Data centre colocation model
MW by phase, lease-up, rent per kW, power, debt and exit. Build a colocation campus from its megawatts up: capacity by phase and its cost per MW, pre-leasing and absorption to stabilised occupancy, rent per kW with an escalator, power from utilisation and PUE billed at cost plus a margin or sold all-in, NOI, a construction loan converted to a term loan sized on LTV and DSCR, and a sale on forward NOI, then read the yield on cost against the exit cap rate.
By Surojit Chakraverti, founder of L3VLUP and an investor running a long-short healthcare and technology equities strategy.
On this subject: data centre colocation model.Updated 4 November 2026
Who builds it, and for whatThe model a data centre developer-operator’s investment team builds before it commits to a campus, and that an infrastructure fund, a real estate private equity fund or a construction lender rebuilds before backing one. The developer uses it to decide what it can pay for land and a power connection and how much of each phase must be signed before it builds; the lender reads the same cash flows for the term loan it will size at conversion; both run the downside in which tenants sign later than planned and the equity has to fund the gap.
Every sheet below, in your browser, with its outputs and checks
Starter workbook free with an account. Worked model: one a month with Plus, every one with Pro. Downloads
| A | B | C | D | E | F | G | |
|---|---|---|---|---|---|---|---|
| 1 | Assumptions | ||||||
| 2 | $m unless stated. Campus inputs in column C; phase inputs in the three phase columns. All figures are illustrative. | ||||||
| 4 | Input | Unit | Phase 1 | Phase 2 | Phase 3 | Notes | |
| 5 | Scenario and timing | ||||||
| 6 | Scenario | # | 1 | 1 = base, 2 = slower lease-up, 3 = higher power price. Selects a column of the Scenarios sheet. | |||
| 7 | Construction loan converts to the term loan, end of year | year | 8 | After the last phase is built; sized on the next year’s NOI | |||
| 8 | Exit, end of year | year | 10 | Sold on the next year’s NOI; after the conversion and at most Y10 | |||
| 10 | Development budget | ||||||
| 11 | Land, power connection and site works | $m | 60.0 | Paid at close: the substation and grid connection are shared by every phase | |||
| 12 | Build profile, first year | % | 40.0% | Share of a phase’s cost spent two years before it is ready | |||
| 13 | Build profile, second year | % | 60.0% | Share spent in the year before it is ready; the two sum to 100% | |||
| 15 | Leasing and rent | ||||||
| 16 | Stabilised occupancy | % | 95.0% | Of the MW in service; space held back for churn and moves | |||
| 17 | Rent, Y1 | $ per kW a month | 120.00 | Per kW of critical IT capacity leased, whether or not the tenant draws it | |||
| 18 | Rent escalator, annual | % | 2.5% | Applied to every lease from Y1 | |||
| 20 | Power | ||||||
| 21 | Utilisation of leased capacity | % | 70.0% | IT load the tenants draw as a share of the MW they lease | |||
| 22 | PUE (power usage effectiveness) | x | 1.3x | Facility power / IT power: cooling, power conversion and lighting on top of the IT load | |||
| 23 | Hours in a year | hours | 8760 | ||||
| 24 | Power price, Y1, base case | $ per MWh | 80.00 | What the operator pays the utility; the scenario scales it | |||
| 25 | Power price escalation, annual | % | 2.5% | ||||
| 26 | Share of leased capacity billed for power at cost | % | 75.0% | Wholesale contracts: IT energy x PUE at cost plus the margin below | |||
| 27 | Margin on power billed at cost | % | 5.0% | Mark-up on the power passed through | |||
| 28 | All-in power charge, Y1 | $ per kW a month | 60.00 | The rest of the leased capacity: a fixed charge per kW leased, escalating with rent, whatever power costs | |||
| 30 | Operating costs | ||||||
