FugakuNEXT (R-CCS)

FugakuNEXT is the codename for the follow-on flagship supercomputer for Japan after Fugaku. RIKEN is the development lead for the system, Fujitsu holds the design contract,¹ and NVIDIA will supply the GPUs.² It is slated to be deployed in Kobe in 2029 and enter operations in 2030.

• Over 3,400 nodes³ and 15K GPUs⁴
• 20x application speedup over Fugaku and 60x over K computer
• FP64 performance:
  • 2.6 EFLOPS vector from accelerators⁵
  • 48 PFLOPS vector from CPUs⁵
• bf16 performance:
  • 150 EFLOPS from accelerators⁵
  • 3.0 EFLOPS from CPUs⁵
• fp8 performance:
  • 300 EFLOPS from accelerators
  • 600 EFLOPS with 2:1 sparsity²
• Memory:
  • CPUs: 10 PiB with over 7 PB/s aggregate bandwidth⁵
  • Accelerators: 10 PiB with 800 PB/s aggregate bandwidth⁵
• Less than 40 MW⁶ (target 30 MW)⁴

System architecture

Per the initial announcement,⁴

• CPU: Fujitsu Monaka-X (ARM with SME and a possible NPU)
  • 1.4 nm with 3D chiplets
  • SVE2 for vector, SME2 for matrix
  • NVLink-C2C
• GPU: NVIDIA “next-gen” GPU
• Host-Device: TBD coherent interconnect
• Scale-up: NVLink within nodes and possibly between nodes
• Scale-out: Custom high-speed interconnect

It will also have a colocated IBM quantum system with a 156-qubit IBM Heron attached to the scale-out fabric.

The high-level design is as shown:⁵

Undecided as of 2026

The following decisions were stated as being undecided in the 2026 Basic Design Technical Report⁵

• Scale-up network: NVL4 vs NVL72
• Scale-out network:
  • Fujitsu proposed a quad-rail, rail-optimized fat tree with 64 nodes per group
  • NVIDIA proposed a two-level fat tree with Spectrum-X Ethernet
• Storage (see below)

Storage subsystem

FugakuNEXT will have a two-tier storage hierarchy:⁵

1. Tier 1: High-speed node-local storage. All-flash.
2. Tier 2: High-capacity storage shared by all compute nodes

The Tier 2 file system is a traditional parallel file system with a few new feature requirements such as virtual namespaces.

The Tier 1 file system is less well-defined. VAST, Weka, and Scality are all mentioned as candidates for this, but only Weka meets the requirement of being able to deploy node-local. The 2026 design report⁵ also discusses a lot of uncertainty around how node-local SSDs will be presented to applications: as literal local file systems, through a distributed client-side namespace, or something else.

Performance targets

Satoshi Matsuoka presented the following targets for simulation workloads at Salishan 2025:⁶

• Raw hardware performance gain: 10x - 20x
• Mixed precision or emulation: 2x - 8x
• Surrogates / PINN: 10x - 25x
• Total: 200x - 1000x or more over Fugaku (“Zettascale”)

The system performance requirements for the RFP³ and reinforced in the initial NVIDIA announcement⁴ are:

Metric	CPU	GPU
FP64 vector	48 PF	3,000 PF
FP16/BF16 matrix	1,500 PF	150,000 PF
FP8 matrix	3,000 PF	300,000 PF
FP8 matrix, 2:1 sparse		600,000 PF
Memory capacity	10 PiB	10 PiB
Memory bandwidth	8 PB/s	800 PB/s

The storage subsystem will be two tiers:³

Tier	Architecture	Implementation	Bandwidth	IOPS	Capacity
First	Near-node local	Something like CHFS, BeeOND	Write memory in less than 1 minute	Open/close/stat file per process in under 1 second	2x memory
Second	Shared	Lustre, DAOS	20% of first tier	10% of first tier	30x memory

Timeline

The project timeline is as follows:³

After the announcement of NVIDIA being selected as the GPU provider,⁴

• 2025: Phase-1 testbed (~200 GPUs)
• 2026: Phase-2 testbed (~2000 GPUs)
  • “more than” 400 GB200 superchips (inferred from the claims of 40 TF FP64/GPU)
  • XDR InfiniBand (nonstandard with B200)
  • Liquid-cooled Supermicro servers
  • Photo of a GB200 NVL72 suggests this is NVL72
  • “Quantum HPC Collaboration Platform” will be merged in as well to provide ~400 more GPUs
• 2027: Phase-3 testbed with Fugaku-NEXT-like architecture (?? GPUs)
• 2030: Full Fugaku-NEXT with 15K GPUs

Early vision

The details of the project were summarized on a digital poster at SC24:

Satoshi Matsuoka has been talking about their vision for FugakuNEXT since around 2022. The vision for its CPU is:⁷

Themes that may be relevant to a processor or node include:⁷³

• 3D stacking of memory and logic (as depicted above)
• Silicon photonics
• Large SRAMs, a la AMD 3D VCache
• Specialized tensor core-like data paths for scientific motifs like stencils, convolution, FFTs
• CGRA instead of or in addition to SIMD
• Processing-in-memory (PIM)

The CGRA is called out as a “strong scaling accelerator” candidate, so perhaps the CPU socket will have tiles of general-purpose CPU cores as well as CGRA tiles.

Footnotes

1. Fujitsu awarded contract to design next-generation flagship supercomputer FugakuNEXT ↩

2. https://www.riken.jp/pr/news/2025/20250822_1/index.html ↩ ↩²

3. スパコンを使った 最先端の天気予報研究 (science.osti.gov) ↩ ↩² ↩³ ↩⁴ ↩⁵

4. https://www.mext.go.jp/content/20250822-mxt-jyohoka01-000044312_41.pdf ↩ ↩² ↩³ ↩⁴ ↩⁵

5. Basic Design Technical Report for the New Flagship System “FugakuNEXT” ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰

6. Satoshi Matsuoka’s presentation at Salishan 2025. ↩ ↩²

7. スパコンを使った 最先端の天気予報研究 (bnl.gov) ↩ ↩²