Project, Equity and Sponsor IRR: Three Numbers for the Same Asset, and Only One Is About the Asset
One project. One set of operating cash flows. Three returns:
Project IRR (unlevered) 9.12%
Equity IRR (levered) 15.07%
Sponsor IRR (net of development fee) 17.92%
Nothing about the asset changed between those three lines. The first gap is leverage. The second is a fee the sponsor pays itself out of the financing. And the number most often quoted in a marketing document is the third.
Only the first of the three tells you anything about the project. The other two tell you about the capital structure and about how the sponsor is compensated, which are worth knowing and are not the same question.
This post opens Series F with what each measure actually captures, how the gaps arise, and which one to use for which decision.
ℹ️ Note: All figures are labelled assumptions, carried forward from the capital structure built across Series E.
What Does Each One Measure?
| Cash flows | Answers | |
|---|---|---|
| Project IRR | Total project cost out, CFADS in — before any financing | Is this a good asset? |
| Equity IRR | Equity contributed out, distributions in — after debt service | Is this a good investment at this capital structure? |
| Sponsor IRR | The sponsor's own cash, in and out, at its own level | Is this a good deal for me? |
The distinction is not pedantic. A mediocre asset with cheap leverage produces a strong equity IRR. A strong asset with no leverage produces a modest one. Quoting a single number without saying which conflates the quality of the project with the quality of the financing.
Project IRR: The Asset
The discount rate at which the project's total cost equals the present value of the cash flow it generates, before any debt.
Total project cost $271,532,502
CFADS $30,000,000/yr
Asset life 20 years
----------
Project IRR 9.12%
This is the number to compare across projects, across technologies and against a cost of capital. It is unaffected by how the project is financed, which is exactly its value: two sponsors bidding for the same asset with different debt packages will compute the same project IRR and different equity IRRs, and only the first is a statement about the asset.
It is also the number that a capital allocator inside a large developer should be looking at when ranking a pipeline, because the financing is a portfolio-level decision and the asset is not.
Equity IRR: The Capital Structure
The return to the equity actually contributed, after every dollar of debt service.
Equity contributed $67,883,125
Annual debt service (18 years) $19,520,732
Distributions, years 1–18 $10,479,268/yr
Distributions, years 19–20 $30,000,000/yr
----------
Equity IRR 15.07%
The jump from 9.12% to 15.07% is leverage, and the mechanism is simple: the project earns 9.12% on total cost and borrows 75% of that cost at 6.5%. The spread between the asset return and the cost of debt accrues to the equity, amplified by the ratio of debt to equity.
Two properties of that amplification are worth stating because they are frequently forgotten.
It works in both directions. If the project IRR falls below the cost of debt, leverage makes the equity return worse than the unlevered one, and it does so with the same amplification factor. A project at 9.12% levered to 15.07% is a project at 6.0% levered to about 4%.
It is not free. Everything Series E described — the lock-up, the sweep, the reserve carry, the covenant package — is the price of that amplification, and none of it appears in the equity IRR as a separate line. It appears as the distributions that were smaller, later, or blocked.
Sponsor IRR: The Deal
The return on the sponsor's own money, which is not the same as the project's equity.
Equity contributed $67,883,125
Less development fee received at close ($10,000,000)
Net sponsor cash out $57,883,125
Same distributions
----------
Sponsor IRR 17.92%
The development fee, as the sources and uses post established, is funded 75% by the lender. So the sponsor writes a $67.9m cheque and receives $10m back on the same day, of which $7.5m is the bank's money. Its net investment is $57.9m, and 285 basis points of the headline return come from that alone.
Other items move the sponsor number in the same way and belong in the same calculation.
A share of the project sold to a partner changes the sponsor's contribution and its share of distributions, usually in different proportions.
Tax equity takes most of the early tax benefits, so the sponsor's cash profile in a flip structure looks nothing like the project's equity profile — the partnership flip post covered the mechanics.
Back-leverage at the holding company level adds a second layer of debt that the project-level equity IRR does not see at all, and it can add several hundred basis points more.