| 31 | Facility operating costs, Y1 | $m per MW a year | 0.12 | Staff, maintenance and security, per MW in service, leased or not | |||
| 32 | Cost inflation, annual | % | 3.0% | ||||
| 33 | Property tax and insurance | % | 0.8% | % a year of development cost built by the start of the year | |||
| 34 | Campus overhead, Y1 | $m | 3.0 | Sales, leasing and management, from the first year a phase is in service | |||
| 35 | Replacement capex, Y1 | $m per MW a year | 0.05 | Batteries, chillers and switchgear; below NOI, per MW in service | |||
| 37 | Construction loan | ||||||
| 38 | Loan-to-cost | % | 60.0% | Drawn pro rata with development cost, including the land at close | |||
| 39 | Interest rate, construction loan | % | 7.5% | Floating, taken as fixed; on the opening balance, paid in cash | |||
| 40 | Arrangement fee, construction loan | % | 1.0% | % of each draw | |||
| 42 | Term loan | ||||||
| 43 | Maximum loan-to-value at conversion | % | 60.0% | Of the campus valued on forward NOI at the valuation cap rate | |||
| 44 | Valuation cap rate at conversion | % | 7.0% | ||||
| 45 | Sizing DSCR | x | 1.4x | Forward CFADS / the first full year’s interest and amortisation | |||
| 46 | Interest rate, term loan | % | 6.0% | Fixed; on the opening balance | |||
| 47 | Amortisation, annual | % | 3.0% | % of the original loan each year; the rest is repaid at the exit | |||
| 48 | Arrangement fee, term loan | % | 1.0% | % of the loan | |||
| 49 | DSCR covenant, minimum | x | 1.2x | Tested each year the term loan is outstanding | |||
| 51 | Exit and returns | ||||||
| 52 | Exit cap rate | % | 7.0% | On NOI in the year after the exit | |||
| 53 | Sale costs | % | 1.5% | % of the sale price | |||
| 54 | Equity discount rate | % | 12.0% | For the equity NPV: the return the developer requires | |||
| 56 | Phases | ||||||
| 57 | Critical IT capacity | MW | 24 | 24 | 32 | The IT load the phase can deliver to tenants | |
| 58 | Development cost per MW | $m per MW | 11.50 | 10.50 | 10.50 | Shell, electrical and mechanical plant and fit-out; the first phase carries more of the shared plant | |
| 59 | Ready for service | year | 3 | 5 | 7 | Energised and able to bill from the start of this year; built over the two years before | |
| 61 | Phase leasing | ||||||
| 62 | Pre-leased at energisation, base case | % | 50.0% | 40.0% | 40.0% | Share of the phase’s MW signed before it is ready | |
| 63 | Absorption, base case | % | 20.0% | 20.0% | 20.0% | Share of the phase’s MW leased in each later year, to stabilised occupancy | |
Select a cell to see its row, its schedule and the type of cell. Font colours: blue for inputs, black for formulas, green for links to other sheets. Scroll sideways to see every period.
Download
Data centre colocation model: the workbook
An Excel workbook with working formulas, no macros and no external links: the same model you can work with on this page.
- Browse without an accountPreview every model online, sheet by sheet, with its outputs and checks.
- Free accountWork with every model online: change the inputs, switch the case, see every formula. And every starter workbook, with its key schedules left blank to build.
- L3VLUP Plus$4.99/monthEverything in Free, and one worked model each billing month as an Excel file: every schedule built, formulas live, checks at zero, yours to keep.
- L3VLUP Pro$25/monthEverything in Plus, every worked model in the library, and the Pro applications.
- The worked model
- 13 sheets and 1,540 live formulas, all 11 schedules built and linked, every check reading zero. Change an input in Excel and the whole model moves.
- The starter
- The same inputs and layout with the data centre power: PUE and pass-through left blank for you to build, each blank cell marked. The worked model is its answer.
