Japan has announced plans to begin construction of its first “zeta-class” supercomputer, Fugaku Next, next year. Once fully operational, it will be 1,000 times faster than today’s most powerful supercomputers.
See also: The world's most powerful supercomputer will have 100,000 AI chips
The machine,which could cost more than $750 million, will help Japan keep up with the pace of development of artificial intelligence (AI) and is expected to be fully operational by 2030.
Plans for the new machine — first released on August 28 by Japan's Ministry of Education, Culture, Sports, Science and Technology (MEXT) — reveal that the Fugaku Next supercomputer could reach zetaFLOPS -scale speeds , something never achieved before.
FLOPS are used to measure how fast computers can solve problems. A supercomputer with a speed of 1 zetaFLOPS could perform one six-million (1 followed by 21 zeros) calculations per second. Today's most powerful supercomputers have just broken the exaFLOPS, meaning they can perform just over one five-million (1 followed by 18 zeros) calculations per second.
See also: Meta: Our new AI supercomputer will be the fastest in the world
The decision to build such a powerful machine was made "to keep pace with the development of scientific research using artificial intelligence," Japanese news website Nikkei in an article.
The proposed supercomputer , Fugaku Next, is being touted as the successor to the Japanese supercomputer Fugaku ( 0.44 exaFLOPS ), which previously held the title of the world's fastest supercomputer until it was dethroned in 2022 by the US Frontier supercomputer ( 1.2 exaFLOPS ) at Oak Ridge National Laboratory in Tennessee. Fugaku is currently considered the fourth most powerful supercomputer in the world.
See also: Microsoft & UK build world's most powerful weather supercomputer
One of the biggest challenges engineers in building the new supercomputer is finding a way to make it run efficiently. In 2023, computer experts predicted that a zeta-class machine built using current supercomputer technologies would require the energy equivalent of the output of 21 nuclear power plants, as previously reported by the computer news website HPCwire.
Source: livescience
