Endicott GB300 Cluster (Global AI)

Global AI's Endicott, New York site, where the company announced on 16 March 2026 that it had completed NVIDIA GB300 NVL72 systems that it calls the largest GB300 NVL72 cluster in New York State (a company claim). The same release says Global AI is 'actively deploying' 7,000 GB300s, so the completed deployment and the 7,000 figure may not describe the same state; both are company statements. If the count means GPUs it equals about 97 NVL72 racks of 72 GPUs, but the release gives no rack count. Initial critical IT load is 16 MW, which the company says will ramp to 100 MW through 2026 (company-wide targets of 250 MW in 2027 and 1 GW by end 2029); the release does not say how much of that sits at Endicott. Supermicro server platforms underpin the rack-scale design; no network, storage or cooling vendor is named. A Vera Rubin NVL72 rollout is planned for Endicott and is not part of this record. Both source pages are the same press release (HPCwire and a WOWK-TV syndication), so the tier is reported.