What the base case says
- Equity IRR
- 21.4%
- Unlevered IRR
- 14.6%
- Equity multiple
- 3.5x
- Yield on cost
- 8.3%
- Development spread over the exit cap rate
- 1.3%
- Total development cost
- $1,116.8m
- Stabilised NOI, untrended
- $92.8m
- Term loan
- $796.5m
- Released to equity at conversion
- $234.1m
- Minimum DSCR
- 1.4x
- Sale price
- $1,715.6m
- Power margin over the hold
- $19.1m
- Equity NPV at the discount rate
- $185.1m
Read from the workbook as served, every input at its default. Periods: Close, Y1, Y2, Y3, Y4, Y5, Y6, Y7, Y8, Y9, Y10, Y11. The figures are invented and move with whatever you type in.
What this model is
A developer-operator’s model for an invented colocation data centre campus of 80 MW of critical IT capacity, built in three phases, leased to tenants who bring their own servers, financed with a construction loan that converts to a term loan, and sold once it is stabilised.
Each phase is built over two years at a cost per MW and bills from its ready-for-service year: a pre-leased share at energisation, then absorption each year to stabilised occupancy. Tenants pay rent per kW of capacity leased, escalating each year, and pay for power either at cost plus a margin or through a fixed all-in charge. Facility power is the IT load drawn times PUE. NOI pays operating costs, property tax and a campus overhead; the term loan is sized on the next year’s NOI at the lesser of an LTV and a DSCR; the campus is sold on forward NOI at an exit cap rate.
Uses: learning how a data centre earns, why rent is per kW rather than per square foot, why power passed through protects NOI and power sold all-in does not, how lease-up decides the term loan and the equity cheque at conversion, and how yield on cost against the exit cap rate makes the development worth doing. It leases powered space; it does not own the servers or sell compute, which is the GPU cloud model’s business.
Seats: Private equity, Investment banking.
Careers that do this work: infrastructure & real assets, debt capital markets, private credit.
How the schedules connect
Every row in the workbook is tagged with the schedule it belongs to; the inspector above shows the tag when you click a row. These are the schedules this model is made of and where each one lives.
Development budget and draw profile
Land, hard and soft costs, contingency, fee, S-curve
Assumptions, rows 11–13 · Assumptions, rows 57–59 · Development, rows 15–17 · Development, rows 20–22 · Development, rows 25–27 · Development, rows 30–36
Data centre capacity and lease-up
MW by phase, pre-leasing, absorption, rent per kW
Assumptions, rows 16–18 · Assumptions, rows 62–63 · Scenarios, rows 6–9 · Leasing, rows 6–8 · Leasing, rows 11–13 · Leasing, rows 16–18 · Leasing, rows 21–26
Data centre power: PUE and pass-through
IT load, PUE, power cost and how it is recovered
Assumptions, rows 21–28 · Power, rows 6–10 · Power, rows 13–20
NOI build
Gross rent to net operating income
Assumptions, rows 31–35 · Operations, rows 6–15
CFADS
Cash flow available for debt service
Operations, rows 16–17
Construction funding
Draws, interest during construction, fees, equity
Assumptions, rows 38–40 · Financing, rows 6–12
Loan sizing: LTV, DSCR and debt yield
The lender lends to the tightest of three constraints
Assumptions, row 7 · Assumptions, rows 43–45 · Financing, rows 15–22
Debt schedule
Tranches, interest, amortisation and the sweep
Assumptions, rows 46–48 · Financing, rows 25–33
Coverage ratios
DSCR, LLCR and PLCR
Assumptions, row 49 · Financing, rows 36–39 · Returns, rows 32–38
Development returns and spread
Yield on cost, development spread, profit on cost, IRR
Assumptions, row 8 · Assumptions, rows 52–54 · Operations, rows 20–27 · Cash Flow, rows 6–7 · Cash Flow, rows 10–18 · Returns, rows 5–11 · Returns, rows 14–23 · Returns, rows 26–29
Sensitivity tables
Two assumptions at once, without a data table
Sensitivities, rows 5–7 · Sensitivities, rows 10–13 · Sensitivities, rows 17–22 · Grid Workings, rows 6–8 · Grid Workings, rows 11–15 · Grid Workings, rows 18–23 · Grid Workings, rows 26–34 · Grid Workings, rows 37–51
What you should be able to explain
- Why a colocation operator prices capacity in kW of critical IT load rather than square feet, and why rent is due whether or not the tenant draws the power.
