A project loan is the lower of two tests. On a data centre let for fifteen years, the one written as the sizing covenant is not the one that sizes it.
Northgate’s net operating income would carry 538.1 million of senior debt at a cover ratio of 1.45 times. The loan signed is 427.3 million, because a second test caps it at 65 per cent of cost. The cap refuses 110.7 million, 20.6 per cent of what the cover test would lend, and the 1.45 in the term sheet plays no part in the size of the loan until the cap is raised past 80.08 per cent. What the 110.7 buys is the part worth pricing.
Worked in full in The Project Finance Handbook by Julian R. Sterling, with every figure reproduced in a free workbook.See the book on Amazon →
Northgate is a 96 MW data centre that costs 614.4 million to build and is let for fifteen years on a triple-net lease at 56.00 dollars a kilowatt a month, indexed at 2.5 per cent. Year-one net operating income is 61.61 million. The loan costs 6.10 per cent all in, a swap of 4.10 and a margin of 2.00, and pays a level instalment on a twenty-five year profile over a fifteen-year term, so a balloon falls due in the month the only lease expires.
The lender sizes it twice. The cover test divides year-one income by 1.45, which permits 42.49 million a year of debt service, and multiplies by the annuity factor on the profile, 12.6628: 538.1 million. The cost test takes 65 per cent of the funding requirement. That requirement is itself a function of the loan, since interest during construction, the arrangement fee and the reserve all scale with it, and the model solves the loop on its own sheet: 657.4 million, of which 65 per cent is 427.3. The loan is the lower of the two, so the cost test binds, and by a wide margin.
The level instalment on 427.3 million is 33.75 million, so year-one cover is 1.826 times, not the 1.45 the covenant describes. Cover is lowest in year one, because the rent indexes and the instalment does not, and it only rises from there.
Raise the cap and the loan follows it, while the cover test stays exactly where it was. The table runs the cap up with nothing else changed.
| Loan to cost | Loan | Funding requirement | Equity | Year-one cover | Balloon | Equity IRR |
|---|---|---|---|---|---|---|
| 65 per cent | 427.3 | 657.4 | 230.1 | 1.826 | 247.2 | 16.60% |
| 70 per cent | 463.3 | 661.9 | 198.6 | 1.684 | 268.0 | 17.75% |
| 75 per cent | 500.1 | 666.8 | 166.7 | 1.560 | 289.3 | 19.20% |
| 80 per cent | 537.4 | 671.8 | 134.4 | 1.452 | 310.9 | 21.16% |
| 80.08 and above | 538.1 | 671.9 | 133.8 | 1.450 | 311.3 | 21.19% |
Two things in that table are easy to miss. The funding requirement rises with the loan, by 14.4 million between the first row and the last, because a bigger loan carries more interest during construction, a bigger fee and a bigger reserve; so the crossover is not 538.1 divided by 657.4 but 538.1 divided by 671.9, which is 80.08 per cent. And the equity return climbs steadily all the way up, from 16.60 to 21.19 per cent, while the cheque falls from 230.1 million to 133.8. On the sponsor’s own spreadsheet every step up the cap looks like a gain.
On this deal the 1.45 times in the term sheet is not the sizing constraint anywhere between 65 and 80 per cent of cost. The number that decides how much the lender advances is the cost cap, and a sponsor negotiating the cover ratio is negotiating a term that does not bind.
The cover test is a forecast: it believes year-one income and lends against it. The cost test distrusts the forecast and lends against what was spent. The difference between the two loans is the price of that distrust, and it can be read off the model in four places.
| Measure | Loan at 65% of cost | Loan at the cover test |
|---|---|---|
| Senior debt (million) | 427.3 | 538.1 |
| Rent at which cover reaches 1.00, debt held ($/kW-month) | 31.81 | 39.40 |
| Fall in rent that takes cover to 1.00 | 43.2% | 29.6% |
| Loan life cover ratio at close | 1.619 | 1.286 |
| Balloon at the end of the term (million) | 247.2 | 311.3 |
| Rent at which a 65% lender refinances the balloon ($/kW-month) | 27.58 | 33.78 |
| as a share of the expiring rent of 79.13 | 34.9% | 42.7% |
The second and third rows are the operating cushion. With the loan at 427.3 the rent can fall 43.2 per cent before income only just meets the instalment; at 538.1 it can fall 29.6 per cent. The cap buys 13.6 points of rent. On a signed lease that cushion is not for a weaker market; it is for a tenant that pays less than it signed for, or stops paying, which is the one event a single-tenant building cannot diversify away.