Management and asset management fees paid by the project to the sponsor are a cost to the project and income to the sponsor.
The result is that a sponsor IRR can exceed the equity IRR by a wide margin without any of the difference being attributable to the project performing better.
None of which makes the sponsor number illegitimate. It is the correct answer to the question the sponsor is actually asking — what did I put in, what did I get out — and a sponsor that has genuinely arranged a structure returning 17.92% on its own capital has created value by doing so. The development fee is compensation for years of at-risk development spend, and back-leverage is a real financing decision with real risk attached.
The problem is only ever presentational, and it is specific: a sponsor IRR quoted without its basis is a number the reader cannot reproduce or compare. Two developers quoting 18% may be describing a project at 9% with an aggressive fee and a project at 12% with none, and nothing in the headline distinguishes them.
The convention that solves it is simple and not widely adopted: state the level at which the return is computed, and disclose the items between it and the project. A single line — "sponsor IRR 17.92%, after a $10m development fee and before back-leverage" — makes the number usable. Without it, the reader's only rational response is to assume the most favourable available treatment has been chosen, which is usually right and is unfair to the sponsors for whom it is not.
Which One for Which Decision?
| Decision | Use |
|---|---|
| Should we build this asset? | Project IRR |
| How should we finance it? | Project IRR vs cost of debt |
| Is this equity investment attractive? | Equity IRR |
| Ranking a development pipeline | Project IRR |
| Comparing to a fund's target return | Sponsor IRR, at the right level |
| Buying a stake from a developer | Equity IRR on your contribution, not theirs |
That last row is the one that causes disputes. A developer selling a stake will quote its sponsor IRR, which includes a development fee it received and a development cost base the buyer is not inheriting. The buyer's return is computed on the price it pays, which is a different number entirely — and that gap is the subject of the farm-down post later in this series.
What IRR Does Not Tell You
Three limitations, each of which matters more in infrastructure than in most asset classes.
It says nothing about scale. A 25% IRR on $2m and a 15% IRR on $200m are not comparable investments, and the IRR cannot distinguish them. That is what multiple of invested capital is for, and it is the subject of the next post.
It assumes reinvestment at the IRR. The internal rate of return implicitly assumes intermediate cash flows are reinvested at the same rate, which for a 15% infrastructure return is usually not achievable. The higher the IRR, the more distorting the assumption.
It is unstable with respect to timing. An IRR rewards early cash disproportionately, which is why a development fee paid at close moves it by 285 basis points while the same $10m paid in year ten would move it barely at all. That is arithmetically correct and it means an IRR can be engineered by moving cash forward without creating any value — the cash sweep post showed the same effect running in reverse.
What About a Mid-Life Exit?
Everything above assumes the asset is held to the end of its life. Most equity is not, and the exit assumption usually matters more than anything in the operating model.
An exit replaces the remaining cash flows with a single terminal value, computed as a multiple or as a discounted cash flow of what the buyer expects to earn. The return then depends on two things the operating model does not contain: when the exit happens and what the buyer pays.
Both are assumptions, and the second is almost always the dominant one. On a twenty-year asset sold in year seven, the terminal value represents thirteen years of remaining cash flow — typically more than half the total value in the model. A return computed on an exit is therefore, to a substantial degree, a return computed on a view about what somebody else will believe in seven years' time.
The merchant tail post made the related point: a terminal value resting on wholesale prices fifteen years out is the least defensible number in most infrastructure models and the one that most often decides whether a deal clears its hurdle.
Two disciplines help.
Report the share of value in the terminal assumption. If seventy percent of the present value sits in a year-seven exit multiple, say so. That single disclosure changes how a committee reads the rest of the analysis.
Test the exit as a yield, not a multiple. An exit at a stated multiple of EBITDA implies an exit yield; comparing that yield to the entry yield says whether the model is assuming value creation or multiple expansion. A model earning its return from the latter is making a market call, not an operating one — and it should be labelled as such.
The Reinvestment Assumption, Quantified
The textbook warning about IRR is the multiple-roots problem: a cash flow stream with more than one sign change can have more than one internal rate of return, and Excel returns whichever its seed converges to.
