Next-generation NVIDIA Vera Rubin NVL72 capacity on a purpose-built campus in the Charlotte metro, booked by the data hall: each 48-rack hall carries 3,456 GPUs, priced at $7.90/GPU-hr on a 5-year term with a 25% deposit. Phase 1 brings 31 data halls online — up to 107,136 GPUs behind 400 MW of off-grid power — with a ready-for-service date of April 15, 2027, and the full campus doubles that to 62 halls and up to 214,272 GPUs.
This is the largest single block of Rubin-generation capacity on the Ax3 marketplace, and the data-hall unit of sale lets a buyer scale from one hall to tens of thousands of GPUs on the same site, fabric and term sheet. Direct-to-chip liquid cooling and InfiniBand are specified across the campus; the dedicated off-grid supply removes the utility-interconnect timing risk that stretches most 2027 deliveries. The public tile shows the headline specification; the complete data room — hall-level configuration, node and fabric specs, exact location, counterparty and full term sheet — is shared once both sides agree to connect.
Cluster size, site and counterparty details are confirmed by Ax3 before any introduction.
