The Reserve Bank of India’s special dollar-rupee swap window is routinely described as the central bank absorbing a hedging cost that banks would otherwise have paid. That description is accurate about the effect and misleading about the shape of what the RBI took on.

The facility opened on 8 June 2026 for fresh FCNR(B) deposits of three-to-five-year tenor and was scheduled to run to 30 September. It closed a month early, on 31 August, after drawing $127.23 billion. Adding external commercial borrowings and overseas foreign currency borrowings, the window took in $136.38 billion - roughly four times what the two comparable 2013 windows raised between them, and it did so in weeks rather than months. What follows concerns one component of that outcome: the currency exposure the RBI now carries, and why it is a different instrument from the one it wrote in 2013.

At Par Is Not the Same as Subsidised

The mechanics are unusual and worth stating precisely. A bank sells dollars to the RBI at the FBIL reference rate, and at maturity buys the same dollars back at the same rate. Both legs are fixed at inception. Nothing in the arrangement is priced off a forward premium.

The effect on banks is exactly as reported: a fully hedged dollar-rupee position at zero explicit hedging cost, worth roughly 280 to 300 basis points a year, which is what allowed FCNR(B) rates to reach the 6% to 7% range against a market hedging cost of about 3.5% before the window opened.

But the RBI has not paid that cost. It has taken the position the hedge existed to lay off. The distinction matters because the two have different shapes. A market counterparty selling a hedge charges a premium at inception and is finished: its cost is known on day one and capped. A counterparty transacting at par carries the exposure for three to five years, and its cost is whatever the rupee actually does over that period, with no ceiling.

2013 Capped the Subsidy. 2026 Removed It.

The 2013 window is the natural comparison, and the difference is not one of degree. In September 2013 the RBI swapped FCNR(B) dollars into rupees at a fixed swap cost of 3.5% a year, compounded semi-annually over the tenor of the swap. That rate sat below the market forward premium, so the facility carried a subsidy. But the subsidy was the gap between two numbers known on the day, and the RBI’s whole cost schedule was settled when the swap was struck.

In 2026 the premium is not discounted. It is absent. That is the difference between a subsidised fixed price and no fixed price at all: the first is a known transfer whose size can be written down on the day, the second is an open position whose size is unknown until maturity. The two 2013 windows raised $34 billion between them on capped terms. The 2026 facility raised close to four times that on uncapped ones.

The Facility Sits on Its Own Break-Even

Paying the market premium and transacting at par cost the same at exactly one point: where realised annual rupee depreciation equals the premium waived. Below that point the RBI ends up better off than the rate it declined to charge. Above it, worse, and without a ceiling.

Cost to the RBI over the term on the $127.23 billion FCNR(B) notional, against realised annual rupee depreciation. The RBI has not published a tenor breakdown, so both ends of the eligible three-to-five-year band are shown rather than a single weighted assumption. The flat line is a premium fixed at inception; the rising lines are an open position.

The crossing point is worth locating, because USD/INR has roughly doubled over the past two decades, which compounds to about 3.5% a year. The facility is priced, in effect, at the historical mean, with no margin built in either direction.

The near term is running above it. The rupee stood at about 94.45 to the dollar on 7 September 2026, roughly 7% weaker than a year earlier - and that is after the $136 billion of inflows and continued RBI intervention had pushed it back to two-month highs. One year is not five, and these swaps mature between 2029 and 2031. But the position was opened at a level where the recent trend is running at twice its break-even.

Why Published Cost Estimates Diverge So Widely

Multiplying a premium by a term by a notional produces a number that moves enormously on assumptions that are easy to get wrong. The same $127.23 billion at 2% over three years gives about $7.8 billion. At 3.5% over five years, about $23.9 billion. The tenor assumption swings the answer further than the rate assumption does.

The RBI has not published a tenor breakdown of the deposits it took in. Any single headline figure for “the cost” is therefore a choice of assumption presented as a measurement. The range is the finding; the point estimate is not.

Reserves Bought With Debt Are Still Owed Back

A second distinction is easily lost in the headline number: what kind of capital this is. A deposit is borrowing. The $127 billion goes back out in dollars at maturity, with interest, and sits in the external debt column while it is here. Prasanna Tantri of the Indian School of Business projects external debt rising from roughly $765 billion toward $900 billion on the back of it, and flags clustered maturities, the displacement of ordinary remittances by borrowed dollars, and ₹7.7 lakh crore of surplus reserve money as the risks that follow.

Equity capital carries no repayment date and no coupon. The provider’s return depends on how the underlying businesses actually do, and the currency risk stays with the provider rather than moving to the central bank. The two are not substitutes, and they are not equally easy to attract. Deposits can be raised quickly by paying enough for them, which is precisely what a 280 to 300 basis point hedging saving passed through to depositors amounts to. Equity responds only to slower work - predictable rules, deeper markets, businesses worth owning - and cannot be summoned to a deadline.

What This Means for Allocators

None of this settles whether the facility was worth doing. The RBI received $127 billion of reserves for it, and reserves have their own value in defending a currency; the swap exposure is one side of a ledger, not the whole of it. Three implications follow regardless of where one lands on that question.

Read reserve adequacy with maturity dates attached. A reserve stock built through par swaps is not equivalent, for stress-testing purposes, to one built through purchased flow.

Note that the central bank’s reaction function has changed. Having guaranteed the future rupee value of $127 billion, the RBI now holds a direct balance-sheet interest in the rupee not depreciating past roughly 3.5% a year through 2031. Anyone pricing INR risk over that horizon is pricing it against a counterparty with a new and quantifiable reason to defend the level.

And treat the exposure as open rather than incurred. The reserves are real. The position is open. The balance comes due between 2029 and 2031, and not before.

Back to Writing