We firmly believe this segment has major global impact potential over the coming years and clearly many far smarter folks, with access to significant budgets, are well underway building systems required. This overview attempt to describe, simply and effectively as possible, the recent past, present, and near future plans for next-gen. High Performance Computing (HPC), running with Quantum component, in Japan.
Quantum operates at speeds beyond our ability to grasp. However blazing fast calculation rates create noise errors which affects precision accuracy, hence this tandem approach with ‘Ai’ assisted guidance while running complex chemistry and physics simulations. The combined power enables basic and applied R&D to explore improved material recipes and operational design at speed and scale across all industries.
In 2020 Fujitsu delivered the Fugaku supercomputer to Rikan in Kobe, this petascale class system claimed #1 position worldwide and is still ranked on the Top-10 list. In late 2023 Rikan announced an MOU with IBM for their upcoming Quantum System Two Heron machine, the first unit shipped outside US, to integrate with Fugaku. The primary budgets to build and operate secured via government funding, based on the Japan National Quantum Strategy released in January 2020, with est. $2.5Bn for all projects over five years, on par with other G7 countries.
In June 2025 Rikan and IBM clipped the ribbon to launch this system, we do not see a comparable example deployed anywhere else in the world. They ran a full day symposium, with an interesting line-up of speakers, and have several industry partners already signed to utilize this new capability. They also hosted their 5th annual Quantum Innovation conference in Osaka from late July, with +200 presentations on a wide range of topics with discussions on new Fujitsu 256-qubit rig which was unveiled in April and their 1,000-qubit build expected in 2026.
Looking ahead: Rikan confirmed the FugakuNEXT project in January 2025, with delivery by 2030, will target Zettascale class. We also noted, via one of symposium keynotes above, it’s training model will be in the 1-Trillion parameter range (xLLM) as described by the MEXT working group back in 2023. While there are many parts of the puzzle, from current install observations on tuning the middle-ware between Quant and HPC to new funding now locked-in, Fujitsu remains a key partner and Nvidia just confirmed for GPU design: interesting times ahead.. 😎
As stated at the top.. a simple brief overview on a significant topic segment, for more details please do feel free to contact us here anytime.
— 5×5 Editors