Worth checking rather than repeating, so here is the check. Take the worked project and add a decommissioning outflow in the final year, at increasing sizes:
Decommissioning $40m → single root at 8.76%
Decommissioning $80m → single root at 8.32%
Decommissioning $150m → single root at 7.46%
Decommissioning $250m → single root at 5.60%
No multiple roots at any realistic magnitude. A sign change is necessary for multiple IRRs but not sufficient, and an infrastructure profile with steady positive interim cash flows and one terminal outflow produces a single root even when the outflow is very large. The textbook concern is real in principle and is largely not a project finance problem.
The reinvestment assumption is a different matter, and it bites hard.
An IRR assumes every interim cash flow is reinvested at the IRR itself. The modified internal rate of return replaces that with an explicit reinvestment rate. On the same project:
IRR 9.12%
MIRR, reinvestment at 5% 6.69%
MIRR, reinvestment at 8% 8.44%
Assuming distributions are reinvested at 5% rather than at 9.12% costs 243 basis points of reported return. That is not a rounding difference, and it is the gap between a number that assumes the sponsor can redeploy every distribution into another 9% asset and one that assumes it cannot.
The practical instruction is not to abandon IRR — it is the market's language and there is no point fighting that. It is to compute MIRR alongside it at a realistic reinvestment rate, and to treat the gap as a measure of how much the headline depends on redeployment the sponsor may not achieve. A fund holding a single asset and distributing to investors has a very different reinvestment reality from a developer rolling capital into the next project.
What Is a Good Project IRR?
A question with no absolute answer and a useful relative one.
The absolute level is set by the market and moves with rates. What does not move is the relationship that determines whether the structure works:
Project IRR 9.12%
Cost of debt 6.50%
Spread 2.62 pp
That spread is what leverage amplifies. A wide spread makes a project financeable at high gearing and produces strong equity returns; a narrow one means leverage does very little and a negative one means it destroys value.
Three things move the required level.
Contract status. A project with a long investment-grade PPA needs a lower return than a merchant one, because the distribution around that return is far narrower. Everything Series C documented about basis, curtailment and capture rate widens that distribution for a merchant asset.
Technology and construction risk. A repeat-build solar project on a proven design carries less execution risk than a first-of-a-kind facility, and the required return reflects it.
Rate environment. As the cost of debt rises, the project IRR required to preserve the same spread rises with it — which is why projects that cleared a hurdle at one point in the rate cycle stop clearing it later without anything about the asset changing.
The useful discipline is to state the hurdle as a spread over the available cost of debt rather than as an absolute number. A 9% hurdle set when debt cost 4% is a very different requirement from the same 9% when debt costs 7%, and a hurdle rate that does not move with rates will either wave through bad projects or reject good ones depending on which way the cycle has gone.
How Do You Build This in Excel?
As three separate cash flow streams from one model, with the bridge between them reported.
The three streams
PROJECT = −Total_Project_Cost, then CFADS each period
EQUITY = −Equity_Contribution, then Distributions each period
SPONSOR = −(Equity_Contribution − Development_Fee − Other_Sponsor_Receipts),
then Sponsor_Share_of_Distributions + Fees_Received
The bridge, which is the output worth publishing
Project IRR 9.12%
+ leverage effect 5.95 pp
Equity IRR 15.07%
+ development fee effect 2.85 pp
Sponsor IRR 17.92%
A reader given three numbers has to work out why they differ. A reader given the bridge can see immediately that nearly nine points of an eighteen percent headline return come from the capital structure and the fee rather than from the asset.
The leverage check
Project_IRR 9.12%
Cost_of_Debt 6.50%
Spread 2.62 pp ← positive, so leverage helps
IF Project_IRR < Cost_of_Debt THEN leverage reduces the equity return
Publish that comparison. A project whose unlevered return is below its cost of debt is being made worse by its financing, and an equity IRR alone will not say so clearly.
