Growth Capex and Maintenance Capex: What Separates Them

Maintenance capital expenditure keeps the earnings a business already makes where they are. Growth capital expenditure buys earnings that do not exist yet. The gap between keeping earnings and buying new ones decides which of the two carries a return question at all: maintenance is tested by what is lost if it is skipped, and growth by whether the return on the money committed beats what the existing capital already earns.

Two spending decisions land on the desk of Devyani Kulkarni, who is chief financial officer at Harivansh Packaging Limited, in the same week. One is an overhaul of the boilers at a plant that has been running for eleven years. The other is a new line that would let the plant make a film it cannot make today. Both are capital spending. Both take cash out this year and put an asset on the books. In the published accounts they will sit in the same total, added together, with no label separating them.

And they are not remotely the same decision. Getting them confused is one of the two or three most expensive habits in reading a set of accounts. Some things can honestly be said about the split between them, and some things cannot.

What is maintenance capital expenditure?

Start away from companies altogether. Think about an auto driver in a city, running one vehicle, twelve hours a day. Tyres wear out. The clutch plate wears out. Every so often the engine needs work that costs a serious fraction of a month's earnings. None of that spending makes the auto carry more passengers than it carried before. All of it is the price of the auto continuing to carry as many as it carried before.

Maintenance capital expenditure is that spending in its entirety. The money replaces what wears out, and next year's capacity, and therefore next year's earnings, then match this year's. Maintenance buys continuity. Maintenance buys nothing new.

The uncomfortable thing about it is what it looks like from the outside while it is being skipped. Maintenance spending buys nothing new, so a business that stops doing it reports more cash for a while and then reports less of everything. The driver who postpones the clutch plate has a better month. He has a better month again. Then, at some point that no one warned him about, he has three days with no auto at all, and after that the vehicle never quite runs the way it did.

Notice how badly the timing works. The saving is immediate, exact and easy to point at. The cost is delayed, spread out, and never labelled as the consequence of the saving. Nobody writes down that the auto broke because the clutch plate was postponed sixteen months ago. The auto simply broke.

The same asymmetry runs through a company. Harivansh Packaging Limited charges depreciation and amortisationThe accounting spreading of what a long-lived asset cost, written off a bit at a time across the years the asset is expected to be used. Depreciation is a bookkeeping allocation, not money going out of the door. of Rs 138 crore against a revenue of Rs 3,180 crore. The charge is an allocation of what the assets cost when they were bought. The charge carries on whether or not a single rupee of replacement spending happens. Stopping the replacement changes the cash. Stopping the replacement does not change the charge. A year of skipped maintenance therefore improves cash and leaves profit almost untouched, and improved cash beside untouched profit is precisely what good management looks like from the outside.

What is growth capital expenditure?

Back to the auto driver. Suppose that instead of a clutch plate he buys a second vehicle and hires a driver for it. Now something genuinely different has happened. There is capacity that did not exist before. There are passengers being carried that were not being carried. There is an amount of money committed, and an amount of extra earning that came out of it, and those two amounts can be set against each other.

Growth capital expenditure is spending that adds capacity, a product, a location or a line. Next year's earnings can then exceed this year's. The new film line at Harivansh Packaging Limited is exactly that: after it is installed the plant can make something it could not make before, and can sell it.

Growth spending is the only capital spending that produces an increment to measure a return against, and so the only capital spending a return calculation can be applied to. That sentence sounds like a technicality and it is the hinge of the whole distinction. A return is a ratio. The ratio needs a numerator that is extra earnings and a denominator that is the extra money committed to get them. Maintenance produces neither. Maintenance produces the absence of a decline. The absence of a decline is real and valuable, and it is not a number sitting anywhere in the accounts.

Where do the two part company, criterion by criterion?

Six tests separate them. Across all six, the distinction stops being a definition and becomes something that can be applied to a real set of numbers.

Each row carries a separate idea. Take them one at a time.

The two pull apart on five criteria and are identical on the sixth, and the sixth is exactly why the arguments about the other five are so often conducted with no evidence at all. Everything that matters about the distinction is invisible in the one place everybody looks.

One row is the one people get backwards, and it deserves a second look. Growth is the exciting half, so deferring growth spending sounds worse than deferring maintenance. Deferring growth is not worse. Postponing a new line costs earnings that were never there. Postponing an overhaul costs earnings that are already there. Only one of those is a loss in any ordinary sense of the word, and it is not the one that gets the meeting.

