TSUBAME4.0

Compute

960 H100e

Chips

960

First operational

Apr 1, 2024

TSUBAME4.0 is a large GPU cluster operated by Tokyo Institute of Technology, located in Japan, in the public sector. It came online around April 2024. At that moment it ranked number 68 among known clusters.

It is built on 960 NVIDIA H100 SXM5 80GB. That amounts to about 960 H100e of AI compute. The cluster draws roughly 1.36 MW. Its hardware is estimated at $34M.

Epoch AI rates the certainty of this record as confirmed.

Full record

Owner: Tokyo Institute of Technology
Users: Cloud
Sector: Public
Status: Existing
Country: Japan
Location: 4259 Nagatsutacho, Midori Ward, Yokohama, Kanagawa 226-0026, Japan
Primary chip: NVIDIA H100 SXM5 80GB
Primary chip count: 960
Total AI chips: 960
H100 equivalents: 960
8-bit throughput: 1.9×10¹⁸ OP/s
Power capacity: 1.36 MW
Hardware cost: $34M
First operational: Apr 1, 2024
Rank when it came online: 68
Certainty: Confirmed