China Unveils CPU-Only Exascale Supercomputer Lingshen, Claiming 2 Exaflops Without Foreign Components

China's National Supercomputing Center in Shenzhen has revealed plans for the Lingshen supercomputer, a system targeting sustained performance above 2 ExaFLOPS using exclusively CPU processing power and no foreign-made components. The announcement, made at a conference on April 24, outlines a system that would pack 47,000 processors into 92 compute cabinets, positioning itself as the first exascale machine designed to reach that performance tier without GPU accelerators.

Technical Architecture and Design

Lu Yutong, director of the Shenzhen supercomputing center and the system's chief designer, presented the technical details at the event. The Lingshen project is structured in two phases: a pilot verification system and a full production deployment. The pilot phase will utilize 100 Huawei Kunpeng servers built on Arm-based Taishan cores, totaling 12,800 cores with a theoretical peak performance above 10 petaflops.

The full production system represents a significant scaling effort, designed to incorporate 1,580 blade servers using x86 CPUs with 101,120 cores. Additional compute capacity comes from 16 four-way servers contributing another 2,048 cores, and four eight-way servers adding 1,280 cores, according to machine translations of Chinese-language press materials.

The system's infrastructure includes 36 network cabinets supporting a million-port interconnect, 650PB of planned storage across 428 nodes, and 67 liquid-cooled storage cabinets capable of handling 10 TB/s of bandwidth. This comprehensive architecture suggests a focus on both computational power and data handling capabilities.

Market Context and Competitive Position

Lingsheng's CPU-only architecture represents a fundamentally different approach to reaching exascale throughput compared to existing global systems. Every other exascale system currently in operation relies heavily on GPUs or specialized accelerator hardware. The U.S. Department of Energy's El Capitan, currently the world's fastest supercomputer, runs on 44,544 AMD MI300A APUs that tightly couple CPU and GPU silicon on a single package.

China's claimed sustained performance of 2+ ExaFLOPS would, if achieved, exceed El Capitan's measured Linpack score of 1.809 ExaFLOPS. While El Capitan's theoretical peak reaches 2.79 ExaFLOPS, real-world Linpack results are consistently lower than theoretical maximums. The absence of any benchmark data for Lingsheng, as the system hasn't been constructed yet, casts some uncertainty over these claims.

Supply Chain Implications

Perhaps most significantly, China emphasizes that Lingsheng will have no reliance on outside vendors for its components. This claim comes amid escalating technological tensions between China and Western nations, particularly the United States. While the announcement specifies that the production system will use x86 CPUs, China's domestic x86 options are limited to Zhaoxin, a joint venture between VIA Technologies and the Shanghai municipal government, and Hygon, which originally licensed AMD's Zen architecture but lost access to updated designs following U.S. export restrictions.

Neither company has demonstrated processors competitive with current-generation offerings from Intel or AMD. The center and its partners haven't named specific suppliers or provided an operational timeline for the completed system, leaving questions about the actual feasibility of their performance claims and component availability.

Strategic Significance

The Lingshen project emerges as part of China's broader technological self-sufficiency initiative, particularly in high-performance computing. By developing a CPU-only exascale system, China appears to be hedging against potential supply chain disruptions while pursuing an alternative architectural path to extreme computing performance.

This approach contrasts with the global trend of heterogeneous computing that combines CPUs with GPUs or other accelerators. While GPUs excel at parallel processing workloads common in scientific computing, they also represent points of potential technological dependency for China given the dominance of U.S. companies like NVIDIA and AMD in this space.

The project's success would not only demonstrate China's growing technical capabilities but also potentially influence future supercomputer design philosophies worldwide. If China can achieve exascale performance with CPUs alone, it could challenge the prevailing assumption that accelerators are essential for reaching the highest performance tiers.

As with any ambitious未完成项目 (unfinished project), the actual performance and capabilities of Lingshen will only become clear once the system is operational and undergoes independent benchmarking. The timeline for completion remains unclear, with only the pilot phase currently confirmed. Nevertheless, the announcement signals China's continued commitment to advancing its position in the global supercomputing race while reducing dependencies on foreign technology.