Where it goes

Chapter 3 found the money. This is the half of the accounts where the arithmetic stops being comfortable.

The expenditure, in full

April to September 2026, in ₹ crore:

	Budget Estimate	Actual to Sept	% of BE	Last year
Revenue expenditure	41,25,494	18,95,288	45.9%	(43.7%)
of which: interest payments	14,03,972	6,46,010	46.0%	(45.3%)
Capital expenditure	12,21,821	6,54,547	53.6%	(51.8%)
of which: loans disbursed	2,78,780	1,51,220	54.2%	(52.6%)
Total expenditure	53,47,315	25,49,835	47.7%	(45.5%)

Revenue expenditure is more than three times capital expenditure. That is the normal shape of a government, which exists mostly to pay people to do things rather than to build things, and it is not in itself a criticism.

The two ratios nobody quotes

Take interest payments of ₹6,46,010 crore and hold them against two other numbers from these same accounts.

Against the fiscal deficit of ₹7,14,413 crore:

6,46,010 ÷ 7,14,413 = 90.4%

Interest on past borrowing is 90.4% of this year's new borrowing. The government borrowed ₹7.14 lakh crore in six months and paid ₹6.46 lakh crore of interest on what it had borrowed previously. To a close approximation, the deficit is the interest bill.

This is the same observation chapter 2 made from the other end: the primary deficit — the fiscal deficit excluding interest — was only ₹68,403 crore, 9.6% of the fiscal deficit. More than nine-tenths of the borrowing requirement is the legacy of earlier borrowing rather than anything decided this year.

Against net tax revenue of ₹12,76,088 crore:

6,46,010 ÷ 12,76,088 = 50.6%

Half of every rupee of net tax revenue goes to interest. Fifty paise in the rupee, before a single school, road, soldier or subsidy.

Neither ratio appears in any coverage of these accounts and both are one division from the printed table. They are, between them, the strongest available argument for caring about a deficit — not because borrowing is wrong, but because the interest on it is a prior claim on all future revenue, and it compounds. Chapter 6 picks this up as a debt dynamic rather than a yearly cost.

Capital expenditure, which is the better news

Capital expenditure at 53.6% of its Budget Estimate is the only major line ahead of a pro-rata 50%, and it is ahead of revenue expenditure at 45.9% by 7.7 points.

Front-loading capital spending is deliberate and sensible. Construction has seasons — the monsoon stops work — and a project started in April does more in a year than one sanctioned in January. A government that spends its capital allowance early is doing the thing it is hardest to do, because capital spending is the easiest line to postpone when cash is short and the easiest to cut when a deficit target is at risk.

One qualification, from the table itself: ₹1,51,220 crore of that capital expenditure is loans disbursed rather than assets built. Lending to a state or an entity is a capital transaction and it is not a road. Strip it out and direct capital spending is ₹5,03,327 crore — still 70.5% of the fiscal deficit, which is the sense in which the borrowing is overwhelmingly funding capital formation.

The subsidies, and what their pace reveals

The accounts publish the four major subsidies separately. April to September 2026, in ₹ crore:

	Budget Estimate	Actual to Sept	Last year	% of BE	Last year %
Food	2,27,629	1,02,506	88,375	45%	(43%)
Nutrient-based fertiliser	54,000	33,664	29,301	62%	(60%)
Urea	1,16,799	85,586	77,974	73%	(66%)
Petroleum	12,085	1,149	6,717	10%	(56%)
Total major subsidies	4,10,513	2,22,905		54%

Total subsidies of ₹2,22,905 crore are 31.2% of the fiscal deficit — roughly a third of the borrowing, if you choose to attribute it that way, which is a choice rather than a fact since money is fungible.

Two lines are worth reading carefully, and they are 63 percentage points apart.

Urea at 73% of its annual allowance by September. Fertiliser subsidy is paid against sales, and the kharif season concentrates in the first half of the year, so some front-loading is structural. At 73% against 66% last year, though, three-quarters of a year's allowance is committed with half the year to run, which usually ends either in a supplementary allocation or in payments pushed into the next year. Both happen; the accounts cannot tell you which is coming.

Petroleum at 10%, against 56% last year. The petroleum subsidy is largely cooking gas, and what it costs depends on the gap between the market price and the administered one. A collapse from 56% to 10% of the annual allowance is most straightforwardly explained by that gap narrowing — crude or LPG prices moving favourably — which costs the government less without any policy changing.

The honest marking the exercise asks for: the urea reading rests on seasonality the accounts do not describe, and the petroleum reading rests on a price I have not sourced here. The first is an inference, the second is a hypothesis, and the accounts support neither on their own. They support only the two percentages, which is still more than most readers have.

What the expenditure side says

The composition is better than the headline and the inheritance is worse.

Capital spending is ahead of pace and ahead of last year. Total expenditure at 47.7% is behind pro-rata, which is fiscally disciplined. And beneath both of those, interest consumes half of net tax revenue and accounts for nine-tenths of the borrowing requirement — a position no single year's decisions created and no single year's decisions can undo.

The point

The Union government's spending is about three-quarters revenue and one-quarter capital, and capital expenditure at 53.6% of its estimate is the only major line running ahead of pace. Interest payments of ₹6,46,010 crore are 90.4% of the fiscal deficit and 50.6% of net tax revenue — half of every rupee of tax, before anything is done with it — and neither ratio is ever quoted. Major subsidies of ₹2,22,905 crore are 31.2% of the borrowing, with urea at 73% of its annual allowance by September and petroleum at 10%.