Huawei's Clustering Strategy: A Powerful Response to US Chip Dominance
Huawei's founder discusses the robustness of his company's in-house chip technology.
In the midst of the ongoing tussle between China and the US over trade, Huawei's founder, Ren Zhengfei, has playfully dismissed the allegations of his company's chip technology being on par with that of the US. In an interview with the state-run "People's Daily," Ren candidly admitted that Huawei's individual chips are still a "generation" behind their American counterparts.
However, he defended Huawei's standing in the competitive chip market, asserting that China boasts numerous powerful chip companies, with Huawei being one of them.
American chipmaker Nvidia has expressed concerns over US restrictions on shipments to China, viewing Huawei as a formidable competitor as a result. Moreover, due to US sanctions, Chinese companies have limited access to advanced AI chips from Nvidia, giving Huawei room to seize market share.
Clustering: The Secret Ingredient
As the trade talks between the US and China unfolded in London, the interview with Ren was published. Both delegations had been engaged in discussions to remove trade barriers since Monday. A contentious issue in these negotiations has been China's export controls on rare earths, with the US seeking relaxation. In exchange, the US might consider easing its trade restrictions in sectors such as computer chips or aircraft components, where China heavily depends on foreign technology.
Ren revealed that Huawei is devising a strategy called "clustering" to address technical deficiencies. This tactic involves combining multiple chips to create a powerful computing system, offsetting the shortcomings of individual chips.
Delving Deeper into Huawei's Clustering Strategy
Huawei aggregates multiple Ascend AI chips to create large-scale, high-performance computing systems. For instance, the AI CloudMatrix 384 system launched in April 2025 links 384 Ascend 910C chips, enabling the company to achieve competitive AI processing power, outperforming Nvidia’s GB200 NVL72 system on certain benchmarks, despite individual chip limitations.
Moreover, Huawei has developed a "compound chip architecture," with two Ascend 910B chips combined into a single 910C module. This ingenious design boosts performance while avoiding US restrictions on advanced fabrication tools.
In addition to systemic innovations, Huawei's MindSpore AI framework supports major AI libraries like TensorFlow and PyTorch, furthering the appeal of its chips among Chinese enterprises and creating a self-reinforcing ecosystem.
The Philosophical Shift at Huawei
Ren Zhengfei has championed a more mathematical and theoretical focus at Huawei. The company seeks to compensate for physical chip limitations by using "mathematics to supplement physics" and employing "non-Moore's law to supplement Moore’s law." This innovative approach recognizes that architecture and algorithm improvements can aid in bridging the technology gap.
Huawei invests billions annually in research, with a significant portion dedicated to theoretical research. This forward-thinking strategy aims to build foundational capabilities rather than replicating existing technologies.
Trade-Offs and Limitations
Cluster computing can lead to increased administrative and energy overheads, and may not always produce the same efficiency as the latest single-chip solutions. Furthermore, Huawei's chip yields remain lower than industry standards, with its 7nm manufacturing process slower than TSMC's 5nm processes.
Conclusion
Huawei's clustering strategy enables the company to stay competitive in AI and advanced computing by optimizing the performance of available hardware through innovative aggregation, compound chip architectures, and a robust software ecosystem. While this strategy doesn't completely eradicate the technology gap with US leaders, it demonstrates resilience and adaptability in the face of ongoing trade restrictions and geopolitical tensions.
- Despite Huawei's individual chips being a "generation" behind those of American counterparts, the company's clustering strategy involving the combination of multiple chips creates a powerful computing system that can outperform Nvidia's GB200 NVL72 system on certain benchmarks.
- Huawei's finance department, in response to US restrictions on advanced AI chips, has invested billions annually in research, with a significant portion dedicated to theoretical research, aligning with Ren Zhengfei's philosophical shift towards a more mathematical and theoretical focus at Huawei, as he seeks to build foundational capabilities rather than replicating existing technologies.
