Four questions to ask anyone advising you.

A buyer entering AI infrastructure for the first time meets a lot of genuinely helpful, genuinely knowledgeable people. Operators with capacity. Integrators with reference architectures. Vendors with roadmaps. Consultancies with frameworks. Advisors of every description.

Almost every one of them is compensated on the basis that a transaction happens. That is not a scandal — it is simply how the market is built, and it is true of us too. What matters is not whether an intermediary is paid, but whether the way they are paid changes what they tell you.

Four questions settle it. None of them are hostile, all of them are quick, and anyone straight with you will answer them without hesitation.

1. Do you hold inventory, listings or resale rights in anything you would show me?

This is the question that determines whether a shortlist is advice or a sales document. An intermediary with a position in one of the options has a reason to prefer it, and that reason will not appear in the deck.

The follow-up is the useful part: is there any counterparty on this list whose selection pays you more than the others? If the answer is yes, you are not looking at a shortlist. You are looking at a preference.

2. Who pays you, and does the amount change depending on which counterparty I choose?

Two separate questions, and the second is the one that matters. Being paid by the market is normal and often better for you than paying a fee — winning business carries a cost of sale that exists whether or not an advisor is involved. The problem is variable compensation across options, because that quietly ranks the options before you see them.

3. How many parties will know about this requirement by the end of next week?

The most underrated question of the four, and the one most buyers never think to ask.

Broad exposure has real uses. It produces fast price discovery in a liquid market, and for a standardised requirement below a certain scale it is the sensible route. But it has a cost: once counterparties know the shape of the field — who else is looking, at what scale, on what timeline — your position is common knowledge and your leverage goes with it. The same applies in reverse to an operator whose availability has been broadcast; softness invites the market to wait you out.

So the answer does not have to be "one." It has to be a number you chose on purpose, rather than one you discovered afterwards.

4. Are you still involved after signature, and are you paid for that?

Most intermediation ends at the close. The parties then discover the operational reality of the terms alone — the ramp that was optimistic, the SLA remedy that turned out to be a gesture, the expansion right nobody secured.

An advisor retained on an ongoing basis has a structural reason to care whether a deal still works in year three. One paid per transaction does not. Neither is dishonest; they are just aimed at different horizons, and you should know which one you have engaged.

What the answers actually tell you

Nothing on this list has a correct answer. A party with inventory can still be the right party to buy from. Wide circulation can be the right strategy. Transaction-based compensation is appropriate for transactional work.

The point is to know which instrument you have picked up, so you can weight the advice accordingly — and so you are not surprised later by an incentive that was visible from the start.

You are not looking for someone with no incentives. You are looking for someone whose incentives you can see.

Our answers, for the record

In order: no — we hold no inventory, no listings and no resale rights in anything we bring you. The ecosystem, and no — we are not paid more for one counterparty over another, which is the whole reason a shortlist from us is worth reading. Only the parties you approve — we approach counterparties directly, one at a time, and anonymous first contact with an MNDA is available before anything specific is shared. Yes, and yes — we are retained on an ongoing basis, so a deal that falls apart in year three is our problem as well as yours.

We are also transparent about the uncomfortable half of that: being paid by the market is a real incentive, and the only thing that makes it workable is structural. We hold no position in the outcome, we do not earn more from one configuration than another, and we are useful to this market only for as long as both sides keep taking our calls. Reputation is the only collateral we have, which makes protecting it the entire job rather than a value statement.

If you are weighing a live decision, the relevant mandates are capacity sourcing, GPU procurement, GPUaaS negotiation, buy-side advisory and, for operators, capacity commercialization.