The loan life cover ratio says the same thing over the whole term. Fifteen years of contracted income discounted at the loan rate is 691.7 million. Against 427.3 that is 1.619 times; against the loan the cover test would have allowed, 1.286. The model does not print a covenant floor, because none follows from the arithmetic, but any floor between the two is cleared by one loan and failed by the other.
The last three rows matter most on a single-tenant building, because they are about the day the lease ends. A twenty-five year profile over a fifteen-year term leaves a balance outstanding at maturity, and a larger loan leaves a larger one: 311.3 million instead of 247.2. For a lender advancing 65 per cent on the same exit cap to take that balloon out, the building has to be worth 478.9 million rather than 380.3, and a new tenant has to pay 33.78 dollars a kilowatt a month rather than 27.58. The expiring rent is 79.13. The cap lets the rent fall 65.1 per cent at renewal and still refinance; the cover-sized loan, 57.3 per cent.
Read the sizing section of a project term sheet from the bottom up. The cover ratio is the covenant everyone discusses, because it is the one that tests the forecast, and on a highly contracted asset with a single strong tenant it is often not the constraint at all. When the cost test binds by 20 per cent, the useful questions are about the cap, the cost it is measured against, and the balloon it leaves:
None of this argues that 65 per cent is right. It argues that on Northgate the cost cap is the covenant that sizes the loan, that it is worth 4.60 points of equity return to the sponsor, from 16.60 to 21.19 per cent, and that the lender is paid for it in 13.6 points of rent cushion and 7.8 points of refinancing headroom. That is a trade to be made deliberately, and it is only visible when both tests are on the page.
Everything above is in Northgate_Model.xlsx. On Debt sizing, B9 is the loan the cover test supports (538.1), B13 the loan the cost test supports (427.3) and B17 names the one that binds; B20 is the level instalment, B21 the balloon, B26 the loan life cover ratio at close and B27 the same ratio had the loan been sized to the cover test (1.286). B28 is the rent that refinances the balloon (27.58).
For the first table, change the maximum loan to cost on Assumptions, B21, and read the loan, the funding requirement on Sources and uses, minimum cover in Returns B5 and the equity return in B22. The Solve sheet re-runs the funding loop and should still read CONVERGED in B26. The book’s figures on the Checks sheet stop matching, as they should, while the identities that hold for any inputs stay at OK. Row 28 on Debt sizing takes the refinancing lender’s advance from the same B21, so to keep that lender at 65 per cent while the construction loan moves, apply its formula by hand: the balloon over 0.65, times the 7.25 per cent exit cap, plus 4.20 million of year-sixteen cost, divided by 1.152.
For the cushion, set the stress switch on Assumptions B34 to YES, put the loan in B35 and cut the rent in B9: cover on Cash flow C13 reaches 1.00 at 31.81 with the signed 427.3 and at 39.40 with 538.1. With the switch at NO the loan re-sizes to every rent and the cover ratio never falls below its target, which is why a sensitivity cannot show this and a stress case can.
The lower of the two. On Northgate, a 96 MW data centre, a 1.45 times cover test on year-one income supports 538.1 million and a 65 per cent loan-to-cost cap supports 427.3, so the cap binds and year-one cover is 1.826 times, not 1.45.
At 80.08 per cent on Northgate. The funding requirement grows with the loan, from 657.4 million to 671.9, so the crossover is 538.1 divided by 671.9, not by 657.4. Below it the 1.45 cover ratio plays no part in the size of the loan.
Cushion. With the loan at 427.3 million instead of 538.1, rent can fall 43.2 per cent before cover reaches 1.00 rather than 29.6, the loan life cover ratio is 1.619 rather than 1.286, and the balloon refinances at a rent of 27.58 rather than 33.78. The sponsor gives up 4.60 points of equity return for it.
This article is one calculation from The Project Finance 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 1706 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.