Capital Work-in-Progress (CWIP): A Balance-Sheet Red Flag

There's a line on the asset side of the balance sheet that does something none of the others do: it can grow for years while producing nothing. Capital work-in-progress — CWIP — is the money a company has spent building assets that aren't finished yet. A half-built factory, a plant being commissioned, a dam mid-construction. In a healthy business CWIP is a way-station: cash flows in, sits there briefly, then graduates into property, plant and equipment (PP&E) where it starts earning. The problem is when it doesn't graduate. When CWIP sits on the books year after year, swelling, never converting — that's not a way-station. That's a parking lot for capital that may never come back.

I'll be honest about the trap I fell into the first time I tried to use this signal: I treated a big CWIP number as the flag. It isn't. A big number can be a company mid-build doing exactly what it said it would. The flag is a big number that stays big — capital that goes in and doesn't come out — and it took me a few embarrassing false alarms to learn to read the movement instead of the level. This piece is about reading the movement, and it's built around one real balance sheet where the movement tells the whole story.

What is capital work-in-progress?

CWIP is the accumulated cost of fixed assets that are under construction or installation but not yet ready for use. Think of it as a holding account between "we spent the money" and "the asset is producing."

Here's the lifecycle in a healthy company: management decides to build a plant → cash is spent on land, civil work, machinery, installation → all of it accumulates in CWIP while the plant is being built → once the plant is commissioned and ready to operate, the whole accumulated cost moves out of CWIP into PP&E (gross block) → from that point the asset starts depreciating and, crucially, starts generating revenue.

The accounting fact that makes CWIP worth scrutinising: assets in CWIP are not depreciated. Depreciation only begins when the asset is "ready for intended use" and capitalised into the gross block. So while a project sits in CWIP it imposes no charge on the income statement at all. The company has spent the cash — it left the building when capex was paid — but the cost isn't yet flowing through profits. That gap between "cash spent" and "cost recognised" is exactly where a problem can hide undetected for years — the same disconnect that makes quality of earnings worth checking whenever reported profit and actual cash drift apart.

Why is CWIP a place value goes to hide?

The non-obvious bit is the asymmetry. Almost every other warning sign on the balance sheet eventually shows up in earnings — bad debt becomes a write-off, falling margins become lower profit, rising interest becomes a bigger expense. CWIP is different. A project that's stalled, over budget, or quietly abandoned can sit in CWIP indefinitely without ever touching the profit-and-loss statement, because you don't depreciate what you haven't commissioned.

That creates a perverse incentive. If a project is going badly — costs ballooning, demand evaporated, the plant no longer viable — capitalising it into PP&E would start the depreciation clock and force the cost into earnings, dragging down reported profit. Leaving it in CWIP avoids that. The asset stays "under construction" on paper, the cash stays buried, and the income statement stays clean. The reckoning only comes if and when the company finally writes it down as an impairment — often years later, often as a "one-time" charge that turns out to have been building quietly the whole time.

So perennial CWIP — a large balance that persists or grows across multiple years without converting to PP&E — is worth treating as a question, not a footnote. The cash is real and already gone. The only thing in doubt is whether it ever earns a return.

One balance sheet where you can watch it happen

Take a company whose CWIP did exactly the thing the textbook warns about — and which discloses, in its own audited notes, that it knows. The state-owned hydropower producer NHPC built capital work-in-progress from roughly ₹15,000 crore in FY14 to nearly ₹40,000 crore on a standalone basis (₹39,834 cr) by FY25 — and to ₹50,398 crore on a consolidated basis (NHPC FY25 annual report; Inve data, FY14–FY25). Over the same stretch, its net fixed assets barely moved — sitting around ₹21,000–22,000 crore for most of FY16 through FY24 (Inve data, balance sheet). Money kept going in. Assets kept not coming out.

(This is an illustration of how to read a balance sheet line, not a view on the stock — NHPC is a regulated utility where long gestation is structural, and nothing here is a recommendation.)

The single most revealing disclosure is the one Schedule III now forces companies to print: the CWIP ageing schedule. Here is NHPC's, consolidated, as on 31 March 2025 — the balance split by how long each rupee has sat unconverted (NHPC FY25 annual report, consolidated, Note 2.2.1):

Read the bottom row, not the total. Nearly half the pile — ₹24,109 crore — had been sitting in construction for more than three years (NHPC FY25 annual report, consolidated). That is not a holding account doing its job. That is capital parked. And the next note in the same report removes any doubt about which projects: NHPC prints a "CWIP completion schedule for delayed projects," naming the Subansiri Lower Project (₹19,468 crore still to complete), Parbati-II (₹12,685 crore), Teesta-VI, Pakaldul and Kiru (NHPC FY25 annual report, consolidated, Note 2.2.1(b)). When a company has to publish a schedule headed delayed projects, the balance sheet has stopped whispering.

