China’s Lingsheng Supercomputer Tops Global TOP500 List
HAMBURG — China’s newly unveiled “Lingsheng” (LineShine) supercomputer has claimed the top spot on the 67th edition of the TOP500 list, the authoritative ranking of the world’s most powerful supercomputers, released on June 23 at the ISC 2026 High-Performance Computing Conference in Hamburg, Germany. With a record-breaking 2.198 exaflops of sustained double-precision performance, Lingsheng becomes the first system ever to exceed two exaflops using only central processing units — and the first Chinese supercomputer to lead the rankings since Sunway TaihuLight in 2017, according to the TOP500.org official announcement.
A Return to the Summit After Nearly a Decade
Lingsheng’s debut at No. 1 marks a significant milestone in China’s high-performance computing ambitions. The system, installed at the National Supercomputing Centre in Shenzhen (NSCS) and built by the Shenzhen Cloud Computing Center, delivers 2.198 exaflops on the High Performance Linpack (HPL) benchmark — roughly 80 percent of its 2.736 exaflops theoretical peak — and draws approximately 42.2 megawatts of power, as reported by Xinhua News Agency.
China had stopped submitting systems to the TOP500 in 2023, despite being widely reported to have fielded several exascale-class machines during that period. Lingsheng’s return to the list — and its immediate ascent to the top — signals that Beijing is once again willing to benchmark its most advanced systems on the global stage.
A Pure CPU Architecture With All-Domestic Components
What makes Lingsheng particularly noteworthy is its architectural approach. Unlike the vast majority of today’s top supercomputers — which rely on GPU accelerators from NVIDIA or AMD — Lingsheng is a pure CPU design built entirely with domestically developed components.
The system is powered by 40,960 custom LX2 processors, each featuring 304 cores based on the ARM v9 architecture running at 1.55 GHz, co-designed with Huawei. Across 20,480 nodes, the machine totals 13,789,440 cores, linked by a proprietary high-speed interconnect called LingQi that delivers 1.6 terabits per second per node, as detailed by DataCenterDynamics.
The LX2 processor is the first domestic Chinese CPU to integrate high-bandwidth memory (HBM) directly on-chip, achieving a tenfold improvement in memory bandwidth over traditional CPUs. The system runs on Kylin OS, developed by China’s National University of Defense Technology, and uses 100 percent full liquid cooling — the largest centralized liquid-cooled deployment in the country.
Lingsheng’s Performance Across Benchmarks
Beyond raw HPL performance, Lingsheng also leads the HPCG benchmark — which measures performance on data-intensive, real-world application patterns — with 22.00 HPCG-petaflops, ahead of the United States’ El Capitan at 17.41 petaflops. On the HPL-MxP mixed-precision benchmark, however, Lingsheng ranks fourth at 7.92 exaflops with a modest 3.6x speedup over its HPL score — a result consistent with its CPU-only design, as GPU-accelerated systems typically achieve much larger gains on mixed-precision workloads.
The New Global Exascale Landscape
Lingsheng’s debut reshapes the top of the supercomputing hierarchy. For the first time, five systems now exceed the exascale threshold of one exaflop, spanning Asia, North America, and Europe simultaneously:
| Rank | System | Country | HPL (Exaflop/s) |
|---|---|---|---|
| 1 | LineShine | China | 2.198 |
| 2 | El Capitan | USA | 1.809 |
| 3 | Frontier | USA | 1.353 |
| 4 | Aurora | USA | 1.012 |
| 5 | JUPITER Booster | Germany | 1.000 |
The United States still leads in total number of systems on the list with 162, while China has 30 systems with total computing power ranking fourth globally. HPE remains the dominant system integrator in the Top 10, supplying six of the ten systems.
Geopolitical Significance: Self-Reliance Under Export Restrictions
Lingsheng’s achievement carries profound geopolitical weight. The system was developed entirely under the shadow of U.S. export controls that restrict China’s access to advanced AMD and NVIDIA GPUs. By achieving world-leading performance without foreign accelerators, Lingsheng demonstrates China’s ability to pursue a fully self-reliant technological path.
Shenzhen’s Science and Technology Innovation Bureau described the system as a “landmark achievement in full-stack self-reliant high-end computing” and a milestone in breaking the “foreign technology blockade,” according to Chinese media reports. The system is part of China’s broader “dual circulation” strategy, which emphasizes technological self-sufficiency alongside global engagement.
Turing Award winner Jack Dongarra praised the system, saying that “China’s ‘Lingsheng’ system lets the world see the hope of a new system architecture for supercomputing toward AI4Science.”
Real-World Applications and Scientific Impact
Lingsheng is not merely a benchmark achievement — it has already demonstrated significant real-world capabilities across multiple scientific domains. The system has been used for global kilometer-scale weather simulation, with Tsinghua University’s Earth System Model achieving the first 30-day global forecast at 1km resolution in just two hours, at a 62 percent cost reduction compared to previous systems.
In drug discovery, researchers at Sun Yat-sen University and China Pharmaceutical University have used Lingsheng to screen 10 trillion compounds per day — a thousandfold efficiency improvement. The system has also been applied to materials science, oil exploration, and electromagnetic simulation for high-end equipment design.
What to Watch For
Lingsheng’s debut raises several questions for the future of global supercomputing. With China having fielded other exascale systems that were not submitted to TOP500 between 2023 and 2025 — including the reported Tianhe-3 and Sunway OceanLight — the full extent of China’s high-performance computing capabilities remains partially obscured. Meanwhile, the United States continues to lead in mixed-precision AI workloads, and GPU-accelerated systems now account for 55.4 percent of the TOP500 list.
As chief designer Lu Yutong noted, Lingsheng represents “six major technological breakthroughs in architecture, performance, energy consumption, programming, scalability, and reliability.” The question now is whether China can sustain this momentum — and how the global supercomputing race will evolve as both the U.S. and China continue to push the boundaries of what is computationally possible.