Khaberni - In a move towards leadership and the development of AI infrastructure in the country, Abu Dhabi unveiled the establishment of a national-scale supercomputer for artificial intelligence in India with a computational capacity of 8 exaflops, where this innovative project was announced on the sidelines of the "Artificial Intelligence Impact Summit 2026" which concluded yesterday in New Delhi, India.
In this context, supercomputers are no longer just scientific tools, but have become an explicit indicator of the balance of technological power among nations, as each additional "petaflop" not only means greater speed but also broader capacity for innovation, more accurate simulations, and faster decisions in fields ranging from national security to drug discovery.
The power of these systems is measured in "petaflops" (PFLOPS) - that is, the number of trillion floating-point operations that can be executed per second - and with the introduction of new systems and updates, the rankings change at an accelerated pace, driven by the development of specialized processors, graphic accelerators, and architectures specifically designed for AI workloads.
Leading the supercomputer rankings in 2026, "El Capitan" boasts an astonishing power of 1809 petaflops, enhancing its position at the peak of the exascale race, and it was inaugurated in late 2024 at the Lawrence Livermore National Laboratory in the United States. It costs about $600 million and is primarily used to support U.S. national security missions, weapon design, and nuclear simulation.
In second place is the "Frontier" with 1353 petaflops, which is the first and fastest "exascale" supercomputer in the world, hosted at a computing facility in Tennessee, USA, and it entered service for the first time in 2022.
Meanwhile, "Aurora" ranks third with 1012 petaflops, funded by the U.S. Department of Energy (DOE) and designed by Intel and Cray for the Argonne National Laboratory.
The top three systems mentioned above exceed the thousand petaflop barrier, marking the practical use of widescale (Exascale) computing.
And in ranks 4 to 10, the following supercomputers are listed:
4- Eagle: 561.20 petaflops
5- "HPC 6": 477.90 petaflops
6- Fugaku: 442.01 petaflops
7- Alps: 434.90 petaflops
8- LUMI: 379.70 petaflops
9- Leonardo: 241.20 petaflops
10- "Summit": 148.60 petaflops
A clear gap
The figures reveal a notable performance gap between the top three and the rest of the list, reflecting a sharp acceleration at the technical pyramid's apex, while a set of systems within the 400-560 petaflop range are closely competing, indicating intense competition in the high-performance tier.
The United States continues to enhance its leadership with several systems topping the list, while Europe maintains a strong presence through systems like "LUMI," "Leonardo," and "Alps," and Japan remains an influential player thanks to "Fugaku" continuing to rank among the top tiers.
Beyond the numbers
Behind each supercomputer, there is a complex system of advanced processors, high-speed networks, high-efficiency cooling systems, and software optimized to work on a massive scale. The race is no longer just about setting record numbers, but about achieving the highest possible performance with energy efficiency, amid growing environmental and economic pressures.
With the escalating demand for computing in healthcare, energy, artificial intelligence, and scientific research, competition is expected to intensify in the coming years, and new architectural designs, more efficient accelerators, and advanced cooling technologies will lead a new wave of performance breakthroughs.
