The cost of the $136 billion FCNR(B) swap
What it costs India to maintain the RBI's dollar-rupee swap — under a rupee that rises, holds, or falls — and how it feeds through to inflation.
An interactive explorer of the annual cost to India of the RBI's $136 billion FCNR(B) dollar-rupee swap, under rising, flat, and falling rupee scenarios, and its inflation implications.
In June 2026, with the rupee at a record low near ₹96.6 and reserves falling through a West Asia oil shock, the RBI reopened its FCNR(B) swap window — only the second time in thirteen years. It expected $50–80 billion. By 31 August it had pulled in $136.4 billion ($127.2bn of it FCNR(B) deposits), and its FX forward book hit a record $136.7 billion. The dollars sit in reserves now; they have to be handed back in 2029–31.
The scheme's headline feature is that the RBI absorbs the currency-hedging cost: banks swap their NRI dollars with the central bank and buy them back later at an RBI-set rate, so the rupee risk sits with the RBI, not the banks. That is what "zero cost" means in the coverage. It is worth being precise about what it does not mean. The cost did not disappear — it moved (bank → RBI), it changed form (a hedging fee became a bet on the rupee), and it got deferred to the redemption window. This explorer prices that bet.
The RBI earns US-Treasury-type yield on the parked dollars while it holds them; that offset is built into the "net of reserve earnings" figure below.
The bet — pick a rupee path
"Normal" is the rupee's long-run slide — roughly 4–5% a year against the dollar (last 15 yr 4.3%, 20 yr 3.5%, since 1991 ~3.7%). Stress and crisis sit beyond that.
Reading the all-in dollar figure. The all-in runs above the coupon because it folds in the currency line, converted at each year's rate. That currency portion is not dollars leaving the country — it is a mark-to-market on the RBI's forward book. It still counts, for three reasons: it crystallises on redemption if the rupee ends weaker and the RBI has spent the dollars rather than held them, forcing it to source them at a worse rate than it took them in; it shrinks the RBI's surplus transfer to the government, so it reaches the exchequer as a smaller dividend rather than a wire; and it measures how much dearer the dollar principal grew in rupee terms. So read the coupon as hard cash and the gap above it as the price of the rupee bet — contingent and deferred, not absent. Waving it off as a paper loss is the same move as calling the swap "zero cost."
The rupee path you picked, against the alternatives
Your scenario (solid), with the firms / normal / crisis references in grey. Tap to enlarge.
Best case to worst case — the annual cost as a function of the rupee
Average all-in cost per year over the tenure, across every rupee path from −5% to +15%. The dot is your current pick. Tap to enlarge.
Every scenario at once — the bill over the tenure
Cumulative all-in cost under each preset, fanning out from the day the swap was signed. Your current pick is bold. Tap to enlarge.
Who is actually short the dollar?
The structure is a leveraged dollar carry trade — and it is built so that nobody running the leverage carries any currency risk. Follow the chain to where the risk finally lands.
Everyone who can be hedged, is — by design. The one balance sheet left short the dollar is the public's. That is the trade: private capital runs risk-free leverage, and the RBI holds the currency short. "Levered short against the dollar" is right about the risk — but the leverage sits with the punters and the short sits with the central bank.
How much that short actually costs turns on what the RBI does with the dollars. Park them in reserves, and the forward loss is offset by those reserves revaluing up as the rupee falls — the net cost shrinks toward the ~280–300 bps hedging subsidy plus any negative carry. Spend them defending the rupee — and inflows can go to intervention rather than headline reserves — and the short goes naked: the RBI has to buy the dollars back in 2029–31 at whatever the rupee is then. The scenario fan above prices that second case. Principal × the rupee's move is the upper bound, not the central estimate.
How it feeds inflation — and why the direction flips
A dollar swap pulls in two opposite directions at once, and the balance reverses between the day the money comes in and the day it leaves.
Money coming in
Money going out
Two readings, equal weight
It worked the defence
A fast, textbook circuit-breaker. It rebuilt reserves that had fallen ~5%, arrested the rupee's slide from ₹96.6 back toward ₹95, and bought disinflation during a live oil shock.
The economics are benign in the central case. NRIs bear no rupee risk, so the deposits are sticky and roll easily. The RBI earns yield on the parked dollars, which offsets much of the coupon. Used exactly as the tool is designed to be used.
The bill is deferred, not cancelled the critique
It is off-balance-sheet borrowing that steadies the symptom — a falling rupee driven by portfolio outflows and an oil-heavy import bill — without touching the cause. The inflows blew past the RBI's own $80bn comfort level to $136bn; the incentives (rate-ceiling lifted, CRR/SLR waived) worked too well.
The central bank is now record-short $136.7bn forward. A sharp rupee fall crystallises large mark-to-market losses that shrink the surplus the RBI transfers to the government — a quiet quasi-fiscal cost — and lands a redemption wall five times 2016's. "Zero cost" is the phrase doing the concealing: the public now owns the currency bet.
Assumptions, method and sources
What is sourced. The $136.37bn total (FCNR(B) $127.22bn, OFCB $5.26bn, ECB $3.89bn) as of 31 Aug 2026; the record $136.7bn FX forward book (July 2026); the 3–5 year tenure, CRR/SLR exemption, and lifted rate ceiling (SOFR+300bps for 3–5yr) — all from RBI Circular RBI/2026-27/99 (8 June 2026) and RBI data releases, as reported by Business Standard and Business Today. Spot near ₹95 and the ₹96.6 record low (19 May 2026) from market data. The 2013 scheme raised roughly $26bn.
What is assumed (adjust them above): spot ₹95, USD deposit coupon 6.5% (banks offered ~6–7.1% in the 3–5yr bucket), reserve yield on parked dollars 4.3%, tenure 4 years (mid-point of 3–5), principal $136.4bn. The annual rupee move is the variable you set.
The cost model. For each year t the rupee is projected as spot₀ × (1 + dep)^t. Dollar interest is valued at that year's rupee: coupon × principal × spotₜ. Principal revaluation is the year-on-year change in the rupee value of the dollars: principal × (spotₜ − spotₜ₋₁). All-in for the year is interest + revaluation; cumulative sums across the tenure; the average is cumulative ÷ tenure. Amounts in ₹ lakh crore (₹1 lakh crore = ₹1,000 bn).
What the model does not claim. The "cost" is spread across three balance sheets — depositors (coupon), banks (which earn a rupee lending spread and are hedged), and the RBI (which bears the currency risk and earns reserve yield). This is a stylised, all-in treasurer's view, not a single-entity P&L. The revaluation line is a mark-to-market that only crystallises on redemption at that year's rate; if deposits roll, it resets rather than settling. It is also an upper bound: it assumes the RBI has deployed the dollars, so it must buy them back at the prevailing rate. To the extent the dollars sit in reserves instead, the forward-book loss is offset by those reserves revaluing up, and the real cost collapses toward the hedging subsidy (~280–300 bps a year, absorbed by the RBI) plus any negative carry. The inflation direction depends on the RBI's sterilisation stance, which is not modelled here. Interest is treated as accruing at year-end spot; a within-year average would shift the carry line slightly. These are flagged, not hidden.
Figures are illustrative and provisional; RBI swap-window data is itself marked provisional and subject to revision. Nothing here is investment advice.