies, potentially offsetting the performance deficit of individual chips. Huawei’s plan involves not only connecting thousands of chips but also enhancing the internal architecture to facilitate faster data transfer within each chip, thereby improving overall processing power through sheer scale.
This announcement comes amidst a wave of revelations regarding advancements in AI chips from other Chinese firms, a notable departure from their previous secrecy. This coordinated effort aligns with Beijing’s growing emphasis on developing domestic chip technology and reducing its reliance on foreign suppliers, such as Nvidia, reflecting a broader strategic goal of technological self-reliance.
The timing, just before a high-level phone call between Presidents Trump and Xi, further emphasizes the geopolitical significance of this technological rivalry.
While acknowledging the performance disparity between its chips and those of Nvidia, Huawei remains confident in its approach. Analysts, however, express reservations, noting the uncertain scalability of Huawei’s ambitious plans and the continued reliance on less advanced manufacturing processes compared to Nvidia’s access to TSMC’s cutting-edge technology.
The long-term success of Huawei’s strategy hinges on its ability to overcome these manufacturing bottlenecks and efficiently produce its advanced chips at scale. Nevertheless, Huawei’s assertive stance signals a significant challenge to Nvidia’s dominance in the AI arena.
