India, in Infographics: Macro and Markets

A Few Billion A Year To Carry. $47 Billion A Quarter To Burn. The Case For The Window.

Critics call the $127 billion FCNR(B) swap window expensive borrowed money. It is borrowed — but the fair test is not its cost against zero, it is its cost against the thing it replaced: selling India’s own reserves into a war-driven oil shock to hold the rupee. That defence had already cost $46.88 billion in a single quarter before the window opened.

The window’s carry, a yearNet of the yield on the parked dollarsReserves burned defending the rupee, one quarterBoth on one dollar scale

The window’s net carry

$2.8bn/yr

≈6.5% coupon on $127bn, less the ≈4.3% earned parking those dollars in reserves. Gross ≈$8bn; no principal before 2029

Burned in one quarter

$46.88bn

What defending the rupee in the open market cost, Feb–Jun 2026 — owned reserves, permanent, sold into a rising dollar

Reserves, kept at a record

$729bn

The window rebuilt the buffer without RBI firing its own dollars into the market; a reserve defence spends it down instead

Added per ₹1 the rupee slides

≈$7bn

The burn isn’t the whole bill — a rupee left to fall lifts India’s ≈$700bn import cost, crude above all

The comparison, in one line

One quarter of defending the rupee by selling reserves cost $46.88 billion — already more than three years of the window’s entire gross coupon ($25 billion), and about 17 times a single year’s net carry. The reserve dollars are gone for good; the window’s dollars are a loan repaid from 2029, with the $127 billion buffer left standing in the meantime.

The honest debit

The window is not free. The Reserve Bank now carries the whole exchange-rate risk on $127 billion — its net forward-dollar book was already past $103 billion at the end of June — so if the rupee keeps sliding, that carry rises and the state, not the banks, is short the dollars. The record reserve total is borrowed, not earned. And the burn figure is a measured quarter, not a certainty for the year: had the storm passed quickly, the old defence might have cost far less. The claim that FCNR is borrowed money is simply true. The point is that it is costly next to what — and its alternative was to spend the buffer itself.

Download The Graphic Post It See The Window’s Scale → How This Was Built ↓

Free to use anywhere with credit to politypolicy.com

The two bills, side by side

Pay the FCNR carry

≈$2.8bn a year

Bounded. A known carry on a fixed $127bn, gross ≈$8bn, net ≈$2.8bn once the yield on the parked dollars is netted off.
Deferred. Not a rupee of principal falls due before 2029, over a 3–5 year tenor.
Buffer intact. The $127bn sits in reserves the whole time, steadying the rupee while it is there.
Reversible. A loan can be rolled, repaid or let mature; it is a liability with a date, not capital spent.

Burn the reserves

$46.88bn in a quarter

Permanent. Dollars sold in the open market do not come back; the buffer is smaller for good.
Open-ended. At the Feb–Jun pace, a full year of the same defence runs to ≈$188bn — more than the entire window.
Sold cheap. The selling happens into a rising dollar and a falling rupee — the worst possible price in the cycle.
Not the whole cost. A rupee left to slide still lifts the import bill and imports inflation; the reserve loss buys you less than it looks.

What the window actually costs

The coupon, and the offset most critics leave out

FCNR(B) deposits pay a rich dollar coupon — around 6.5% — so the gross interest on $127 billion is about $8 billion a year. But those same dollars do not sit idle: they land in reserves, where the Reserve Bank earns a yield of roughly 4.3% on US Treasuries and other official paper. Net the one against the other and the true interest carry is about 2.2% — $2.8 billion a year — and even that overstates it, because the banks earn a further spread lending the swapped rupees at home. The honest cost is single-digit billions a year, not the headline coupon.

The real exposure is the currency risk, not the interest

In 2013 the banks carried a slice of the exchange-rate risk. In 2026 the Reserve Bank has taken the whole of it onto its own book — its net forward-dollar position was already past $103 billion at the end of June. If the rupee keeps falling, the rupee cost of returning $127 billion of deposits rises, and that bill lands on the state. That risk — not the coupon — is the genuine price of the window, and it is a risk, a range of outcomes, not a fixed charge.

Why letting reserves burn is the worse bill

The dollars do not come back. A swap is a loan with the buffer left whole; an open-market sale is capital gone. Reserves fell $46.88 billion in the one quarter before the window — permanent, at a moment India could least spare it.
The selling is at the worst price. A reserve defence means selling dollars into a rising dollar and a war-driven oil spike — buying rupees dear, selling dollars cheap, exactly when the market is moving against you.
The rupee slides anyway, and inflation follows. Reserves rarely hold a currency alone against a shock. A weaker rupee lifts the whole ≈$700 billion import bill — crude above all, at an elevated ≈$150 billion a year — adding roughly $7 billion for every rupee of depreciation, and forcing rates higher.
A thinner buffer is its own penalty. Every dollar burned cuts import cover and India’s credit standing, nudging up what the government and companies pay to borrow abroad — a cost that outlasts the crisis.

How this was built

The carry is an interest differential, deliberately conservative

Net carry is the $2.8 billion that remains after subtracting a 4.3% yield on the parked dollars from the 6.5% deposit coupon on $127.22 billion — about 2.2%. The gross coupon is $8.3 billion a year. Both are printed so the sum can be redone; the true net cost is arguably lower, since banks also earn a domestic lending spread on the swapped rupees.

The burn is a measured quarter; the annual figure is a pace

Reserves fell $46.88 billion between the pre-war peak on 27 February 2026 and the window’s launch on 5 June — both RBI weekly prints, a firm number. The ≈$188 billion annual figure simply runs that quarter’s pace over four quarters; it is a pace, shown to gauge scale, not a forecast that the defence would have continued unchanged.

The import sensitivity is illustrative

India’s merchandise import bill runs on the order of $700 billion a year, so each ₹1 the rupee depreciates adds roughly $7 billion to its rupee cost — a back-of-envelope figure for scale, with crude the most exposed line. The FX risk the Reserve Bank now bears is a range of outcomes, not a fixed charge, and is kept out of the carry number.

Sources: Reserve Bank of India weekly statistical supplement and swap-facility data; contemporaneous reporting on the FCNR(B) deposit rate and RBI’s forward book; trade data for India’s import bill. The window figures are RBI provisional totals at 31 August 2026. The cost model, its assumptions and its arithmetic are the author’s, laid out above so the reader can check them. Provisional.