How do you spot perennial CWIP? The forensic checks

You can't see "stalled" from a single year's number. You see it from the movement across years. Three checks do the work, and NHPC happens to fail all three — which is what makes it a clean teaching case.

Check 1 — CWIP versus net block, tracked over time. A company in active, healthy expansion shows CWIP rise and then fall as projects commission and the cost transfers to PP&E. The danger sign is CWIP that climbs while net block flatlines. NHPC's standalone CWIP roughly doubled across the decade while net fixed assets sat near ₹21,000–22,000 crore through most of it (Inve data, FY16–FY24). The capital was being raised and spent; the earning base wasn't growing.

Check 2 — The CWIP-to-PP&E transfer ("capitalised" column). The fixed-asset note shows additions to CWIP and the amount transferred out to the gross block. In a healthy capex cycle you see large transfers out as projects finish. In NHPC's standalone FY25 movement, against an opening CWIP of ₹27,452 crore and additions of ₹5,682 crore, only ₹150 crore was capitalised into the gross block (NHPC FY25 annual report, standalone, Note 2.2). Five-and-a-half thousand crore in; a hundred-and-fifty out. Money was going in by the truckload and trickling out by the bucket.

Check 3 — CWIP age, where disclosed. Schedule III of the Companies Act, 2013 (Division II), effective for financial years beginning on or after 1 April 2021, requires the ageing schedule shown above — and, for anything overdue against its own cost or timeline, a completion schedule. Read these first. A large balance in the "more than 3 years" bucket is a literal admission, signed off by the auditor, that capital has been parked rather than progressing.

What management said while the meter ran

CWIP comes with a narrative, always. Every long-running project has a concall explanation, and the pattern to watch for is the same explanation with the commissioning date quietly sliding forward each time. NHPC's flagship Subansiri Lower Project — 2,000 MW on the Assam–Arunachal border, under construction since the mid-2000s — is the said-vs-did record in miniature, drawn straight from the calls.

On the Q2 FY24 call (November 2023), the then-CMD said: "I am very hopeful that we can commission two units of the project in fourth quarter of current financial year and remaining six units are expected to be commissioned one-by-one by May '25. The estimated cost of the project is Rs. 21,248 Crore out of which, we have already incurred Rs. 19,459 Crore till September'23" (NHPC Q2 FY24 concall).

Eighteen months later, on the Q4 FY25 call (May 2025), the target had moved: "we will commission three units by June'25 and rest of five units will be commissioned one-by-one by May'26" (NHPC Q4 FY25 concall). Inve's record flags the FY26 commissioning commitment, made on the prior call, as diluted — guidance that was walked back rather than met as stated.

By the Q3 FY26 call (February 2026), two units were finally declared commercial — and the cost had moved too: "We have already spent around Rs. 25,691 Crore against the revised cost of Rs. 27,949 Crore" (NHPC Q3 FY26 concall). The estimate had climbed from ₹21,248 crore to ₹27,949 crore — a roughly ₹6,700-crore overrun — and the levelised tariff the project must now earn had risen from about ₹5.6 a unit to ₹7.50 (NHPC Q2 FY24 and Q3 FY26 concalls). Asked about it, NHPC's finance team didn't reach for spin. Saroj Kumar Roy said the quiet part: "Parbati-II and Subansiri both have taken exceptional time in commissioning. 20-25 years kind of timeline has been there" (NHPC Q3 FY26 concall).

That is the value of matching the balance sheet to the call. The ageing schedule tells you ₹24,109 crore has been parked for over three years; the concall record tells you the commissioning date slid from Q4 FY24 to FY26, the cost rose by a third, and management says so plainly. Inve's Promise Tracker pins each commissioning and capex commitment to the quarter it was made and records what happened — delivered, diluted, missed, or never mentioned again — with the original quote attached. For an investor holding several capex-heavy industrials, matching each company's CWIP trend against its project guidance, quarter after quarter, is precisely the legwork that doesn't get done by hand. Which is how a parked balance stays unexamined.

