How India's government capital expenditure flows from Budget announcements to contractor order books, revenue and cash, and why an order is not revenue for infrastructure and cement companies.
In the 2023 Budget, the government raised its capital expenditure target to about ten lakh crore rupees, a record, and kept the number above eleven lakh crore in the Budgets that followed. Capital expenditure means building things: highways, railway lines, ports, power transmission, defence equipment.
This is the state acting as the economy's biggest customer. When the largest buyer in the country announces it will spend more, the businesses that build, and the businesses that supply the builders, all stand to gain. The market's mistake is not the direction. It is the speed. A Budget number is an intention. It has to become a tender, then a contract, then construction, then a payment, and every one of those steps runs on its own slow clock and loses something along the way.
Demand rose, or more precisely, a promise of demand was announced. The government declared it would buy more construction than last year. Everything else is the long journey from that sentence to cash in a company's bank account.
L&T and its peers win the tenders, so the promise reaches them first, as order books. But an order is a promise of future work, not revenue. The market re-rates the stock on order inflow; the profits arrive years later, if execution holds.
They sell to the builders, one link down. Their volumes move only when sites actually pour concrete, years after the speech. But the advantage of this link is real: the cement maker does not have to win the tender. It only needs the project to actually happen. The contractor carries tender risk, the fear of losing the bid; the cement producer carries only execution risk, the fear that the project stalls. Whoever builds the road, they all buy cement and steel.
Contractors need working capital to survive the gap between building and being paid. Banks that finance that gap between work performed and cash received grow their loan books alongside the capex. The end customer is the government, which is good for eventual recovery, but government projects still carry real payment delays and receivables stress, so the lending is not risk-free.
A Budget is a fixed purse. A record capex line is paid for somewhere: slower growth in subsidies or revenue spending elsewhere. The businesses that lived off the squeezed lines quietly feel it.
Investors who bought infrastructure and railway stocks after two years of capex Budgets, at rich valuations, owned the right theme at the wrong price. Several of those stocks corrected sharply in late 2024 and 2025 even as the actual spending continued.
The single most useful habit this signal teaches: when a company announces a huge government order, do not read it as revenue. Read it as a queue.
The order will convert to revenue over five or seven years, at whatever pace land, clearances and the contractor's own capacity allow. Some of it will be delayed. Some will be renegotiated. The cash will trail the work by quarters, because the government certifies before it pays. A company can hold a record order book and still have a miserable year in cash terms.
So the questions that matter are conversion questions. How fast does this company historically turn orders into revenue? What share of past orders slipped? How much working capital does each rupee of government revenue trap on the balance sheet, and for how long? Two contractors with identical order books can be a good business and a terrible one, separated only by those three answers.
Here is the leak most investors never check. The Budget number that moves markets in February is a target. The number that moves businesses is what actually gets spent by the following March, and the two can part ways.
In the election year of 2024, capital spending ran well behind the announced pace for months, because governments slow decision-making around elections and new project approvals wait. The revised figures later confirmed the year's actual capex came in below the grand target the market had cheered. The stocks had priced the speech. The businesses received the spending, which was smaller and later.
The same gap appears at the state level, where a real share of announced projects stalls on land and clearances. None of this makes the capex theme false. The roads did get built and the spending did grow, year over year, into 2025 and 2026. But every rupee of it arrived later and leakier than the announcement suggested, and a stock priced on the announcement has no room for later and leakier. Follow the monthly accounts the government publishes, not the February speech. The speech is marketing. The monthly accounts are the till.
You do not have to take that on faith. The gap between the announced number and the committed one is published every year. Let us line them up.
If the Budget headline were translating cleanly into real spending, the money the government actually commits should track the number it announced. If the framework is right, it will fall short, and by more in an election year, when approvals slow.
The method again: we do not have to prove the capex story false. We take the single number the market reacted to, the Budget target, and set it against the government's own revised estimate, the honest recount it publishes near year end. If the promise is leaking, the recount will be the smaller number.
Both years, the money the government committed came in below the number it announced: about half a lakh crore short in FY24, and roughly 0.9 lakh crore short in the election year FY25. The market reacted to the bigger number in February; businesses received the smaller one, later, over the following year.
The same gap shows up one link down, inside the biggest contractor. In FY25 L&T won a record 3.57 lakh crore of fresh orders and carried a record order book of about 5.8 lakh crore, yet booked only 2.56 lakh crore of actual revenue (Business Standard). The order book is a queue worth more than two years of work, not money in the bank, and orders won that year outran work actually done. That is an order book behaving exactly like a promise, not a profit.
Reading it fairly: the revised estimate is the government's own later count, not an outside guess, so this is not us calling the capex fake. Spending did grow every year, and the roads got built. The point is narrower, and it is the whole point. The money arrives smaller and later than the February headline, and a stock priced on the headline has no room for smaller and later.
You can watch the promise shrink. Eleven lakh crore announced, ten committed, and at the company that wins the work, two and a half lakh crore actually done in a year. An order is not revenue, and now you have seen the gap rather than been told about it.
| Event | Immediate | ~6 months | ~2 years |
|---|---|---|---|
| A record capex Budget is announced | Infra, railway and defence stocks re-price that afternoon. Not one tender has been issued. | Tenders flow and order books swell. Revenue has barely moved, and impatient money starts leaving. | Construction is underway, contractor revenue and cement volumes show the push, and the market has moved on to worrying about the next Budget. |
Use the pipeline on a different promise. The government announces a production-linked incentive scheme paying electronics manufacturers a bonus on every phone made in India.
Where this signal plays out in depth: the sectors it moves and the companies that lived it.
Government spending is real demand on a government clock. A Budget announcement must become tenders, then contracts, then construction, then payment, and it leaks and slips at every joint. The market prices the speech in an afternoon; the businesses receive the money over years. The durable trades in a capex cycle are the conversion questions nobody asks on Budget day: how fast, how leaky, how much cash trapped on the way.
The speech is marketing, the monthly accounts are the till. An order is a promise, revenue is work done, and cash is a certified bill finally paid. Price the till, not the speech.