China’s LineShine supercomputer, built entirely with domestic processors, has claimed the top spot in the 67th edition of the TOP500 rankings, dislodging the US-built El Capitan with a record 2.198 exaflops of sustained performance. The rankings were unveiled at the ISC High Performance 2026 conference in Hamburg, Germany, marking the first time a Chinese system has led the list since 2017. India placed 16th globally with seven supercomputers on the list, led by the Shakti Cloud system at 32nd position with 84.31 petaflops.
What Is the TOP500 Ranking?
The TOP500 is a biannual ranking of the world’s most powerful supercomputers, first published in June 1993. It was conceived by Hans Meuer, Jack Dongarra, Horst Simon, and Erich Strohmaier at a small meeting in Mannheim, Germany, and has since become the definitive benchmark for high-performance computing (HPC) worldwide.
The ranking is based on the High-Performance Linpack (HPL) benchmark, which measures a system’s ability to solve dense linear equations using double-precision floating-point arithmetic. Results are expressed in flops (floating-point operations per second), with classifications progressing through megaflops, gigaflops, teraflops, petaflops, and exaflops. A single exaflop represents one quintillion operations per second.
A significant development at the ISC 2026 conference was the announcement that the Association for Computing Machinery (ACM) will take over ownership of the TOP500 list from the ISC group. Going forward, each edition will receive a Digital Object Identifier (DOI) for easier reference, though the list will remain free and publicly accessible.
LineShine: China’s New Number One Supercomputer
LineShine (also referred to as Ling Cheng or Ling Shen) is installed at the National Supercomputing Centre in Shenzhen (NSCS) and was built by the Shenzhen Cloud Computing Center. It recorded 2.198 exaflops on the HPL benchmark, more than 20 percent ahead of the second-placed El Capitan (1.809 exaflops). Its theoretical peak performance is 2.736 exaflops, meaning it achieved roughly 80 percent of its peak in sustained operation.
LineShine is the fifth exascale system ever built, joining El Capitan, Frontier, Aurora, and JUPITER as machines capable of crossing the one exaflop barrier. It is the first Chinese system to top the rankings since Sunway TaihuLight in 2017, which held the number one spot at just 93 petaflops, illustrating the enormous leap in computing power over the past nine years.
The machine also claimed the number one position on the HPCG (High-Performance Conjugate Gradients) benchmark, which measures performance on memory-bound, data-intensive workloads that better reflect real-world scientific applications. LineShine posted 22 HPCG-petaflops, ahead of El Capitan’s 17.41. On the HPL-MxP mixed-precision benchmark, it placed fourth with 7.92 exaflops, recording a modest 3.6 times speedup over its standard HPL score, consistent with its CPU-only design.
Inside LineShine: Technical Architecture
What makes LineShine remarkable is that it achieved the top ranking using no GPU accelerators at all, a departure from every other exascale system in the top five. All previous exascale machines, including El Capitan, Frontier, Aurora, and JUPITER, rely on GPU or APU accelerators to boost their floating-point throughput. LineShine is built entirely on CPUs, representing a fundamentally different architectural philosophy.
The system is built on the custom LingKun platform and uses 45,360 LX2 processors, each containing 304 cores running at 1.55 GHz. These chips are Armv9-compliant, support SVE2 and Scalable Matrix Extensions, and are built from two compute dies, each with four 38-core clusters. Each LX2 processor delivers an estimated 3.3 teraflops of FP64 performance at 690 watts. In total, the system uses 13,789,440 cores.
The processors are linked through a proprietary LingQi interconnect providing 1.6 terabits per second per CPU, and the system runs on the Kylin operating system, a Chinese Linux distribution. The entire machine consumes approximately 42.2 megawatts of power, yielding an efficiency of 52.07 gigaflops per watt.
Reports identify the LX2 processor as a Huawei-designed chip, though the official TOP500 listing does not name the designer. The system’s all-CPU architecture also means it excels at double-precision (FP64) workloads that are critical for traditional scientific simulations, while its unaccelerated design limits performance gains in mixed-precision AI inference tasks.
Top 5 Systems at a Glance
All five top-ranked systems in the 67th edition have crossed the exascale threshold. The table below summarises their key specifications.
| Rank | System | Country | HPL Performance (Exaflops) | Cores | Architecture |
|---|---|---|---|---|---|
| 1 | LineShine | China | 2.198 | 13,789,440 | Custom LX2 CPUs (ARM), LingKun platform |
| 2 | El Capitan | USA | 1.809 | 11,340,000 | AMD EPYC + AMD Instinct MI300A APU |
| 3 | Frontier | USA | 1.353 | 9,066,176 | AMD EPYC + AMD Instinct MI250X GPU |
| 4 | Aurora | USA | 1.012 | 9,264,128 | Intel Xeon + Intel Data Center GPU Max |
| 5 | JUPITER Booster | Germany | 1.000 | 4,801,344 | NVIDIA Grace Hopper GH200 Superchip |
Beyond the top five, Italy’s energy company Eni entered the list at No. 6 with its HPC7 system (571.5 petaflops), built on the same HPE Cray architecture as El Capitan. Japan’s Fugaku, which once held the number one spot, now ranks No. 9 with 442 petaflops. Switzerland’s Alps system rounds out the top 10 at 434.9 petaflops.
