NPV and IRR — how companies decide if a project is worth funding
The answer
A company decides whether to fund a project by asking 1 question: will this return more than the money costs us? Net present value answers it in rupees or dollars, and internal rate of return answers it as a percentage. Both are ways of comparing money that arrives later against money that leaves today.
Why this costs you money
A company you own announces a ₹12,000 crore expansion. The stock rises. The press release uses the words "capacity", "growth" and "market leadership".
Four years later the capacity exists, revenue has grown, and the share price has gone nowhere. Nothing was fraudulent. The company spent ₹12,000 crore to earn a return of 7% on it, and the money it used cost about 12%. Every rupee of that expansion made you poorer, and every quarterly report described it as growth.
This is the most expensive thing that happens to long-term shareholders, and it is almost invisible. A company destroying value by growing looks identical, on the surface, to a company creating value by growing. Revenue rises in both cases. Assets rise in both cases. The only difference is a comparison that never appears in the press release.
Growth is not the same as value. A company that grows while earning less than its cost of capital is shrinking your wealth in a way that takes years to show up in the share price, and by then you have held it for years.
How it works
Money later is worth less than money now
₹100 today is not equal to ₹100 in 3 years. If you can earn 10% a year, ₹100 today becomes ₹133 in 3 years. So ₹100 arriving in 3 years is worth about ₹75 today. That conversion is called discounting, and the rate you use to do it is the discount rate.
The discount rate is the company's cost of capital
This is the part that decides everything, and the part most explanations rush past.
A company's money comes from lenders and from shareholders. Lenders charge interest. Shareholders charge nothing visible, but they expect a return, and if they do not get it they sell and the share price falls. Blend the 2 and you get the cost of capital — the return a company must earn just to leave its owners no worse off.
A rough version, which is enough for everything in this article:
Cost of capital ≈ the yield on 10-year government bonds + about 5 to 6 percentage points for the risk of owning a business instead of a bond.
If the 10-year government bond yields 7%, a typical company's cost of capital is somewhere around 12 to 13%. If it yields 4%, around 9 to 10%. Companies with steadier earnings sit below the range. Companies with volatile earnings sit above it.
Everything a company earns below that number is a loss dressed as a profit.
Net present value
Take every rupee the project will produce, discount each one back to today, add them up, and subtract what the project costs.
NPV = (all future cash flows, discounted to today) − (the cost today)
If the answer is positive, the project earns more than the money costs, and doing it makes the owners richer by roughly that amount. The rule is short: positive, do it; negative, do not.
Internal rate of return
Now ask the question backwards. Instead of choosing a discount rate and finding the value, find the discount rate that would make the value exactly zero. That rate is the internal rate of return, or IRR. It is the project's own return, expressed as a percentage.
The rule is equally short: if IRR is above the cost of capital, do it. IRR is popular because a percentage is easy to compare and easy to say in a meeting. "This project returns 18%" is repeated more often than "this project has a net present value of ₹340 crore".
Where the 2 disagree
They usually agree. When they disagree, NPV is right.
The common case is size. A small project can have a spectacular IRR and add almost nothing. Spending ₹10 crore to earn 40% adds ₹4 crore a year. Spending ₹1,000 crore to earn 15% adds ₹150 crore a year. IRR prefers the first. Your wealth prefers the second.
The second case is timing. IRR quietly assumes every rupee the project produces can be reinvested at the same high rate. Usually it cannot.
Use IRR to describe a project. Use NPV to choose between projects.
What it tells you, and what it does not
An NPV calculation tells you what a set of assumptions is worth. It does not tell you whether the assumptions are true.
Every NPV has 3 inputs: the future cash flows, the discount rate, and how long it goes on. All 3 are estimates, and 2 of them are estimates about a decade away. Change the discount rate by 2 percentage points and a long project's value can move by a third. Change the assumed long-run price of a commodity by 15% and the answer flips sign. The number at the bottom of the model has a precision the inputs do not support.
So NPV does not produce truth. What it does is force the assumptions into the open. Somebody has to write down what they believe about prices, volumes and time, which means those beliefs can be argued with now and checked later.
The version that helps you as an outside investor is different, and better. You cannot see a company's NPV models. You can see the results of every NPV decision it has already made, added together, in its published accounts.
The decision rule
You cannot audit a company's forecasts. You can audit its record.
Compare the return the company earns on all the capital it has already deployed against a rough cost of capital.
If the return is comfortably above the cost of capital and has stayed there for several years, then this management has been making positive-NPV decisions, whatever their spreadsheets looked like.
