Colocation customers rarely draw all the power they contract for. The gap is stranded capacity, and selling part of it again turns it into revenue at the price of a risk the operator has to measure.
A 36.0 MW campus has 32.4 MW under contract, but customers peak at only 67% of what they bought. That leaves 10.69 MW stranded. Selling half of it again at $145 per kW per month earns $9.30m a year and lifts contracted capacity to 37.75 MW. The campus stays safe until the customers' peak ratio passes 95.4%.
A course built on one 60 MW campus, from grid tranches and cost per MW to leases, the power recharge, the construction loan, the refinancing and the stress tests, with Excel exercises, a 162-question quiz and the Data Center Development Model, Pro Edition included ($199).
| Item | Value |
|---|---|
| Critical capacity of the campus | 36.0 MW |
| Capacity under contract | 32.4 MW (90%) |
| Peak draw of customers, share of contracted | 67% |
| Average draw of customers, share of contracted | 58% |
| Price of contracted capacity | $145 per kW per month |
| Share of the unused gap the operator sells again | 50% |
Customers pay for what they contract, not for what they draw, so the operator already earns 32.4 × 1,000 × $145 × 12 = $56.4m a year. The question is what to do about capacity that is contracted but never used.
Average draw = 32.4 × 58% = 18.79 MW
Peak draw = 32.4 × 67% = 21.71 MW
Stranded at peak = 32.4 − 21.71 = 10.69 MW, or 29.7% of the whole campus
Excel: =32.4*(1-67%)
Capacity resold = 10.69 × 50% = 5.35 MW
Revenue = 5.346 × 1,000 × 145 × 12 = $9.30m a year
Contracted capacity becomes 32.4 + 5.35 = 37.75 MW, or 104.9% of the campus
Peak draw becomes 37.75 × 67% = 25.29 MW, 70.2% of capacity
Excel: =32.4*(1-67%)*50%*1000*145*12/1000000
The campus is contracted above 100% but its peak draw stays at 70.2% of capacity. The resale works only while customers keep drawing a small share of what they have bought. The revenue is also the least secure on the campus, because the new customers have the same rights to ramp up as the old ones.
| Peak draw ratio | Stranded gap | Resold | Extra revenue | Contracted after | Peak draw after |
|---|---|---|---|---|---|
| 55% | 14.58 MW | 7.29 MW | $12.68m | 39.69 MW | 21.83 MW (60.6%) |
| 67% | 10.69 MW | 5.35 MW | $9.30m | 37.75 MW | 25.29 MW (70.2%) |
| 80% | 6.48 MW | 3.24 MW | $5.64m | 35.64 MW | 28.51 MW (79.2%) |
| 90% | 3.24 MW | 1.62 MW | $2.82m | 34.02 MW | 30.62 MW (85.0%) |
The lower the peak ratio, the more there is to sell and the more room is left. The break-even is how far the peak ratio can rise before the campus cannot serve everyone: with 37.75 MW contracted it is 36.0 ÷ 37.75 = 95.4%. At 80% the peak draw would be 30.20 MW, and at 90% it would be 33.97 MW, 94.4% of capacity, which leaves no margin for a cooling or distribution fault.
| Price per kW per month | Extra revenue a year |
|---|---|
| $100 | $6.42m |
| $120 | $7.70m |
| $145 | $9.30m |
| $170 | $10.91m |
Free Excel template: data center financial model.
Power that is contracted to customers but never drawn. Here customers contract 32.4 MW and peak at 21.71 MW, which leaves 10.69 MW.
In this example half of the gap is 5.35 MW. At $145 per kW per month that is $9.30m a year.
The risk that customers draw more at the same time. With 37.75 MW sold the campus runs out of power if the peak ratio passes 95.4%.
The free companion files for The Data Center Operations Handbook rebuild a 36 MW colocation campus month by month for three years after handover, with the operating checklists in editable form. The customer contracts in this article are invented and are not in them. No account and no email address.
Not on Amazon: 19 paperbacks by Julian R. Sterling on energy, infrastructure and land, printed on demand and shipped worldwide. Browse the bookshop →
Reading guide: infrastructure, data centre and energy investing → · All 1908 articles →