A strong answer. “Space is almost never the binding constraint, and I would say that first. A hall designed for 5 to 10 kW per rack has the floor area for GPU racks and cannot power or cool them — a dense training rack can draw 40 to over 100 kW, so a single row can exceed what a whole aisle was designed for.

So the questions in order. What is our contracted power and our actual draw? The gap is what is genuinely available, and it is usually smaller than people assume because commitments are counted optimistically. What can the distribution deliver to a single rack position? Even with headroom in the building, the busway, PDUs and branch circuits to that spot may cap far lower. Can we reject the heat? This is where it usually fails — air cooling runs out somewhere around 30 kW per rack in most designs, so we are talking about rear-door heat exchangers or direct-to-chip liquid, which needs a water loop the building may not have. Can the floor carry it? Dense racks are heavy, and raised floors have limits.

Then the honest answer, which is usually: we can deploy a small number of racks by spreading them across the hall to distribute thermal load — which wastes space and is often the right pragmatic move — and anything at scale needs either a retrofit with liquid cooling or a purpose-built facility. I would give the business both the number of racks we can take today and the lead time and cost for each option, because ‘no’ without alternatives just gets routed around.”