The asset is described in megawatts. How many of them can run AI racks, and how sure is anyone?
Your situation
A data-centre asset is usually described by contracted power. For an AI thesis the relevant number is the compute the hall can host, which depends on constraints the seller's description does not mention.
Underwriting needs the answer and the confidence in it. A figure that rests on a library default is not the same as one built from the seller's measured data, and a diligence note that does not separate them is hiding the variance that matters.
A portfolio view needs one methodology across assets. Ranking halls by figures prepared under different assumptions compares the assumptions, not the halls.
The decision this informs
What an asset is worth as AI capacity, which assets in a set are worth pursuing, and which questions to put to a seller before a price is set.
THE FIRST QUESTIONHow much of the stated capacity is useful AI compute, and how much of that answer is assumption?
What you probably already have
- The seller's data room: single-lines, capacity records and tenant schedules
- The investment thesis, including the platform and density it assumes
- A short list of candidate assets
The constraints most likely to decide your answer
- Firm supply (grid import and behind-the-meter)
Capacity that cannot be relied on in contingency is not capacity a lender will count.
In core European metros the grid connection is the asset, and it is not growing before the 2030s.
- UPS capacity (protected load)
Protected load is the ceiling on what a tenant can be promised without redundancy being given up.
AI training loads step harder than anything a 2014 UPS was specified for.
- Transformer capacity
Its lead time is the single most common reason a stated uplift arrives later than the model.
The transformer is rarely the tightest constraint and is usually the one that sets the date.
The route through what can be bought
It starts free and each step credits or informs the next; none is required. Prices are read from the live catalogue.
- Density Screen€4,500
5 working days from a complete intake
One asset: a defensible read of what binds it and which inputs the seller has not evidenced.
- Portfolio Screen — deposit€15,000 deposit; opens a scoped engagement priced at €60,000 to €140,000
Five to fifteen assets ranked under one methodology, so the comparison is between halls and not between spreadsheets.
What to ask us before you rely on the answer
- How much of each result is assumed, and what is the weakest evidence class behind the headline figure?
- Is a date in the result dependent on a lead time the seller supplied or one the library assumed?
- Which inputs, if the seller confirmed them, would move the ranking?
What it is not
- It is not a technical due-diligence report and not a valuation. It is an input to both.
- No figure is a forecast of revenue, yield or market growth; the engine does not produce any.
Time to Power sells no equipment, holds no inventory and owns no megawatts. A figure is modelled unless its evidence class says otherwise, and a date that rests on an assumed lead time is labelled as one.