Renew, retrofit or re-let: value each option per MW from the expiry date, and the rent of the next lease turns out to matter far more than the cost of the works.
Value each lease-end option per MW from the expiry date and compare: renew at a reset rent, retrofit the halls to liquid cooling and re-lease, or re-let to smaller tenants. On an illustrative hyperscale hall, a $2.6m per MW retrofit re-leased at $168 per kW-month is worth $26.16m per MW against $24.17m for a renewal, a gain of $1.99m per MW or $63.7m on 32 MW. It pays as long as the new rent clears $157.73, about today's passing rent. But about 70 per cent of the gain rests on a keener exit cap rate.
Worked in full in The Digital Infrastructure Investor by Julian R. Sterling, with every figure reproduced in a free workbook.See the book on Amazon →
A data hall built for air-cooled racks faces a binary question when its first lease ends: will the next generation of chips fit it? A tenant that renews takes the hall as it is, usually at a discount to its passing rent. A retrofit to liquid cooling costs real money and a year of empty halls, but it buys a new long lease at a higher density rent. The arithmetic below puts the two on the same footing.
| Input | Renew | Retrofit | Re-let |
|---|---|---|---|
| Rent, $ per kW-month | 142.67 (90% of 158.52 passing) | 168.0 from year 10 | 120.0 from year 10 |
| Capex in year 9, $m per MW | 0.4 | 2.6 | 0.8 |
| Year 9 | let | empty, half running costs | empty |
| Occupancy | 100% | 100% | 55% then 85% |
| Running costs | base | base | 1.25x base |
| Exit cap rate, sale at end of year 18 | 6.25% | 5.75% | 7.25% |
Common to all: rent escalates 2 per cent a year, landlord running costs are $0.162m per MW in year 9 plus a $0.051m reserve, both indexed at 3 per cent, management is 1 per cent of rent, exit costs are 1.5 per cent, and cash flows are discounted at 7.5 per cent. Energy is passed through to the tenant and left out. The inputs follow the lease-end assumptions of the book's illustrative case; the per-MW valuation here is a simplified version of it.
Value at expiry = Σ (NOIt − capext) ÷ 1.075t−8 + (NOI19 ÷ exit cap × (1 − 1.5%)) ÷ 1.07510
Rent in $m per MW-year = rent per kW-month × 12 × 1,000 ÷ 1,000,000. Passing rent of 158.52 is $1.902m per MW a year.
In Excel, with years 9 to 18 of NOI less capex in C5:L5 and the net sale in L6: =NPV(7.5%,C5:L5)+L6/1.075^10
| Option | Capex | Year-10 NOI | Year-19 NOI | Net sale, year 18 | Value per MW | 32 MW |
|---|---|---|---|---|---|---|
| Renew | 0.40 | 1.509 | 1.779 | 28.03 | 24.17 | 773.3 |
| Retrofit to liquid cooling | 2.60 | 1.776 | 2.098 | 35.94 | 26.16 | 837.0 |
| Re-let to several tenants | 0.80 | 0.522 | 1.106 | 15.03 | 11.93 | 381.8 |
| Stranded: powered land less demolition | 1.225 | 39.2 |
Against a renewal, the retrofit adds $1.99m per MW, $63.7m on the hall. Against the outcome the owner actually fears, the tenant leaving, the comparison is not close: retrofit beats re-letting to colocation tenants by $14.22m per MW, and re-letting beats a stranded hall by $10.71m. The retrofit decision is really two decisions. If the tenant will renew in the old halls, the retrofit is a modest improvement. If it will not, the retrofit is what keeps the building a hyperscale asset rather than a multi-tenant one worth less than half as much.
| Retrofit rent, $/kW-month | Value | Gain over renewal | Retrofit capex, $m/MW | Value | Gain over renewal |
|---|---|---|---|---|---|
| 150 | 22.67 | -1.50 | 2.0 | 26.71 | 2.55 |
| 160 | 24.61 | 0.44 | 2.6 | 26.16 | 1.99 |
| 168 | 26.16 | 1.99 | 3.5 | 25.32 | 1.15 |
| 180 | 28.48 | 4.31 | 4.5 | 24.39 | 0.22 |
The rent matters far more than the cost. The break-even retrofit rent is $157.73 per kW-month, 0.5 per cent below the passing rent at expiry: the retrofit has to win back almost the old lease's rent, 10.6 per cent above the renewal rent. The cost, by contrast, can rise to $4.74m per MW, 82 per cent over budget, before the retrofit loses to a renewal. Owners spend months negotiating the construction contract; the number that decides the outcome is the new tenant's rent.
Crediting the retrofit with a keener exit yield and not noticing how much of the answer it is. Value the retrofitted hall at the same 6.25 per cent exit cap as the renewal and its advantage falls from $1.99m to $0.59m per MW: about 70 per cent of the gain is the assumption that a buyer in year 18 will pay more for liquid-cooled halls. That may be true, but in ten years the liquid-cooled hall will itself face the next generation of chips. Show the retrofit case at both exit yields, and decide on the version that does not need the yield to move.
The case is illustrative and ignores the energy margin, phasing of the works and tax. Retrofit costs, rents and cap rates for liquid-cooled capacity vary widely; nothing here is investment advice.
The four lease-end outcomes of Halvard DC1, with its debt and the full equity cash flows, are in the free workbook for this case. How much of the purchase price the first lease covers is worked in how much of a data centre's price the signed lease covers.
It varies widely with the existing plant, the density targeted and how much electrical work is needed. The illustrative case assumes $2.6m per MW plus a year of empty halls. Cost matters less than rent: the retrofit still beats a renewal at up to $4.74m per MW, 82 per cent over that budget.
The owner chooses between re-letting to several smaller tenants, retrofitting for a new single tenant, or mothballing. In the illustrative hall, re-letting is worth $11.93m per MW against $26.16m after a retrofit and $1.225m as powered land less demolition, so a hall that can be retrofitted keeps most of its value.
Buyers expect higher-density, liquid-cooled halls to stay lettable to the next generation of tenants for longer, so they pay more per dollar of NOI. The illustrative case uses 5.75 per cent against 6.25 for a renewed air-cooled hall, and that half point accounts for about 70 per cent of the retrofit's $1.99m per MW advantage.
This article is one calculation from The Digital Infrastructure Investor. 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 1340 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.