Meta 2017 V100 Cluster

Compute

1.4K H100e

Chips

22,000

First operational

Oct 1, 2017

Meta 2017 V100 Cluster is a large GPU cluster operated by Meta AI, located in the United States, in the private sector. It came online around October 2017. At that moment it ranked number 1 among known clusters.

It is built on 22,000 NVIDIA Tesla V100 SXM2. That amounts to about 1.4K H100e of AI compute. The cluster draws roughly 15.5 MW. Its hardware is estimated at $476M.

Epoch AI rates the certainty of this record as confirmed. Unsure if this cluster has since been decommissioned, since it's relatively old Our only information on the date is sources saying it was "Built in 2017". The V100 wasn't released until June 2017, so I assume this cluster was built in the second half of 2017. October 1st is the middle of the second half of 2017, so I put that as the date

Full record

Owner: Meta AI
Users: Meta
Sector: Private
Status: Existing
Country: United States
Primary chip: NVIDIA Tesla V100 SXM2
Primary chip count: 22,000
Total AI chips: 22,000
H100 equivalents: 1.4K
Power capacity: 15.5 MW
Hardware cost: $476M
First operational: Oct 1, 2017
Rank when it came online: 1
Certainty: Confirmed