Japan · operational · ai training

ABCI 3.0

Operated by AIST at ABCI Data Center, Kashiwa .

Phase 1 seed dataset, compiled by hand. These rows were built from public operator, laboratory and vendor sources. A mechanical second-reader pass has since fetched every cited source: 1092 of 1153 systems have a readable citation that names them, and 40 are genuinely weakly sourced. Every claim carries its source and a confidence tier. Treat anything below verified as a lead, not a citation.

Measured performance

Rmax: 145.1 PFlop/s
Rpeak: 181.5 PFlop/s
Rmax ÷ Rpeak: 79.9%
Cores: 479,232
Power: 3.60 MW
Per watt: 40.3 GF/W

Figures are as last publicly reported for the configuration described below, not a live measurement. Where a system was upgraded in place, the post-upgrade configuration is shown and the earlier one appears in the timeline.

What this machine is made of

One row per supplier relationship. “Supplier at build” is the company that shipped the part at the time; where that company has since been acquired, the parent it rolls up to today is shown beside it. That distinction is what makes ticker-level aggregation possible across a thirty-year dataset.

Role	Part	Supplier at build	Quantity	Confidence	Source
Integrator	- HPE-built	Hewlett Packard Enterprise HPENYSE	-	Verified	top500.org nextplatform.com
Accelerator	NVIDIA H200 SXM Accelerators CUDA (Hopper GH100) · 16,896C · 1.98 GHz · TSMC 4N · 700 W NVIDIA H200 SXM5 141GB	NVIDIA NVDANASDAQ	-	Verified	top500.org nextplatform.com
CPU	Intel Xeon Platinum 8558 CPUs x86-64 · 48C · 2.1 GHz · Intel 7 Xeon Platinum 8558 48C 2.1GHz	Intel INTCNASDAQ	-	Reported	top500.org
Interconnect	NVIDIA InfiniBand NDR (Quantum-2) Interconnect Fat tree or dragonfly+ · 400 Gb/s per port · Under 0.6 microseconds Infiniband NDR200	NVIDIA NVDANASDAQ	-	Reported	top500.org

AI datacenter metrics

Figures beyond an accelerator count, each with who stated it. What can and cannot be compared across AI datacenters.

Metric	Value	As of	Basis	Note	Source
Accelerators	6,128 accelerators	2025-06	Operator stated	766 compute nodes with 8 NVIDIA H200 each, per AIST documentation.	docs.abci.ai unit.aist.go.jp
Facility power	6 MW	2024-11	Operator stated	Electrical capacity of the AIST AI data centre after the ABCI 3.0 upgrade (5.2 MW cooling, 144 racks). Facility figure from the ABCI 3.0 paper.	arxiv.org
PUE	1.15	2025-06	Operator stated	AIST states PUE '< 1.15' as of June 2025 (32 C water cooled by a cooling tower). Value is the stated upper bound.	unit.aist.go.jp

The read

AIST's third physical ABCI generation, not an upgrade of the machine already in this dataset: HPE replaced the NEC and Fujitsu integrators that built ABCI 1.0 and 2.0, on Xeon Platinum hosts paired with NVIDIA H200. The original `abci` row (1.0, 2018, Fujitsu-built, Tesla V100) is a genuinely different, still-correctly-recorded machine, not an outdated figure for this one.

Timeline

2024-07 Announced source
2025-06 List debut at #65 (June 2025) source
2025-06 First operational source

Public list appearances

Pointers only: rank and edition, linking to the canonical entry. We do not reproduce list tables. See sourcing policy.

Site & facility

ABCI Data Center, Kashiwa

Location: Kashiwa, Japan
Cooling: Direct liquid cooling with hot-water loop
PUE: 1.1
Commissioned: 2018

abci.ai

Change history

2026-10-04 Added AI datacenter metrics for abci-3
2026-09-22 Gained a Green500 rank pointer (#65 on the 2025-06 list)

Source check

We fetched this system's own citations and recorded whether each page actually mentions it. This is a corroboration signal, not a fact check, and it is published so you can see how well the sourcing holds up rather than take it on trust.

Cited source	Result	Found on the page
top500.org	names system and part	NVIDIA H200 SXM, Intel Xeon Platinum 8558, NVIDIA InfiniBand NDR (Quantum-2)
nextplatform.com	names system and part	NVIDIA H200 SXM, Intel Xeon Platinum 8558, NVIDIA InfiniBand NDR (Quantum-2)

Checked 2026-10-09 by pnpm verify. Re-run it and the table changes with the web.