Why does only one of them carry a return question?

Here is a question that sounds sensible and is not: what return does maintenance spending earn?

The question sounds sensible. Every other use of capital has an answer to it. A purchase has a return. A new plant has a return. A repayment has an avoided cost that behaves like one. So the mind reaches for the same question and asks it of the overhaul.

The reason it fails is that a return compares an outcome against an alternative, and for maintenance the alternative is not a lower return. The alternative is the loss of the return the existing capital already makes. At Harivansh Packaging Limited, capital employedThe money tied up in a business and available for it to use, taken from the funding side as net worth plus borrowings. Settled in the earlier material and used here as a base to strike a return on. of Rs 2,390 crore is already earning 14.2 per cent, and that figure already includes the earnings of every asset the boilers help run. There is no overhaul to have a return, so skipping the overhaul does not produce a lower return on the overhaul. Skipping it produces a hole in the 14.2 per cent.

So the two kinds get two different questions, and the questions are not interchangeable.

• Maintenance: what is lost if this is not done? That is a question about capacity, output, safety and continuity, and its answer is usually a description rather than a rate.
• Growth: does the return on the money committed clear the hurdle rateThe minimum return a company decides a new commitment must beat before it is worth making. It was settled earlier in this sequence and is applied here rather than rebuilt. of 14.2 per cent that the existing capital already earns? That is a question about ratio, and it has a numeric answer.

Using one question for both is how a company talks itself into cutting the wrong half of its capital spending. Ask what the overhaul returns and the honest arithmetical answer is that no return can be computed. Under pressure that answer reads as a return of zero, and a return of zero reads as the first thing to cut. The mistake is not in the arithmetic. The mistake is in having asked.

Can the split be computed from what Harivansh Packaging publishes?

No. The answer being no is itself the finding, and it is worth more than a manufactured proportion would be.

Look at what the record for Harivansh Packaging Limited actually carries. Revenue, Rs 3,180 crore. Earnings before interest, tax, depreciation and amortisation, or EBITDAWhat a business earns before interest, before tax and before any writing down of assets comes off. Settled in the earlier material, and used here only as the top of the ladder. , Rs 477 crore. Depreciation and amortisation, Rs 138 crore. Finance cost, Rs 60 crore. Tax, Rs 75 crore. Profit after tax, Rs 225 crore, on 18.00 crore shares. A balance sheet carrying borrowings of Rs 740 crore against cash of Rs 140 crore.

Now look at what it does not carry. There is no figure for capital spending in the year. There is no cash flow statement. Without a capital spending total there is nothing for maintenance to be a proportion of, so the growth and maintenance split cannot be computed from this record at all.

The temptation runs the other way, and the missing total is worth sitting with. Writing that maintenance is "around" the depreciation charge and that growth is "the rest" is very easy. Both halves of that claim would be made up. The first is a stand-in with known problems, examined next. The second needs a capital spending total the record never gives.

A fair objection at this point: surely a real listed company has to publish its capital spending somewhere? Publication requirements are a separate question, and the answer sits with the two regulators named below.

Why is depreciation used as a stand-in, and where does that stand-in fail?

Everybody uses depreciation as a stand-in for maintenance spending. There is a good reason and a bad one, and it is worth separating them.

The good reason is that the two really are aimed at the same thing. Depreciation is an attempt to say how much of the asset base got used up in the year, and maintenance spending is what it costs to put that back. Depreciation and maintenance spending describe the same physical fact from two sides.

The bad reason is that it is the only published number that means anything like it, so it gets used whether or not it is any good. On this record it is the only one available at all.

The stand-in fails in three ways, and knowing the ways matters more than knowing that it fails.

Failure one: replacement happens at today's price

Depreciation spreads what an asset cost when it was bought, its historical costWhat an asset actually cost on the day it was acquired, which is the figure the accounts carry it at and spread over time, regardless of what the same asset would cost today.. Replacing it happens at whatever the same asset costs now. On any asset base bought over a stretch of years in which prices rose, the charge is struck on the smaller figure and the spending happens at the larger one. The auto driver bought the vehicle at one price and replaces the clutch at this year's price, and nobody adjusts his mental depreciation for the gap.