- How PUE turns the IT load the tenants draw into the facility’s power bill, and why power passed through at cost protects NOI while power sold all-in does not.
- Why the share of a phase pre-leased before it is energised decides how much equity the campus needs.
- How a term loan is sized at conversion on the next year’s NOI as the lesser of an LTV and a DSCR, and why a slow lease-up means equity is injected rather than released.
- What yield on cost and the development spread over the exit cap rate say about whether a campus is worth building.
- How colocation differs from a GPU cloud, which owns the servers and sells compute by the hour, and from an office development, where rent is per square foot and tenants buy their own power.
What a reviewer looks for
- Revenue that counts only rent, so power billed to tenants is missing from revenue and cost alike and the NOI margin looks like a software company’s.
- Power cost from the capacity built rather than the IT load the tenants actually draw, or without PUE.
- Every phase leased the day it is energised, so the term loan is sized on NOI the campus has not yet earned.
- Yield on cost on the development budget alone, leaving out the construction interest the build costs.
- All-in power pricing treated as a pass-through, so a higher power price appears to cost the operator nothing.
Conventions this workbook uses
Stated on the cover sheet too. A model is only as trustworthy as the decisions it tells you it made.
- Annual periods. Close is financial close, when the land, the power connection and the site works are paid for; Y1 to Y11 follow. Each phase is built in the two years before its ready-for-service year and bills from the start of that year. The campus is sold at the end of the exit year on the following year’s NOI, which is why Y11 is laid out; cash flows stop after the exit.
- Capacity is critical IT load in MW, the power the halls can deliver to tenants’ servers. A phase’s leased share is its pre-leased share in its first year, plus absorption for each year after, capped at stabilised occupancy. Rent is per kW of capacity leased per month, the same for every lease and escalating each year from Y1; a tenant pays it whether or not it draws the power.
- Power: the IT load drawn is leased MW times utilisation; facility load is that times PUE; energy is facility load times the hours in a year. The operator buys all of it. Capacity billed at cost (the pass-through share) pays its share of the power cost plus a margin, so a higher price passes through; the rest pays a fixed all-in charge per kW leased, escalating with rent, so a higher price is the operator’s. Revenue therefore includes the power billed, which lowers the NOI margin without changing the NOI.
- Operating costs per MW in service run whether the space is leased or not; property tax and insurance are a share of the development cost built by the start of the year; the campus overhead starts when the first phase is in service. Replacement capex sits below NOI, so the cap rate and the LTV read NOI before it and the DSCR reads CFADS after it. Everything is before tax.
- A construction loan is drawn at the loan-to-cost with every dollar of development cost, including the land at close; interest is charged on the opening balance and paid in cash, so nothing is circular. At conversion it is repaid by a term loan sized on the next year’s NOI as the lesser of the maximum LTV at the valuation cap rate and the loan the next year’s CFADS covers at the sizing DSCR; the surplus is released to equity, a shortfall injected. The term loan amortises at a share of the original loan a year and is repaid at the sale.
- Total development cost is the development cost plus the construction loan’s interest and fees to conversion. Yield on cost is stabilised NOI, every phase at stabilised occupancy at Y1 rent, power price and costs, over that cost; the development spread is the yield on cost less the exit cap rate.
- The equity IRR grid recalculates the equity cash flow on Grid Workings rather than with an Excel data table, resizing the term loan on each row’s NOI; the yield on cost grid is computed directly from the rows it moves.
What this model does not cover
Where its answer stops, so you know what to add before relying on it.
- Tax and depreciation. Returns are before tax, and there is no income statement below NOI.
- Individual leases. Every tenant pays the same rent per kW on the same escalator; there is no rent roll, no lease expiries, renewals or churn, and no interconnection revenue.
- Monthly timing. A phase bills from the start of its ready-for-service year, and the construction loan is drawn annually.
- Power procurement and grid risk: hedges, renewable power agreements, a delayed energisation or a capacity cap from the utility.
