China's 'LingSheng' Supercomputer Tops TOP500 for First Time in 8 Years, Breaks 2 Exaflop/s Barrier

The June 2026 edition of the TOP500 supercomputer rankings, unveiled at the ISC 2026 conference in Hamburg, Germany, has a new champion. China’s “LingSheng” (灵晟) supercomputer stormed to the number one position with an Rmax of 2.198 Exaflop/s, marking the first time a Chinese system has topped the HPL benchmark in eight years — and the first system anywhere in the world to break through the 2 Exaflop/s ceiling.

Built on a pure CPU architecture, LingSheng runs on the domestically developed “LingKun” platform. At its heart are 304-core LX2 processors clocked at 1.55 GHz, connected via a custom “LingQi” interconnect fabric. The system packs a staggering 13.79 million cores in total and runs the Kylin operating system — a completely homegrown stack from processor to OS.

LingSheng didn’t just dominate the headline HPL benchmark. It also claimed first place in the HPCG (High Performance Conjugate Gradient) test, a metric that stresses memory and communication subsystems more heavily than traditional Linpack. In the HPL-MxP mixed-precision benchmark, the system placed fourth overall.

On the efficiency front, LingSheng draws approximately 42.2 megawatts of power, yielding an energy efficiency of 52.07 Gigaflops per watt — good enough for 50th place on the GREEN500 list.

The other major newcomer to the top ten is Italy’s HPC7, built by HPE for energy giant Eni. Powered by AMD’s “Zen 4” EPYC processors and Instinct MI300A APUs, HPC7 delivers 571.50 Petaflop/s and debuts in sixth place. Eni’s existing HPC6 system also held on at number eight, giving the Italian energy company two systems in the global top ten.

LingSheng’s ascent represents a significant milestone for China’s homegrown semiconductor and HPC ecosystem. The system was previously reported to have been lit up in Shenzhen, with sustained performance already exceeding 2 Exaflops before the official TOP500 submission — a claim now validated by the rankings.