If the return is below the cost of capital and the company keeps investing, then growth is making you poorer, and more growth will make you poorer faster.
Unless the company is in the middle of a large project that has not started earning yet. Money spent on a half-built factory counts as capital today and earns nothing today. Check the capital work-in-progress line before you conclude anything.
Try this now
Five minutes, 1 company you own, 1 annual report. This is the single most useful calculation in this cluster.
- Search for "[company name] annual report" and open the latest PDF from the company's own website. Go to the consolidated financial statements.
- From the Statement of Profit and Loss, write down Profit before tax and Finance costs. Add them. That is your operating profit, close enough. In a US filing, use Operating income directly from the income statement.
- From the Balance Sheet, write down Total equity and Total borrowings (add long-term and short-term). Then find Cash and cash equivalents and any short-term investments, and subtract those.
Total equity + Total borrowings − Cash = capital employed. - Divide.
Operating profit ÷ Capital employed = return on capital employed, usually shortened to ROCE. Express it as a percentage. - Now look up the current yield on the 10-year government bond — search "India 10 year bond yield" or "US 10 year Treasury yield". Add 5 percentage points. That is your rough cost of capital.
- Put the 2 percentages side by side.
What you should see. One of 3 pictures.
ROCE clearly above the cost of capital, by 5 points or more. Every rupee this company reinvests adds value. Growth is genuinely good news here, and a decision to hold back cash instead of paying it out is probably in your favour.
ROCE close to the cost of capital, within about 2 points. The company earns back roughly what its money costs and no more. Expansion announcements are neutral events. Read them as neither good nor bad until you know the return on the specific project.
ROCE clearly below the cost of capital. Every expansion announcement is bad news for you, no matter how it is written. If this company also has rising capital work-in-progress, it is currently spending money that will make the gap wider.
Do this for 2 companies in the same industry and the comparison becomes sharper still. Same customers, same input costs, same regulations — and often a 15-point gap in ROCE, which is entirely a difference in how well the 2 managements decide what to fund.
One caution. Skip banks, insurers and finance companies for this test. Capital means something different in a lending business, and ROCE calculated this way will mislead you.
Three real cases
1. Reliance Jio, from 2016 (India) — a very large bet with a very long payback Reliance built a national telecom network from nothing and launched commercial service on 5 September 2016, initially giving the service away. The capital committed ran into tens of billions of dollars, the largest single project the company had undertaken. On any short-term measure it looked indefensible: huge spending, no revenue, entrenched competitors. Jio's first quarterly net profit came in the December 2017 quarter, announced in January 2018. This is what a genuine NPV decision looks like from outside — years of negative cash flow accepted because the discounted value of a decade of later cash flow was judged larger. Note what it required: an assumption about how many Indians would use mobile data 10 years out. That assumption, not the arithmetic, was the decision.
2. Tata Steel and Corus, from 2007 (India and United Kingdom) — the assumption failed, not the maths Tata Steel completed the acquisition of the Anglo-Dutch steelmaker Corus in April 2007 for about £6.2 billion, winning a bidding contest to do it. The price implied a view about European steel demand and prices for many years ahead. The financial crisis arrived the following year and European steel demand never recovered to the assumed level. In the 2012-13 financial year Tata Steel wrote down about $1.6 billion, mainly against those European operations, and in 2016 it sold its UK long products business for a nominal amount. The arithmetic in 2007 was almost certainly correct. The input was not. NPV is only ever as good as the number somebody typed into the cell marked "long-term price".
3. ExxonMobil and XTO Energy, 2009 to 2020 (United States) — a discount rate cannot save a wrong forecast Exxon announced the acquisition of natural gas producer XTO Energy on 14 December 2009, in a deal valued at about $41 billion including debt. The logic was a forecast of rising US natural gas prices. Instead, shale drilling expanded and gas prices fell for most of the following decade. In the fourth quarter of 2020 Exxon recorded impairment charges of $19.3 billion, mostly against dry gas assets in the United States. One of the most disciplined capital allocators in the world, using every technique in this article, was wrong about 1 number for 10 years.
The question that resolves it
A novice reads a capital expenditure announcement and asks: how big is it?
An expert asks: what return does the company earn on the capital it already has, and why would this be different?
The past return is a fact you can calculate. The future return is a claim. A company that has earned 22% on capital for a decade is making a claim supported by evidence. A company that has earned 8% for a decade is making the same announcement and contradicting its own history.
