NVIDIA Circe

Compute

36.38 H100e

Chips

576

First operational

Nov 1, 2018

NVIDIA Circe is a large GPU cluster operated by NVIDIA, located in the United States, in the private sector. It came online around November 2018. At that moment it ranked number 22 among known clusters.

It is built on 576 NVIDIA Tesla V100 SXM2. That amounts to about 36.38 H100e of AI compute. The cluster draws roughly 0.4 MW. Its hardware is estimated at $13M.

Epoch AI rates the certainty of this record as confirmed.

Full record

Owner: NVIDIA
Users: NVIDIA
Sector: Private
Status: Existing
Country: United States
Primary chip: NVIDIA Tesla V100 SXM2
Primary chip count: 576
Total AI chips: 576
H100 equivalents: 36.38
Power capacity: 0.4 MW
Hardware cost: $13M
First operational: Nov 1, 2018
Rank when it came online: 22
Certainty: Confirmed