Failure two: the life in the books need not be the working life

An asset's useful lifeThe number of years over which an asset's cost is spread in the accounts. The assumed life is set when the asset is recorded, and it need not match how long the asset actually keeps working. is an assumption made when the asset goes on the books. How long the machine actually keeps producing is a physical fact discovered later. Where the assumed life runs longer than the working life, the annual charge is spread too thin, and it sits below what replacement will really cost per year. Where the assumed life is shorter, the charge runs ahead of the spending instead. So this one has a direction only once it is settled which of the two is longer.

Failure three: replacement described as expansion

The third failure is not accounting at all, it is language. A line that was going to have to be replaced anyway gets replaced with a slightly better one, and the project is described internally, and sometimes externally, as an expansion. Nothing dishonest has necessarily happened. But spending that was going to be needed to stand still has now been counted in the half that is supposed to represent moving forward.

Two of the three failures understate maintenance with no condition attached, and the third does so whenever the working life is shorter than the assumed one, so nothing in the set pushes the other way on its own and the errors accumulate rather than cancelling. That is why the stand-in tends to understate rather than to scatter, and why whatever residual gets called growth comes out systematically too large. An error that scatters can be lived with; an error that leans has to be named.

The worked instance: what can honestly be said about this capital spending

Devyani Kulkarni is preparing a note for the board on how much of the year's cash is genuinely free. Ashwin Rege, who leads the transaction team, wants a figure. Here is what can be built, and where it stops.

Start with what is missing, because that is the finding

The record carries depreciation and amortisation of Rs 138 crore for Harivansh Packaging Limited and carries no capital spending figure at all. So the growth and maintenance split cannot be computed. That sentence goes in the note, first, and it is the most valuable line in it. Every reader further down the chain now knows what standard of evidence the rest of the note is built on.

What can be done with the stand-in

Take the Rs 138 crore as the maintenance claim, with its three failure modes stated, and set it against what the business actually generated.

Two honesty flags on that table before anything is done with the Rs 204 crore. The Rs 204 crore is a construction, not a published figure and not a free cash flowA measure of the cash a business has left after paying for what it must to keep operating. Several definitions exist and they differ, so any figure called by this name needs its build shown. figure from a statement. This record carries no cash flow statement at all. Other income of Rs 21 crore is not operating, so the table leaves it out, along with the gap between a tax charge and cash tax paid, and any movement in working capital. The omissions are choices, they are stated, and a reader who disagrees with them can rebuild the table.

Now stress the stand-in in the one direction that is arithmetic

The first failure mode says replacement costs more today than the historical amounts being depreciated. Suppose it costs 1.6 times more. The 1.6 multiple is chosen to show what an uplift does to the answer, and a real one would come from asking what the same assets cost to buy today.

Rs 138 crore multiplied by 1.6 is Rs 220.8 crore. Rs 342 crore less Rs 220.8 crore is Rs 121.2 crore. So Rs 82.8 crore of room disappears, a fall of 40.6 per cent in the capital available for everything else.

The free capital falls by 40.6 per cent, from Rs 204 crore to Rs 121.2 crore, without a single figure in the published accounts changing. The uplift is therefore an assumption that has to be stated in the open rather than buried in a model. Anyone who reads the Rs 204 crore without being told what maintenance was assumed to cost has been handed a number whose largest single input is invisible.

The growth side, where a return question does exist

Tapti Crossing Infrastructure Private Limited is growth capital spending in its purest available form: a single toll crossing, built from nothing, with no existing business underneath it. Project cost Rs 1,800 crore. Annual EBITDA Rs 248 crore. Because there is nothing there but the new asset, the increment is the whole thing, and a return question applies cleanly.

Rs 248 crore on Rs 1,800 crore is 13.8 per cent, and that is struck before depreciation. The comparison worth making is against the 14.2 per cent the existing capital at Harivansh Packaging Limited already earns. The 14.2 per cent is struck after depreciation, on capital employed of Rs 2,390 crore.

Two figures on two different bases would normally be the point at which a careful reader stops. Not here, and the reason is worth understanding rather than memorising. Depreciation is subtracted and is never negative, so earnings after depreciation are always below earnings before it. So the comparable, after depreciation return on the crossing must be below 13.8 per cent. 13.8 per cent is therefore a ceiling, and a ceiling that already sits below the 14.2 per cent hurdle settles the comparison without the missing depreciation figure ever being found.

