On
The World's 10 Fastest Supercomputers in 2026

The latest Top500 supercomputer rankings reveal a fascinating shift in global computing power. Five of the world's most powerful machines call the United States home, while China and Japan each claim two spots. Finland and Italy round out the top ten with one system each. These rankings come directly from the prestigious Top500 supercomputer project, which tracks the planet's most formidable computing systems.

The Top500 initiative releases updated rankings twice annually—once in late May or early June, and again in November. These findings are presented at major international conferences dedicated to supercomputing. The project has been running continuously since 1993, led by respected researchers including Jack Dongarra from the University of Tennessee Knoxville; Erich Strohmaier and Horst Simon from the U.S. Department of Energy's Lawrence Berkeley National Laboratory; and Hans Meuer from the University of Mannheim.

Article Contents

  • What is a supercomputer?
  • How supercomputer speed is measured
  • The world's top 10 supercomputers

What Exactly is a Supercomputer?

Supercomputers, or High Performance Computers (HPC), represent the cutting edge of computational power. These machines operate at performance levels far beyond what most people can imagine. You'll find them humming away in university research labs, national laboratories, and other critical scientific facilities across the globe.

Supercomputers play an indispensable role in computational science. They tackle extraordinarily complex calculations across numerous fields: quantum mechanics, weather forecasting, climate research, oil and gas exploration, and molecular modeling (which involves calculating the structures and properties of chemical compounds, biological macromolecules, polymers, and crystals). They also run physics simulations—everything from modeling the universe's first moments after the Big Bang to simulating aircraft aerodynamics, spacecraft dynamics, nuclear weapon detonations, and nuclear fusion reactions. Throughout their history, supercomputers have also proven critical for cryptanalysis and code-breaking.

How Do We Measure Supercomputer Speed?

Unlike conventional computers measured in MIPS (million instructions per second), supercomputer performance is measured in FLOPS—floating-point operations per second. By 2015, the fastest machines could execute up to 10 petaFLOPS (that's 10 quadrillion calculations per second), expressed in the unit PFLOPS. Most modern supercomputers run Linux-based operating systems.

The machines on this list are all measured in petaflops. One petaflop equals 10^15 (10 quadrillion) calculations per second. Yes, that's a mind-bogglingly large number.

The World's Top 10 Supercomputers

1. Frontier, United States

Built in 2022 by Hewlett Packard Enterprise in collaboration with its Cray subsidiary, Frontier holds a historic distinction: it's the world's first exascale supercomputer. This means it can perform at least 10^18 calculations per second—a full exaflop of computing power.

Frontier boasts an incredible 8,730,112 cores and achieves 1.1 exaflops during Linpack benchmark tests. The system uses HPE's latest Cray EX235a architecture, combining 64-core AMD EPYC 7A53 processors running at 2GHz with AMD MI250X GPUs. What's particularly impressive is its energy efficiency ranking of 52.23 gigaflops per watt, making it the world's most power-efficient supercomputer. Each of its 74 computing cabinets weighs approximately 3.63 tons, and the entire system cost roughly $600 million to build.

2. Fugaku, Japan

Performance: 442.010 petaflops, peak efficiency 537.212 petaflops

Fugaku supercomputer
Fugaku supercomputer

Developed by Fujitsu and installed at RIKEN's Center for Computational Science (R-CCS) in Kobe, Fugaku smashed the previous world record with 442 petaflops—more than three times faster than the second-ranked system on this list. RIKEN's director, Satoshi Matsuoka, noted that they finally achieved something significant: "we could use the entire machine instead of just a small fraction of it."

After the initial rankings, Matsuoka's team spent months fine-tuning the code to squeeze maximum performance from the system. His assessment? "I don't think we can improve much further than this."

3. Aurora, United States

Aurora represents one of the newest entries on this list, with potential to become the most powerful machine in the coming years.

