Supercomputers/Systems/SoftBank GB200 NVL72 Platform
Japan · operational · mixed
SoftBank GB200 NVL72 Platform
Also known as SoftBank NVIDIA GB200 NVL72 AI computing platform
Operated by SoftBank Corp. at SoftBank Corp. (undisclosed site) .
Researched with AI assistance, checked by scripts. These rows were researched by AI agents reading public operator, laboratory and vendor sources, appended by scripts and checked by automated validation; no subject-matter expert has reviewed them (how this site is made). A mechanical second-reader pass has since fetched every cited source: 1350 of 1581 systems have a readable citation that names them, and 64 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
- -
- Rpeak
- -
- Rmax ÷ Rpeak
- -
- Cores
- -
- Power
- -
- Per watt
- - 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 | - SoftBank built and operates the platform | SoftBank Corp. | - | Reported | softbank.jp |
| Accelerator | NVIDIA GB200 NVL72 Accelerators Arm CPU + CUDA GPU · TSMC 4NP 1,224 NVIDIA Blackwell GPUs in NVIDIA GB200 NVL72 racks, to expand beyond 4,000 | NVIDIA | 1,224 reported | Reported | softbank.jpsbintuitions.co.jp |
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 | 1,224 accelerators | 2025-12-22 | Operator stated | Initial 1,224 Blackwell GPUs; SoftBank plans to expand to more than 4,000. | softbank.jp |
The read
SoftBank's liquid-cooled AI computing platform built on NVIDIA GB200 NVL72, which began operating on 22 December 2025 (release dated 25 December). It starts with 1,224 Blackwell GPUs, exactly 17 NVL72 racks of 72 GPUs (derived), with expansion planned to more than 4,000 GPUs; SoftBank cites 10.6 EFLOPS for that expanded platform without stating the precision, so no FLOPS figure is recorded. It uses direct-to-chip liquid cooling. Stated uses are a rental service of dedicated NVIDIA GPU resources, development of SB Intuitions' Sarashina Japanese LLM, and service to Japanese companies and research institutions; the Japanese release links the build to the Ministry of Economy, Trade and Industry's certification of SoftBank's cloud-programme supply plan under the Economic Security Promotion Act. No site is named, so it sits on SoftBank's undisclosed-site record. It is distinct from CHIE-4 (chie-4, a DGX B200 cluster), and SoftBank does not say whether it relates to the Tomakomai or Sakai campuses (softbank-tomakomai-ai-data-center, softbank-sakai-ai-data-center). Both pages are SoftBank group sources (SoftBank Corp. and its subsidiary SB Intuitions), so the tier is reported.
Timeline
- 2025-12-22 First operational source
Site & facility
Change history
- 2026-10-29 Added SoftBank GB200 NVL72 Platform
Nearby in scale
The machines closest to SoftBank GB200 NVL72 Platform by accelerator count, smaller to larger.
- Sferical One · Sweden, operational · 1,152 accelerators
- Miyabi · Japan, operational · 1,120 accelerators
- SK Telecom Titan · South Korea, operational · 1,040 accelerators
- Pioneer-EUS · USA, operational · 1,312 accelerators
- Pioneer-SCUS · USA, operational · 1,312 accelerators
- Pioneer-WEU · Netherlands, operational · 1,312 accelerators
Related systems
Same site
- CHIE-4 · Japan, operational, 2025
- CHIE-2 · Japan, operational, 2024
- CHIE-3 · Japan, operational, 2024
Same operator
- SoftBank Tomakomai AI Data Center · Japan, under construction
- SoftBank Sakai AI Data Center · Japan, planned
Same accelerator
- CoreWeave Vennesla (Bulk N01) · Norway, operational
- MAVERIC · Australia, under construction
- Microsoft Fairwater (Atlanta) · USA, operational
- Mistral Campus AI (Bruyeres-le-Chatel) · France, operational
- Nano 4 GB200 · Taiwan, operational, 2025
- Stargate Abilene · USA, operational, 2025