Energy mandate · CAP-05
Power first. The compute follows.
Land is available, buildings can be built and GPUs can be ordered. Energised megawatts on a date you can plan against are the scarce good, and every other part of an AI infrastructure plan is downstream of them.
What an energy mandate covers
- —Queue and constraint analysis
Where power is genuinely available on your timeline, which is a different map from where land and fibre are. Interconnection queue depth, transmission constraint and substation headroom, by market.
- —Utility engagement
Tariff class, large-load interconnection agreements, and what the utility will actually commit to in writing versus what it has indicated verbally.
- —Supply structuring
Physical and virtual PPAs, sleeved arrangements, retail supply contracts and hedging — matched to your risk appetite rather than a standard template.
- —Bridge and behind-the-meter
On-site generation, storage and bridge power to get a facility earning while the permanent interconnection completes.
How buyers actually get power sooner
The interconnection queue is a public constraint, which means everybody is looking at the same list. Advantage comes from the options around it.
- Flexibility as a bargaining chip. A load willing to curtail during system peaks is a materially easier load for a utility to accommodate, and AI training is more curtailable than most industrial demand. Offering flexibility deliberately, in exchange for an earlier date or a better tariff, is the most underused lever in this market.
- Existing industrial sites. Retired or shrinking industrial and heavy manufacturing sites frequently retain interconnection capacity that would take years to establish new. So do some converted crypto mining sites, though the substation and cooling reality varies enormously.
- Behind-the-meter and bridge generation. On-site gas, fuel cells or storage to carry initial load while the grid connection completes. It changes the economics and the permitting picture, and it is not right everywhere, but it can move an in-service date by years.
- Phased energisation. Contracting for capacity in tranches that match utility delivery rather than a single date, so the first tranche is earning while the rest builds.
- Geography. The blunt lever. ERCOT, the Nordics and parts of LATAM have materially different queue dynamics and power prices from Northern Virginia or the FLAP-D core, and for latency-tolerant training work the tradeoff is often obvious once it is quantified.
- Tariff class and demand charge structure. Two facilities on the same grid with the same draw can pay very different bills depending on how the load is classified and how demand charges are calculated. This is a permanent operating-cost difference and it is set early.
How much of this is available to you depends heavily on jurisdiction. Interconnection speed and the willingness to entertain behind-the-meter generation differ sharply between US markets, Canadian provinces and European grids — and where power is the gate on a new build, it is the development mandate that has to be structured around it.
Asked on this mandate.
01Is power really the binding constraint on AI data centers?
In most major markets, yes. Space and hardware can be arranged on commercial timelines; energised capacity is set by utility and transmission timelines that no amount of budget compresses. Any credible capacity plan starts with the power position and works backwards.
02Can you get us a better interconnection date?
Not by moving a queue. What we can do is find the sites and markets where the position is already better, structure the load so the utility can accommodate it sooner, and build a bridge strategy so the facility earns before permanent power lands.
03Do you negotiate PPAs?
We advise on structure and negotiate terms alongside your team and your counsel — physical, virtual and sleeved. We hold no position in any generation asset or offtake, so we have no preferred structure to sell you.
04Which markets do you cover for power?
ERCOT and the wider US including PJM and MISO, the Nordics, the UK and Ireland, continental Europe, APAC and Latin America. Power price and queue dynamics vary so sharply between them that the comparison is usually the exercise.
05Should we consider a converted crypto mining site?
Sometimes. Retained interconnection capacity is genuinely valuable, but substation configuration, cooling, structure and network position frequently are not fit for AI workloads without substantial rebuild. The diligence matters more than the headline megawatts.
Powered shell & build-to-suit
Site selection and delivery structure for buyers who need capacity that does not exist yet.
Open the mandate →CapacityData center capacity sourcing
Retail and wholesale colocation up to 150 kW per rack, sourced globally and benchmarked against real comparables.
Open the mandate →ComputeGPU cluster procurement
H100, H200, GB300 and the supply agreements around them — allocation, delivery, and remedies when the vendor slips.
Open the mandate →Start with the power question.
Tell us the load and the date, and we will tell you where in the world that is actually achievable.