This supercomputer delivers 585 petaflops (0.59 exaflops) and resides at Argonne National Laboratory in Illinois. It's the second exascale system ever built. Aurora is a joint creation of Intel and HPE, having gone live in June 2023. The system integrates scientific tools and analytics, executing modeling, simulations, and artificial intelligence workloads.

According to ALCF representatives, Aurora has the potential to reach 2 exaflops—double Frontier's power. Its computational muscle could generate accurate models across multiple domains: climate prediction, materials science, energy storage, and nuclear fusion research. The real concern is nuclear fusion—that's where Aurora's power is being primarily directed.

4. Eagle, United States

Unlike other supercomputers confined to laboratory facilities, Eagle operates as part of Microsoft's broader Azure cloud data center infrastructure. This means anyone with an Azure account can access its computational power through the cloud platform.

Eagle delivers 561 petaflops (0.56 exaflops) and became operational in August 2023. Microsoft equipped this machine with 48-core Intel Xeon Platinum 8480C processors using Sapphire Rapids architecture, paired with Nvidia H100 GPUs running Hopper architecture. The system totals 1.1 million processing cores.

5. LUMI, Finland

LUMI (Large Unified Modern Infrastructure) was constructed by HPE in 2022 and deployed in Finland, where it became Europe's fastest supercomputer. The system contains 1,110,144 cores and achieves 151.9 petaflops.

LUMI runs on identical processors as Frontier and achieves an energy efficiency rating of 51.63 gigaflops per watt, making it the world's second-most power-efficient supercomputer.

6. Leonardo, Italy

Leonardo utilizes Intel Xeon Platinum 8358 32-core processors combined with Nvidia A100 and HDR100 processing chips, delivering 238.7 petaflops of computing power. The system began operations in November 2022 at Bologna and cost $240 million to construct. Intel and Nvidia jointly handle the software infrastructure running the machine.

7. Summit, United States

Performance: 148.600 petaflops, peak efficiency 200.795 petaflops

Summit supercomputer

Based at Oak Ridge National Laboratory (ORNL) in Tennessee, Summit was built by IBM and stands as America's fastest supercomputer. It launched in 2018 with 148.8 petaflops of performance, powered by 2,282,544 IBM Power9 cores and 2,090,880 Nvidia Volta GV100 cores across 4,356 nodes. Each node contains two 22-core Power9 CPUs and six Nvidia Tesla V100 GPUs.

Recently, two research teams working on Summit won the prestigious Gordon Bell Prize for outstanding achievement in high-performance computing—often called the "Nobel Prize of supercomputing."

8. Sierra, United States

Performance: 94.640 petaflops, peak efficiency 125.712 petaflops

Sierra supercomputer
Sierra supercomputer

Housed at Lawrence Livermore National Laboratory (LLNL) in California, Sierra achieved an HPL benchmark score of 94.6 petaflops. Its 4,320 nodes each feature two Power9 CPUs and four Nvidia Tesla V100 GPUs, giving it an architecture nearly identical to Summit's design.

Sierra also ranked 15th on the Green500 list of the world's most energy-efficient supercomputers.

9. Sunway TaihuLight, China

Performance: 93.015 petaflops, peak efficiency 125.436 petaflops

Sunway TaihuLight supercomputer

Installed at China's National Supercomputing Center in Wuxi, Sunway TaihuLight once held the number one ranking for two consecutive years (2016 and 2017). However, its position has steadily declined since then. It dropped to third place a year ago and now ranks fourth.

Built by China's National Research Center of Parallel Computer Engineering and Technology (NRCPC), Sunway TaihuLight achieved 93 petaflops on the HPL benchmark. The system uses exclusively homegrown Sunway SW26010 processors—a significant point of national pride for China's computing independence.

10. Perlmutter, United States

Performance: 64.6 petaflops, peak efficiency 89.795 petaflops

Perlmutter stands as the only brand-new system making its debut on the top 10 list this cycle. Built on the HPE Cray Shasta platform, it combines AMD EPYC nodes with 1,536 Nvidia-accelerated A100 nodes for its processing architecture.

Related Articles