A financial PPA settles on the generator's output and price, while the data centre pays for its own load; the gap between the two is the real price.
A virtual PPA does not deliver power: the data centre keeps buying from its utility, and each month the generator pays it (market price minus strike) times the megawatt-hours generated, or receives the difference when the price is below the strike. On an illustrative 100 MW campus with a 200 MW solar contract struck at $55, the hedge covers 54.9 per cent of annual consumption and narrows the all-in power cost range from $50 to about $22.5 per MWh. It does not fix the price at $55: on the hedged volume the effective price is the strike plus the gap between the load price and the solar capture price, $75 in the base case.
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 load is the illustrative campus used in the annual electricity calculation: 100 MW of critical IT at 70 per cent average utilisation and an annual PUE of 1.30. The contract is a financially settled solar PPA, sometimes called a contract for differences, at the generator's hub. Every figure is illustrative.
| Input | Value |
|---|---|
| Critical IT capacity, MW | 100 |
| Average utilisation | 70% |
| Annual PUE | 1.30 |
| Annual consumption, MWh | 797,160 |
| Contracted solar capacity, MW | 200 |
| Capacity factor | 25% |
| Annual generation settled, MWh | 438,000 |
| Strike price, $ per MWh | 55 |
Two prices matter, and they are not the same number. The load price is what the data centre pays its supplier or the market at its own node, averaged over a flat 24-hour load. The capture price is the hub price averaged over the hours the solar farm actually generates, weighted by output. Midday prices in a solar-heavy market sit below the daily average, so the capture price is usually lower. The base case assumes a load price of $70 and a capture price of $50.
Settlement to the data centre = (capture price - strike) x generation
Net power cost = consumption x load price - settlement
Net cost per MWh = net power cost / consumption
Base case: settlement = (50 - 55) x 438,000 = -$2.19m, so the data centre pays the generator. Bill = 797,160 x $70 = $55.80m. Net = 55.80 + 2.19 = $57.99m, or $72.75 per MWh.
In Excel, with hourly generation in column B, hub price in C, load in D and load price in E: =SUMPRODUCT(B:B,C:C-55) for the settlement and =(SUMPRODUCT(D:D,E:E)-settlement)/SUM(D:D) for the net cost per MWh.
The annual version above uses averages; a real settlement runs hourly or by settlement interval, which is why the weighted capture price, not the simple average hub price, goes into the formula.
| Year | Load price | Capture price | Utility bill | Settlement | Net cost | Net $ per MWh |
|---|---|---|---|---|---|---|
| Low | 50 | 30 | 39.86 | -10.95 | 50.81 | 63.74 |
| Base | 70 | 50 | 55.80 | -2.19 | 57.99 | 72.75 |
| High | 100 | 80 | 79.72 | 10.95 | 68.77 | 86.26 |
Unhedged, the power cost per MWh would swing from $50 to $100, a range of $50. With the contract it swings from $63.74 to $86.26, a range of about $22.5, which is 45.1 per cent of the unhedged range. The hedge removes exactly the share of consumption it covers, 54.9 per cent, because in this example the capture price moves one for one with the load price.
The data centre pays $10.95m in the low year. That is the hedge working, not a loss: the utility bill fell by more. Boards that see only the settlement line in a cheap-power year tend to ask why the company signed it.
The one-for-one assumption is the weak point. As more solar connects to a market, midday prices fall relative to the daily average, and the gap between load price and capture price widens. Each dollar of extra gap costs the data centre $0.44m a year on 438,000 MWh. Holding the load price at $70:
| Capture discount, $ per MWh | Capture price | Settlement, $m | Net $ per MWh | Effective price on hedged MWh |
|---|---|---|---|---|
| 10 | 60 | 2.19 | 67.25 | 65.00 |
| 20 | 50 | -2.19 | 72.75 | 75.00 |
| 30 | 40 | -6.57 | 78.24 | 85.00 |
| 40 | 30 | -10.95 | 83.74 | 95.00 |
The effective price on the hedged volume is the strike plus the capture discount: $55 plus $20 is $75. A $55 strike is a $55 price only for a buyer whose load has the same shape as the solar farm and sits at the same node. A flat data centre load never has that shape, and the further the campus is from the generator's hub, the more basis risk sits on top.
To match consumption over a year, the contracted capacity has to be consumption divided by capacity factor times 8,760 hours: 797,160 / (25% x 8,760) = 364 MW of solar for 100 MW of IT. That is annual volume matching. Hour by hour the campus would still be short every night and long every sunny afternoon, so a contract sized for 100 per cent annual matching over-hedges the middle of the day and leaves the night open. Developers chasing hourly carbon-free targets add wind, storage or firm clean capacity for exactly that reason.
The frequent error is to model a virtual PPA as a fixed price of $55 on the contracted volume. That figure ignores shape and basis. In the base case it understates the true cost of hedged energy by $20 per MWh, and in a market where solar cannibalisation pushes the capture discount to $40, by $40. The second error is to size the contract on IT load rather than facility load: IT energy here is 613,200 MWh, so sizing on it leaves the 183,960 MWh of cooling and conversion overhead uncovered.
Compute the settlement on generation and capture price, net it against the bill on consumption and load price, and quote the result per MWh consumed. On this campus a 200 MW solar contract cuts the power cost range by 54.9 per cent and costs about $75 per hedged MWh, not $55. Power procurement sits alongside site selection and grid interconnection in the book; the levered return on its 180 MW campus is rebuilt in the free workbook for this case, and the tariff side of the bill is in the demand charge per MWh.
A physical PPA delivers the generator's power to the data centre's meter, so the buyer takes the output and its shape. A virtual PPA is a financial swap: the data centre keeps its utility supply and settles (market price minus strike) on the generator's output. On the illustrative 200 MW solar contract that settlement is -$2.19m in a base year and $10.95m in a high-price year.
Divide annual facility consumption by the capacity factor times 8,760 hours. At 70 per cent utilisation and a PUE of 1.30 the campus uses 797,160 MWh, so at a 25 per cent capacity factor it needs 364 MW of solar to match on an annual basis. Hour by hour it is still uncovered at night.
When the hub price during solar hours is below the strike, the buyer pays the difference to the generator. In the illustrative low-price year the capture price is $30 against a $55 strike, so the data centre pays $10.95m, while its utility bill falls by $15.94m against the base year. The net cost still falls, to $63.74 per MWh.
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.