- PUE that changes with load, and the power a part-filled hall draws before it is leased.
- A real company. Equinix and Digital Realty are candidates for a case built from their filings; this model uses invented figures.
Build it yourself
The starter workbook
The Power sheet has been cleared: the IT load the tenants draw from their leased MW, the facility load at PUE, the energy over the year, the power price in the selected case, the power cost, the cost and the billing of the capacity on pass-through at cost plus a margin, the all-in power charges, and the power margin split between the two. Build them so that Operations, Financing and Returns calculate again. The NOI margin, the power margin over the hold and the higher power price scenario show whether the treatment is right.
Blanks: Data centre power: PUE and pass-through. Download it beside the worked model.
The path around this model
Understand it, drill it, read the build, then apply it to a real company.
Build · Lab · ~12 min
DSCR Sculpting
Sculpt debt service to CFADS at a target DSCR, size the loan as its present value, split a payment into interest and principal, and explain why a level repayment profile fails a project lender.
Build · Lab · ~3 min
IRR by Eye
Say the IRR for any multiple and hold before the interviewer finishes the sentence, and the multiple for any IRR.
Read · Guide · 12 min
Project Finance Modelling: Cash Flow Waterfalls, DSCR Sizing and the Debt Sculpt
Read · Guide · 13 min
Real Estate Financial Modelling: Why Corporate Finance Mechanics Break Here
Read · Guide · 10 min
Scenario and Sensitivity Architecture: Building a Model That Can Be Stress-Tested
Apply · Skill
Model Audit
Find the errors in a financial model before someone senior does.
Vocabulary: Yield on Cost, Development Spread, Absorption, Stabilisation, Net Operating Income (NOI), Cap Rate (Capitalisation Rate), Interest During Construction (IDC), DSCR (Debt Service Coverage Ratio), Loan-to-Value (LTV), IRR.
Questions about this model
Why is data centre rent priced per kW?
Because what the tenant buys is power delivered to its servers, with the cooling and backup behind it, not floor space. Here a tenant pays $120 per kW of critical IT capacity a month in Y1, $1.44 million a year for each MW leased, escalating at 2.5% a year, whether or not it draws the power. Across the 80 MW campus at 95% occupancy that is $109.4 million of rent at Y1 prices.
How is power treated?
The operator buys all of it. A MW leased draws 0.7 MW of IT load and, at a PUE of 1.3, 7,972 MWh a year, which costs $0.64 million at $80 a MWh. Three quarters of the leased capacity is billed for that power at cost plus 5%, so a higher price passes through. The rest pays a fixed all-in charge of $60 per kW a month, $0.72 million a year per MW. When the power price is 50% higher, the all-in contracts lose money and the power margin over the hold falls from $19.1 million to minus $10.0 million.
What does the base case return?
A yield on cost of 8.3% on $1,116.8 million of total development cost, a development spread of 1.3 points over the 7.0% exit cap rate. The term loan at conversion is $796.5 million, set by the DSCR, which repays the construction loan and releases $234.1 million to equity. Sold at the end of Y10 for $1,715.6 million, the campus returns an equity IRR of 21.4% and 3.47 times the equity, on an unlevered IRR of 14.6%.
What happens if the lease-up is slower?
With half the pre-leasing and 60% of the absorption, the NOI the term loan is sized on is lower, so the loan is $538.1 million and equity injects $21.7 million at conversion instead of receiving $234.1 million. The equity IRR falls to 12.1%. The yield on cost does not move, because it is measured on a stabilised campus: lease-up risk shows in the return and the equity cheque, not in the yield.
Related models
Infrastructure and project finance · Advanced
GPU cloud infrastructure model
GPU cohorts, contracts and backlog, power, facility debt
Real estate · Advanced
Real estate development model
Budget, construction loan, lease-up, refinance, development spread
Infrastructure and project finance · Advanced
Project finance model
Sculpted debt, reserves, lock-up, coverage ratios
Questions about this model
Discussions open here soon.