What would make this wrong
If capital allocation did not matter, then companies in the same industry with the same growth rates would produce the same returns for shareholders. They do not, and the gap between the best and worst allocator in most industries is larger than the gap between the industries themselves.
The honest limits are 4, and they are serious.
First, ROCE is a backward-looking measure applied to a forward-looking question. A company can earn a high return today on assets built cheaply 20 years ago, and be unable to repeat it at today's construction costs.
Second, accounting distorts it. A company that has bought other companies carries goodwill on its balance sheet, which inflates capital employed and depresses ROCE. A company that has written that goodwill off looks better having destroyed more value.
Third, the calculation punishes companies mid-investment. Money spent on a factory that opens in 2 years sits in capital employed today and earns nothing today.
Fourth, a low return is not always a bad decision. A regulated utility may be allowed to earn only a set return and may still be a reasonable holding, because the return is certain. The test is whether the return beats the cost of that company's capital, which for a regulated monopoly is lower than for a miner.
In India
Indian companies are not required to publish project-level NPV or IRR calculations, and almost none do. What you can find is scattered across 4 places in the annual report, and it is enough.
- Management discussion and analysis. Companies describe planned capacity additions here, usually with a rupee figure and a completion date.
- Capital work-in-progress, on the balance sheet. Money already spent on assets not yet operating. A large and rising figure means a large bet is in flight. Compare it against total capital employed.
- The cash flow statement, under investing activities. "Purchase of property, plant and equipment" is the actual capital spending for the year, and it is harder to arrange than any figure in the profit and loss statement.
- Related party transactions, in the notes. In a group with many companies, this shows where money moved inside the group. It is the most useful page in an Indian annual report and it is near the back.
Two Indian features shape capital decisions. First, most listed Indian companies have a controlling promoter family, so the decision to fund a large project is often made by an owner rather than a board answering to dispersed shareholders. That can mean long horizons and real conviction. It can also mean nobody in the room is able to say no.
Second, the cost of capital in India is structurally higher than in the United States, because the risk-free rate is higher. A project earning 11% can be value-creating in the United States and value-destroying in India, on identical operations. When you read American writing about acceptable returns, translate the numbers before you use them.
For money raised in an IPO, SEBI requires companies to state the objects of the issue in the prospectus and to appoint a monitoring agency to report on how the proceeds were actually used. Those reports are filed with the exchanges. If a company raised money for a stated purpose, you can check whether it did it.
In the United States
American filings give you more, in 3 places.
- The 10-K annual report carries capital expenditure plans and, in many industries, segment-level asset and profit figures. Segment data lets you calculate a return for each part of the business separately, which often reveals that 1 division is subsidising another.
- Earnings calls. Analysts ask directly about hurdle rates and expected returns on new projects, and the transcripts are public. Some companies state a specific hurdle — a minimum IRR a project must clear to be approved.
- Capital allocation frameworks. Large US companies increasingly publish an explicit ranking: maintenance spending first, then growth projects above a stated return, then dividends, then buybacks. Where this exists, you can hold management to a standard they wrote themselves.
The discipline is also external. Activist investors buy stakes in companies they believe are investing badly and campaign publicly for the money to be returned instead. That mechanism barely operates in India.
Where they differ, and what that tells you
The mechanical difference is the risk-free rate, and it changes the arithmetic. An Indian company faces a higher hurdle for the same project, because the money costs more. That is a reason Indian companies should reject projects American companies accept, and a reason to be suspicious when an Indian company describes a 10% return as attractive.
The more instructive difference is who says no.
In the United States, a large capital project is scrutinised by a board, by analysts on a public call, and sometimes by an activist investor who will run a campaign if the numbers do not work. Several independent parties can stop a bad project.
In India, most large capital decisions are made inside a controlling family or by a government owner. The only party who can stop a bad project is inside the company. When the controlling owner is a skilled allocator, this produces decisions with a 15-year horizon that no quarterly-reporting American company could make. When the controlling owner is wrong, there is often nobody with the standing to stop them.
What that tells you is where to look. For a US company, judge the process: the stated hurdle rate, the segment returns, whether management does what its own framework says. For an Indian company, judge the person: what has this promoter actually earned on capital over 10 years, across every business they have started? Past allocation matters more in India precisely because there is less external correction on the next decision.
Carry this
- Money later is worth less than money now, and the discount rate is what the company's money costs.
- NPV in currency chooses between projects. IRR in percent describes 1. When they disagree, NPV wins.
- You cannot check a company's forecasts. You can check its record: return on capital employed against roughly the 10-year bond yield plus 5 points.