The sensitivity that matters most
CFADS −10% → Project IRR 7.68% Equity IRR 10.36% Sponsor IRR 12.55%
CFADS −20% → Project IRR 6.17% Equity IRR 5.41% Sponsor IRR 7.00%
Note what happens at −20%: the equity IRR falls below the project IRR, because the unlevered return has dropped beneath the cost of debt and the amplification has reversed. That crossover is the single most useful thing a levered sensitivity can show, and it is invisible if only one return is reported.
ℹ️ Note: Always label which IRR is being quoted, and at what level of the structure. "The project returns 18%" is not a statement anyone can act on, and in a document with three defensible answers it is closer to a choice than to a fact.
To build the three streams, the bridge and the leverage crossover, prompt Dezzmond with your capital structure and cash flow profile.
What Do Investors Actually Check?
- Which IRR is being quoted, and at what level?
- What is the project IRR, and how does it compare to the cost of debt?
- How much of the equity return is leverage rather than asset performance?
- What is in the sponsor number that is not in the equity number — fees, back-leverage, a partner's contribution?
- At what CFADS level does leverage start working against the equity?
- What is the reinvestment assumption implicitly embedded, and is it realistic?
- What is the absolute amount of value created, not just the rate?
Frequently Asked Questions
What is the difference between project and equity IRR?
Project IRR is computed on total project cost and total cash flow, before any financing. Equity IRR is computed on the equity actually contributed and the distributions received after debt service. The gap is leverage.
Why is the sponsor IRR higher than the equity IRR?
Because the sponsor's own cash position differs from the project's equity. A development fee received at close, fees paid by the project, back-leverage and partner contributions all change the sponsor's net investment without changing the project.
How much does a development fee move the return?
On the worked example, a $10m fee received at close against a $67.9m equity contribution moves the return by 285 basis points — from 15.07% to 17.92% — because IRR rewards early cash disproportionately.
Can leverage reduce the equity return?
Yes. Whenever the project IRR falls below the cost of debt, leverage amplifies in the wrong direction. On the worked sensitivity, a 20% fall in CFADS takes the equity IRR below the project IRR.
Does the multiple-IRR problem affect infrastructure projects?
Rarely. Adding a decommissioning outflow of $40m to $250m to the worked project produces a single root in every case. A sign change is necessary for multiple roots but not sufficient, and typical infrastructure profiles do not trigger it.
How much does the reinvestment assumption overstate returns?
On the worked project, 243 basis points — an IRR of 9.12% against a MIRR of 6.69% at a 5% reinvestment rate. Compute both and treat the gap as a measure of how much the headline depends on redeployment.
Which IRR should be used to rank a development pipeline?
Project IRR. Financing is a portfolio-level decision and the asset is not, so ranking on a levered return confuses the quality of the project with the quality of the debt package available at the time.
Closing: Three Right Answers to Three Different Questions
There is nothing wrong with any of the three numbers. Each is correctly computed and each answers a real question. The problem is that they are usually presented as one number, chosen by whoever is presenting.
A developer marketing a stake quotes the sponsor IRR, because it is the highest. A lender's model shows the project IRR, because financing is what it is deciding about. A fund's investment committee sees the equity IRR at the level it is investing. All three describe the same asset and differ by nine percentage points.
The fix is not to pick one. It is to publish the bridge — 9.12% for the asset, plus 5.95 points of leverage, plus 2.85 points of fee — because that decomposition is what tells a reader how much of the return is the thing they are buying and how much is the structure around it.
It also has a useful disciplinary effect. A bridge showing that two thirds of the headline return comes from leverage and fees is a document that invites the right questions: what happens if the leverage is not available on those terms, and what happens to the fee if the buyer negotiates it.
The next post takes the limitation IRR cannot fix: what a rate of return says about how much money was actually made, and why infrastructure funds report two numbers rather than one.
Sources: Wall Street Prep — Distinctive Features of a Project Finance Model · Ryan O'Connell, CFA — Building a Project Finance Financial Model · Edward Bodmer — Project Finance Exercises · Energy IB Guide — Renewable Energy Valuation: Contracted Cash Flows, Merchant Tail and Yield Frameworks