SSPLAXData Center Power Readiness
An AI campus wants to bring new GPU capacity online, but the compute plan is ahead of the physical site. Firm utility capacity arrives in stages. Non-firm power can bridge part of the gap, but only for flexible load. Onsite generation is available, but it is more expensive and pushes against carbon policy. Transformer gear, cooling, and GPU hall commissioning all move on different schedules.
How much IT load can come online in the current ramp window, and which physical limit prevents the rest of the GPU plan from becoming usable capacity?
The model treats planned IT load as MW moving through firm grid power, non-firm grid service, and onsite generation before merging at the transformer/switchgear yard. That powered load then passes through cooling readiness and GPU halls before becoming online compute capacity. Constraints cap firm utility service, non-firm flexible load, onsite generation, transformer capacity, cooling load, hall readiness, monthly spend, and carbon policy.
- Cooling readiness fully used
- Firm utility capacity full
- Non-firm flex allowance full
- Hall A ready capacity full
The plan is GPU-ready before it is power-and-cooling ready. Firm utility service and the non-firm flex allowance fill first, then cooling readiness becomes the hard ceiling on usable IT load.