US Still Leads in Number of Systems
Although China claimed the top spot, the United States remains the most represented country on the list with 161 systems, down from 171 in the previous edition. China has 30 systems on the list, placing second by system count. Other significant presences include Germany (29 systems), Japan (27), France (22), and the United Kingdom (15).
In terms of aggregate computing power, however, the US lead is narrower. China’s contribution to the list’s total computing power surged significantly in this edition, driven almost entirely by LineShine, which alone accounts for more than half of the new compute power added to the 67th edition. A total of 44 new systems entered the list, with an average performance of 91 petaflops.
On the vendor front, HPE is the dominant system integrator, supplying El Capitan, Frontier, HPC7, HPC6, and Alps. On the processor side, AMD powers four of the top eight systems directly, contributing more than 40 percent of the combined top 10 performance.
India’s Position in the 67th TOP500
India is ranked 16th globally with seven supercomputers on the list. The country’s most powerful system, Shakti Cloud, operated by Yotta Data Services Private Limited in Mumbai, secured the 32nd position with a performance of 84.31 petaflops. Shakti Cloud, which debuted on the TOP500 list in November 2025, is built on Dell PowerEdge XE9680 servers with Intel Xeon Platinum processors and NVIDIA H100 GPUs, connected via InfiniBand NDR400.
The remaining Indian systems on the list are:
| System | Institution/Location | Primary Use |
|---|---|---|
| PARAM Siddhi-AI | C-DAC, Pune | AI and HPC workloads, ranked in TOP500 since 2020 |
| Pratyush | Indian Institute of Tropical Meteorology (IITM), Pune | Weather and climate research |
| Mihir | National Centre for Medium Range Weather Forecasting (NCMRWF), Noida | Monsoon and weather forecasting |
| Aditya | Indian Institute of Tropical Meteorology (IITM), Pune | Climate modelling |
| SahasraT | Supercomputer Education and Research Centre (SERC), IISc, Bengaluru | General purpose scientific computing |
| Arka | Indian Institute of Tropical Meteorology (IITM), Pune | AI/ML workloads for weather (ranked 493rd) |
| Param Brahma | IISER, Pune | Research and scientific computing, 1.75 petaflops |
Pratyush and Mihir are both Cray XC-40 class systems operated under the Ministry of Earth Sciences. Pratyush, installed at IITM Pune in 2018, has a peak performance of 4 petaflops and is used for ocean-atmosphere climate system research. Mihir, at NCMRWF Noida, delivers 2.8 petaflops and supports the Monsoon Mission and climate change research.
PARAM Siddhi-AI, installed at C-DAC Pune in 2020, was India’s first AI-specific supercomputer and has been a consistent presence on the TOP500 list since its debut at rank 62. It delivers 4.6 petaflops and is built on the NVIDIA DGX A100 architecture.
India’s Supercomputing Journey: The National Supercomputing Mission
India’s growing presence in the TOP500 is driven by the National Supercomputing Mission (NSM), a flagship programme launched in 2015 with a vision to build a comprehensive indigenous supercomputing ecosystem. The mission is jointly steered by the Ministry of Electronics and Information Technology (MeitY) and the Department of Science and Technology (DST), and implemented by the Centre for Development of Advanced Computing (C-DAC), Pune, and the Indian Institute of Science (IISc), Bengaluru.
C-DAC itself has a storied history in supercomputing. It was established in 1988 after India faced denial of access to advanced computing technology from the West. Under the leadership of Dr. Vijay P. Bhatkar, C-DAC delivered the PARAM 8000 in 1991, India’s first supercomputer, achieving 1 gigaflop of performance.
As of early 2026, the NSM has deployed 38 supercomputers across the country with a cumulative computing capacity of 44 petaflops. Among the mission’s key achievements are the development of the Rudra server, India’s indigenously designed and manufactured HPC server at par with global HPC-class servers, and the Trinetra high-speed interconnect, rolled out in three phases: Trinetra-PoC, Trinetra-A (100 Gbps), and Trinetra-B (200 Gbps). Indigenous systems have been deployed at IISc, IITs, NITs, C-DAC centres, and other research institutions. A 20-petaflop PARAM Rudra supercomputer is under construction at C-DAC Bengaluru, which will be India’s largest indigenous system when completed.
The PARAM series of supercomputers includes notable systems such as PARAM Pravega (3.3 petaflops at IISc Bengaluru, the largest in Indian academia), PARAM Shakti (3.1 petaflops at IIT Madras, inaugurated January 2026), PARAM Rudra (3 petaflops at IIT Bombay), and PARAM Shivay (the first indigenously assembled system at IIT BHU in 2019).
Key Takeaways
- The 67th edition of the TOP500 rankings was released at the ISC High Performance 2026 conference in Hamburg, Germany.
- LineShine, China’s CPU-only supercomputer, topped the list with 2.198 exaflops, becoming the first system ever to exceed two exaflops of sustained performance and the first Chinese number one since Sunway TaihuLight in 2017.
- LineShine is built on 45,360 custom LX2 processors (304 cores each) using the LingKun platform with no GPU accelerators, consuming 42.2 MW of power.
- The United States leads the list with 161 systems, but China tops the performance charts with LineShine at number one.
- India ranks 16th with seven supercomputers on the list, led by Shakti Cloud (84.31 petaflops) at 32nd position.
- The National Supercomputing Mission (NSM), launched in 2015, has deployed 38 supercomputers with a cumulative capacity of 44 petaflops, with C-DAC as the nodal implementing agency.
