Free · directional · no contact details required
Most halls do not fail on cooling. They fail on a 63 A tap-off.
Thirteen constraints across electrical, thermal and physical domains decide how much AI compute an existing hall can carry. One of them binds first, and it is rarely the one people expect. Give the engine seven numbers you already know and it will tell you which — and which of your inputs nobody has actually measured.
Capacity qualifier
What actually stops your hall?
Seven numbers you already know. The same engine that produces our paid study returns the binding constraint, the deployable rack count, and the inputs nobody has measured yet. For scenarios with indicative capital cost and time to full capacity, the free Time-to-Power comparison goes further; a human-reviewed engagement is the paid step.
Want to see what the paid study looks like before you ask? Read one in full — the whole document, the deck, and the tables behind every number.
What you get
The binding constraint and why, deployable racks as the hall stands and after relief, the item that sets the date, and the list of inputs still assumed.
Where the paid line is
The free qualifier shows the binding constraint and the rack count. The free Time-to-Power comparison also shows scenarios with indicative capital cost and time to full capacity, each with its evidence class, and a time that rests on an assumed lead time is labelled as one, not as a date. What is paid is engineering you can rely on: human-reviewed engagements, and metered engine usage beyond the free allowance.
Why it is honest
Every figure carries an evidence class. A value you supply is customer data; a value we fall back to is an assumption, and it is named as one. A model applied to measured data stays a model.