Behind-the-meter generation is priced on the energy the campus uses, and the redundant engines are paid for whether they run or not.
On-site gas power costs a data centre the fuel (heat rate times delivered gas price) plus variable O&M plus the plant's annualised capital and fixed O&M, divided by the megawatt-hours the campus actually uses. For an illustrative 100 MW IT campus on eight 20 MW engines, with gas delivered at $4.00 per MMBtu, that is $78.00 per MWh: $35.20 of fuel, $8.00 of variable O&M and $34.79 of capital and fixed O&M, against an illustrative grid price of $70. Paying for capacity that sits idle, not fuel, is what pushes the cost above the grid.
Worked in full in The Data Center Development Handbook by Julian R. Sterling, with every figure reproduced in a free workbook.See the book on Amazon →
The question comes up whenever the utility's energisation date is later than the tenant's: bridge or permanent behind-the-meter generation, usually reciprocating engines or aeroderivative turbines, sized to carry the whole campus. The load below is the illustrative campus used in the annual electricity calculation. Plant costs and fuel prices are illustrative, not market quotes.
| Input | Value |
|---|---|
| Critical IT capacity, MW | 100 |
| Utilisation / annual PUE | 70% / 1.30 |
| Peak facility load, MW | 130 |
| Annual consumption, MWh | 797,160 |
| Unit size, MW | 20 |
| Units: 7 for load plus 1 spare (N+1) | 8 |
| Installed capacity, MW | 160 |
| Installed cost, $ per kW | 1,400 |
| Life and cost of capital | 20 years, 9% |
| Heat rate, MMBtu per MWh | 8.8 |
| Gas commodity plus transport, $ per MMBtu | 3.50 + 0.50 |
| Variable O&M, $ per MWh | 8 |
| Fixed O&M, $ per kW-year | 20 |
Fuel per MWh = heat rate x delivered gas price = 8.8 x $4.00 = $35.20
Capital recovery factor = r / (1 - (1 + r)-n) = 0.1095 at 9 per cent over 20 years
Annual capital = 160,000 kW x $1,400 x 0.1095 = $224.0m x 0.1095 = $24.54m
Capital per MWh = $24.54m / 797,160 MWh = $30.78
Fixed O&M per MWh = 160,000 x $20 / 797,160 = $3.20m / 797,160 = $4.01
Total = 35.20 + 8.00 + 30.78 + 4.01 = $78.00 per MWh
In Excel: =-PMT(9%, 20, 224) gives the annual capital of 24.54, and =8.8*(3.5+0.5)+8+(24.54+3.2)*1e6/797160 the total.
The denominator is the campus's consumption, not the plant's capability. An island plant has to be sized for the peak hour plus a spare unit, and it only sells what the campus draws: 797,160 MWh against 160 MW installed is a plant utilisation of 56.9 per cent. Every idle megawatt is still paid for.
| Component | $ per MWh |
|---|---|
| Fuel | 35.20 |
| Variable O&M | 8.00 |
| Capital recovery | 30.78 |
| Fixed O&M | 4.01 |
| Total | 78.00 |
At $78.00 the plant is $8.00 per MWh above the illustrative $70 grid price, about $6.37m a year on this campus, on a total annual cost of $62.2m. Whether that is expensive depends on the alternative. If the choice is between on-site power now and grid power in three years, the comparison is not $78 against $70: it is $78 against the rent forgone while the hall sits empty, which is a very different number.
Fuel is the variable input; each $1.00 per MMBtu moves the cost by $8.80 per MWh, the heat rate.
| Gas commodity, $ per MMBtu | Delivered | Cost per MWh |
|---|---|---|
| 2.50 | 3.00 | 69.20 |
| 3.50 | 4.00 | 78.00 |
| 4.50 | 5.00 | 86.80 |
| 5.50 | 6.00 | 95.60 |
On these assumptions the plant matches the grid at a delivered gas price of $3.09, or $2.59 of commodity. Redundancy moves the answer more than gas does. Tenants used to a 2N electrical design sometimes ask for the same on generation: 14 units, 280 MW installed for a 130 MW peak. Capital rises to $392.0m, plant utilisation falls to 32.5 per cent and the cost becomes $104.09 per MWh, $26.10 more than N+1. That uplift is worth the equivalent of a $2.97 rise in the gas price, and it is a design choice rather than a market risk.
Island mode is the expensive case. A plant that runs alongside a grid connection can be smaller, because the grid is the spare; one that bridges to an energisation date can be leased and returned, which turns most of the capital line into a shorter rental.
Two inputs deserve diligence beyond the cost line. Firm gas transport, not interruptible capacity, is what lets the plant run through a cold snap when local distribution companies curtail industrial customers, and it is priced into the transport charge. And the heat rate should be the one at the load the engines will actually carry: with eight units sharing a 91 MW average load, most run well below their rating for much of the year, and part-load heat rates are worse than the data sheet.
The usual error is to spread the plant's capital over its own output at high availability: 160 MW x 8,760 hours x 95 per cent is 1,331,520 MWh, which gives $64.03 per MWh and makes on-site gas look cheaper than the grid. That denominator is energy the campus cannot use. The plant was built to cover the peak hour and a failed unit, and its capital has to be recovered from the 797,160 MWh actually consumed. The difference, $13.96 per MWh, is the whole answer to whether the project clears the grid price.
Price on-site gas per MWh consumed, with the capital of the full redundant fleet in the numerator. On this campus that is $78.00 at N+1 and $104.09 at 2N. Generation and grid timing feed straight into the development return; the book's 180 MW case, with its twelve-month energisation delay costed, is in the free workbook for this case, and the same capital recovery logic applied to a solar plant is in how to calculate LCOE.
Use the heat rate at the part load the units will actually run at, not the full-load figure on the data sheet. The illustrative case uses 8.8 MMBtu per MWh, so each $1.00 per MMBtu of gas adds $8.80 per MWh. A plant running at 56.9 per cent utilisation spends much of the year below full load, where heat rates are worse.
Rarely on energy cost alone once redundancy is paid for. The illustrative N+1 plant costs $78.00 per MWh against a $70 grid price, and only matches it at $2.59 of gas commodity. The case for it is time: rent earned while the grid connection is still years away can outweigh an $8.00 per MWh premium.
For a 130 MW peak on 20 MW units, N+1 needs 8 units and 160 MW, while 2N needs 14 units and 280 MW. Capital rises from $224.0m to $392.0m and the cost per MWh consumed from $78.00 to $104.09, an uplift of $26.10 that equals a $2.97 rise in the gas price.
This article is one calculation from The Data Center Development Handbook. The book takes the same case from first principles to the decision, chapter by chapter, and every figure it prints is a live formula in the free companion workbooks.
Get the book on Amazon →Free companion files
Also on Amazon UK · Amazon Germany · Amazon France · Amazon Canada
Reading guide: infrastructure, data centres and energy → · All 453 articles →
If this book helped, or didn’t, a few lines on Amazon are worth more than they look: they are what the next reader goes on. Write a review. The workbook stays free either way.