China tops supercomputer rankings with domestically built LineShine

China took the top spot on the latest TOP500 list with LineShine, a domestically built supercomputer installed at the National Supercomputing Centre in Shenzhen and assembled by the Shenzhen Cloud Computing Center. The machine posted 2.198 exaflops on the High Performance Linpack benchmark, more than 20 percent ahead of El Capitan in the United States, and became the first system to exceed two exaflops of sustained double-precision performance using CPUs only.

The June 23 announcement at ISC 2026 in Hamburg gave China its first No. 1 ranking in three years and returned a Chinese system to the top after a long gap in submissions under U.S. export pressure. LineShine uses 13,789,440 cores on custom LX2 processors, the LingKun platform, the LingQi interconnect and Kylin OS, a stack that underscores Beijing’s push for technological self-sufficiency as much as its pursuit of prestige.

That top-line result does not settle the question of AI leadership. LineShine placed fourth on HPL-MxP, a mixed-precision benchmark designed to resemble AI-style work more closely, where it reached 7.92 exaflops. It also ranked No. 1 on HPCG at 22.00 petaflops, but the TOP500 list still measures a narrow slice of performance, not the full spread of workloads that matter for training and serving modern AI systems. Jimmy Goodrich put that limitation bluntly, saying that if hyperscalers submitted their systems, the “world’s fastest” would not crack the top five.

El Capitan, at Lawrence Livermore National Laboratory, dropped to second place but remains a pillar of the U.S. national-security computing base. The system is used by the U.S. government to develop and maintain its nuclear weapons stockpile, a reminder that the race is tied to defense, science and industrial policy as much as to bragging rights. TOP500 said five systems now exceed one exaflop on HPL, with exascale machines spread across Asia, North America and Europe.

China first reached the top of the TOP500 in 2010 and traded the lead with the United States and Japan until 2023. Its last No. 1 system before LineShine was Sunway TaihuLight in 2017. The broader contest now turns less on who can win a headline ranking and more on who can keep building AI-optimized chips, manage soaring power demands and sustain the software ecosystems needed to turn raw compute into usable advantage.