The caution belongs in the note, so say it out loud as well. Capital employed at the packaging business and project cost at the crossing are not identically constructed quantities, so this is a comparison of two returns on capital committed rather than a like-for-like ratio built from one definition. The comparison can support the direction. The comparison cannot support a precise gap.

The swap that makes the distinction worth the trouble

Harivansh Packaging Limited holds cash of Rs 140 crore. Imagine that amount going out as a share repurchaseA company buying back its own shares from the market and cancelling them, so the same earnings are divided among fewer shares afterwards. How it is decided and what it signals are covered later in this sequence. instead, funded by postponing the boiler overhaul and the rest of the year's replacement work.

The repurchase side of that swap can be computed. At the illustrative share price of Rs 300/- and earnings per share of Rs 12.50/-, the earnings yieldYearly earnings per share divided by what was paid for the share, written as a percentage. The yield says how much current earning a rupee spent on the share actually buys. is 4.17 per cent. Rs 140 crore at 4.17 per cent retires about Rs 5.83 crore of annual earnings for the holders who remain.

The postponement side cannot be computed. Postponing removes some amount of capacity from a business whose existing capital earns 14.2 per cent, at some point in the future, and this record does not carry a single figure that would let anyone size either the amount or the timing.

One side of that exchange has an arithmetic answer of about Rs 5.83 crore and the other side has no answer at all in this record, and the published accounts would show only that capital spending fell. That asymmetry, one computable side and one uncomputable side, is the reason a reader has to hold the distinction in mind rather than waiting for the accounts to hold it for them. Whether that exchange is worth making is not something the figures settle; what sits on each side of it is all they show.

What happens when maintenance spending is postponed?

Postponement deserves its own section because it is the most attractive bad decision available to a management team, and attractiveness is a structural property here rather than a moral failing.

A household version. Picture a house run on one salary, needing a repaint and a new water tank in the same season. Money is tight that year, so the people in it postpone both. Nothing bad happens this year. Nothing bad happens next year either. In the fourth year the tank fails in the middle of a summer, the ceiling below it is damaged, and the cost of fixing everything is several times what the replacement would have been. Ask them in year one whether they made a mistake and they will point at a bank balance that says otherwise.

Now the structure of it, stated plainly. The cash released by postponing maintenance is visible and immediate, and the capacity it costs is invisible and delayed. A reporting cycle rewards exactly that shape of decision. Every quarter and every year that separates the release from the cost is a period in which the decision looks good, and the decision keeps looking good for as long as the postponement lasts.

Two further things follow from that shape, and both are worth carrying.

The first is that postponement is self-concealing. When capacity finally goes, it arrives as an operational event: a line down, an order not filled, a run rateThe level a figure implies if the current period is simply carried forward unchanged. A useful shorthand and a poor forecast. The shorthand assumes nothing about the period was unusual. that quietly resets lower. The reset does not arrive labelled as the consequence of a decision taken four years earlier, and the person who took that decision may well have moved on.

The second is that postponement compounds. The shed not repaired this year has deteriorated further, and prices have moved, so it costs more to repair next year. Postponing growth spending has no equivalent property: the new line costs what it costs whenever it is built, and waiting a year does not make the untaken opportunity worse in the same mechanical way.

What does confusing the two cost a reader?

The confusion runs in both directions, and both directions are expensive in a way that survives for years.

Read all capital spending as growth, and a business that is standing still looks like a business that is investing. Every rupee of replacement gets counted as an addition. The capital spending line looks healthy and forward-leaning. The capacity it is buying is exactly the capacity that existed before, and nobody notices. The accounts do not distinguish, and the story is a good one.

Read all capital spending as maintenance, and the opposite happens. A genuine expansion disappears into the run rate. A business that has just committed serious money to a new line looks like a business grinding along, and the earnings that arrive two years later arrive without any visible cause. Earnings with no visible cause look like luck instead of a return on something.

Whichever way the reader has assumed, the accounts present one figure and go on presenting it, so both errors survive for years. There is no moment at which the accounts contradict the assumption. A wrong reading of a ratio gets corrected the next time somebody recomputes the ratio. A wrong reading of a capital spending line gets corrected when the capacity turns up or fails to. The correction may come four years later, and it will not be presented as a correction.