Hall A · longevity
Take-or-pay or owner-user workload. No experimental plant on this bus. This is rent, and the only thing that finances a building.
From explainer to experiment
A live hall with a fenced test pad next door.
Proposed architecture. Not an existing facility. Not a claim that NVIDIA, AMD, Microsoft, or Google participates.
Proposal
The experimental pad must be electrically isolated. A failed skid cannot take down paying load. Insurance and the utility will require that even if a slide deck does not.
Selling line that is true: 4 MW in production in Texas, under typical large-load thresholds, plus a 1 MW instrumented island where power, water, and silicon get a witnessed run next to a live hall.
Selling line that is not used here: “the world’s first,” or a named hyperscaler as a tenant.
Pad B · proof
Cooling skids, CDUs, dry coolers, power systems, storage, DCIM, accelerators, neuromorphic experiments. Full metering. Classes of possible participants: semiconductor, cooling, power, utilities, universities, labs, startups — none named as committed.
Built to give technology companies a place to prove what works in the real world. A 5 MW production hall that is “also a lab” on the same bus is not financeable. Split it, or do not build it.
Data last checked 2026-09-04 · Current public data is a snapshot, not a live meter
Show me the receipts