The Capacity Desk
High-density AI colocation, sourced globally.
Retail, wholesale and high-density AI-ready colocation up to 150 kW per rack — sourced globally, benchmarked against what this market is actually paying, and negotiated on your behalf at no cost to you.
Requirements from a single cabinet to campus scale. Above roughly 40 kW per rack the list of operators who can genuinely deliver narrows sharply, and knowing which claims survive a site visit is most of what the desk is for.
Where high-density colocation gets difficult
Above roughly 40 kW per rack the list of operators who can genuinely deliver narrows sharply. Knowing which claims survive a site visit is most of what this desk is for.
40–70 kW per rack
Achievable in many modern facilities with contained aisles and rear-door heat exchangers. The constraint is usually cooling redundancy at full draw rather than the electrical path, and it is the question operators are least often asked.
70–150 kW per rack
Direct-to-chip liquid cooling territory. A materially shorter list of operators, and the gap between designed-for and deployed is where most of the disappointment lives. We ask what is running today, at what density, and for how long.
What we verify
- Deliverable density in the specific hall, not the site's best case
- Cooling headroom at your full draw with a unit down
- Whether liquid cooling is deployed or only designed for
- Floor loading and structural provision for the rack spec
- Water position and any regulatory exposure in that jurisdiction
The full diligence list is here.
Densifying instead of leasing
Where power exists at the meter but the cooling cannot reject the heat, a retrofit often unlocks capacity you are already paying for. Our partner OptiCool Technologies builds two-phase refrigerant cooling for exactly that case — water-free at the rack, and non-invasive enough to deploy against live workloads.
Three steps, one business day to the first.
- 01
Send the specification
Megawatts or racks, density per rack, cooling method, region, latency tolerance, redundancy tier and the date you need it live. Anonymous is fine, and MNDA is available before anything is shared.
- 02
We return a shortlist and a benchmark
Operators who genuinely hold what you need, in markets where power is available on your timeline, with a view on what the market is actually paying for comparable capacity. You see the reasoning behind every name, including the ones we ruled out.
- 03
We negotiate, you contract
We advise from behind, lead alongside your team, or run the table outright — your choice, reversible at any point. You contract directly with the operator and own that relationship outright.
Power availability decides the map.
Not demand, and not the market nearest you. For latency-tolerant training work, geography is usually worth more than every clause in the contract combined.
Northern Virginia, Dallas–Fort Worth and ERCOT, Phoenix, Atlanta, Chicago, Columbus, plus secondary markets that frequently beat all of them on time-to-energisation.
Iceland, Norway and Sweden — the strongest cost-per-megawatt case in Europe for latency-tolerant training.
London and Slough for latency-bound inference; Ireland where grid position allows it.
Frankfurt, Amsterdam and Paris, with Nordic and Iberian alternatives when FLAP-D power is constrained.
Singapore, Johor and wider Malaysia, and Japan.
Brazil, Chile and Mexico — renewable-heavy grids and a favourable cost basis.
Have capacity to fill?
The desk runs both ways. We qualify demand before we introduce it, never the reverse — so what reaches your desk is a requirement that is real, funded and ready to move.
- —Qualified offtake sourcing
We establish the requirement, the budget and the decision authority before an introduction is made.
- —KYC and compliance
Ultimate offtaker identified, ownership verified, sanctions and AML screening run — worked through alongside your compliance team before anything reaches your credit committee.
- —Capacity commercialization
Anchor tenants and term structure for new builds and expansions.
- —Nothing circulated
Your availability, pricing floor and commercial position stay with us. We approach demand directly.
Asked at the desk.
01What rack density can you actually source?
Up to 150 kW per rack, with air, rear-door heat exchanger and direct-to-chip liquid cooling depending on the facility and region. Above roughly 40 kW the list of operators who can genuinely deliver narrows sharply, and above 70 kW it narrows again.
02What does this cost the buyer?
Nothing directly. Winning business carries a cost of sale that exists whether or not we are involved, and that is where we are paid from. You contract directly with the operator at the rate we negotiated.
03How quickly can capacity be identified?
First response within one business day, and a credible shortlist inside the first week for most requirements. Delivery is set by the operator and the utility — above a few megawatts, energisation date is usually the binding constraint rather than available space.
04Can you help if we already have a proposal?
Yes, and it is a common entry point. We benchmark what you have been offered, tell you where it sits against the market, and identify which terms are actually movable. Mid-negotiation is not too late.
Send the specification. One business day to a first response.
Megawatts, density, region and date. Anonymous first contact is fine and MNDA is available on request.