When the parking lot finally clears its books

The reckoning, when it comes, comes through impairment — and NHPC's notes show that too, on the smaller, genuinely dead projects. The expenditure on the Bursar Project, ₹128.74 crore, was "fully provided for" after the company concluded no further reimbursement was coming (NHPC FY25 annual report, consolidated, Note 34(24)). The Kotli Bhel projects in Uttarakhand have been frozen since a 2013 Supreme Court order halting environmental clearances, and roughly ₹380 crore of CWIP against them has been provided for (NHPC FY25 annual report, Notes 2.2.3 and 34(25)). In all, the FY25 standalone note carries ₹554.98 crore of survey-and-investigation-stage CWIP "provided for" — capital the company has now admitted, in writing, may never earn (NHPC FY25 annual report, standalone, Note 2.2.3).

Here is the honest steelman, because the case isn't one-sided. NHPC is a regulated utility, not a momentum manufacturer chasing a fad. Under the cost-plus tariff regime, an approved cost overrun on a project delayed for reasons "beyond the control of the management" can be folded into the capital cost the regulator allows it to recover — which is exactly the argument management makes on the call. Subansiri's delays were real and externally caused: a National Green Tribunal halt to construction ran from 2011 to 2019, and the geology of the Himalayan foothills is nobody's fault. A dam is not a textile mill; twenty-year build-outs are the nature of the asset, and "long CWIP" in hydropower is not automatically "value destroyed." If you can't underwrite the regulatory recovery and the eventual tariff, you shouldn't be in the name at all — and that's a fair reason to leave perennial-CWIP utilities to people who can.

Where the steelman runs out is the part the tariff regime can't fix: time. A rupee spent in 2013 and earning its first revenue in 2026 has surrendered thirteen years of compounding, and no cost-plus formula returns that. The cash was real and gone the whole time the income statement looked clean.

How does CWIP connect to returns and cash flow?

The link is to return on capital. Capital sitting in CWIP is capital employed — on the balance sheet, often funded partly by debt — but it earns nothing. So a large, stagnant CWIP balance silently depresses return on capital employed (ROCE): the denominator keeps growing while the numerator doesn't, because the assets aren't producing. A company can report a respectable ROCE on its operating assets while a fat CWIP balance drags the true, all-in return well below the headline. Read CWIP alongside the return ratios and the picture sharpens — the kind of linked reading that separates real analysis from definition-collecting, and that connects to the leverage story in ROCE versus ROE.

There's a debt dimension, and NHPC shows it. Its consolidated borrowings climbed from about ₹20,800 crore in FY14 to ₹39,557 crore in FY25 — roughly in step with the CWIP build (Inve data, balance sheet). Projects stuck in CWIP are often debt-funded, which means the company pays interest on assets that aren't yet earning. NHPC's own note shows ₹2,166 crore of borrowing cost capitalised into CWIP in FY25 alone — interest folded back into the unfinished asset rather than charged to profit (NHPC FY25 annual report, consolidated, Note 2.2.2). That keeps reported earnings clean today and makes the eventual all-in cost larger. A long-delayed, debt-funded project is one of the cleaner routes into a debt trap — servicing borrowings against an asset that hasn't started paying for itself.

A short routine for reading CWIP

1. Pull the CWIP line for three to five years from the balance sheet, alongside net PP&E.
2. Check whether CWIP rises and then falls as projects commission, or just keeps rising while net block flatlines. Falling-after-rising is healthy; permanently elevated is the flag.
3. Read the CWIP ageing schedule in the notes. Anything in the "more than 3 years" bucket deserves an explanation — and check for a "delayed projects" completion schedule beside it.
4. Compare additions to the "capitalised" column in the fixed-asset note. Crores going in but almost nothing transferring out means projects aren't finishing.
5. Watch for capitalised borrowing cost. Interest folded into CWIP keeps today's profit clean and tomorrow's project more expensive.
6. Adjust your view of returns. Mentally back out idle CWIP when judging ROCE; a flattering operating return can mask a poor all-in return.
7. Match it to the concall. Has the commissioning date held, slipped repeatedly, or vanished? Check the project-completion commitments in the Promise Tracker.

Frequently asked questions

For a five-year owner, CWIP is a question about discipline and disclosure: does this management finish what it starts, does it tell you the truth when a project slips, and — if the asset genuinely takes a decade to build — can you actually underwrite the eventual return? A balance that rises and then converts into earning assets is capex working as intended. A balance that ages in place, funded by debt, with the commissioning date sliding call after call, is capital parked where the income statement can't see it. NHPC at least prints the ageing schedule and says "20-25 years" out loud; plenty of companies leave you to find the parked capital yourself. The line is easy to skim. Reading it — bottom row first — is where the edge is.

Related reading

• Asset Turnover Ratio: Formula, DuPont & ROE Link
• Cash Conversion Cycle Explained: DSO, DIO, DPO
• Contingent Liabilities in Balance Sheet Explained