India has raised $57 billion in ten weeks. The rupee is exactly where it started.
The FCNR window will be remembered as a success on reserves and a non-event on the currency. Both are already true, ten weeks in, and the second was largely knowable in advance.
Since 8 June, $56.85 billion has come through the RBI's three concessional windows — $52.3 billion of it FCNR(B).¹ Set that against FCNR(B) inflows of $946 million for the whole of FY26,² and you have a subsidy that manufactured a flow rather than amplified one. Nothing about diaspora appetite changed in June. The price did.
On 14 August the RBI closed the window five weeks early.³ The rupee closed that day at 95.42, up three paise — broadly where it sat when the scheme opened.⁴
The instrument is nearly identical to 2013. Everything around it is different, and the differences determine whether it works.
The cleanest test of what this subsidy is worth is to set it against a bank's alternative source of dollars.
State Bank of India priced $500 million of five-year Reg S paper on 13 August at 5.25%, or 88 basis points over the five-year Treasury.⁵ Adjust for the five-year USD bond-swap spread and that is roughly SOFR+116. The RBI's swap takes those dollars to rupees at SOFR flat. So SBI's all-in rupee funding via FCNR sits around 5.25% — identical to its dollar coupon, and comfortably inside the five-year sovereign.
That comparison excludes issuance fees, which on a deal of this type run between 0.5% and 1.5% of notional — ten to thirty basis points a year amortised over the tenor — and which push the true cost of the bond route above its headline coupon.
Look at where SBI has set its FCNR deposit rate and the logic is visible in the pricing: 6.00% above $1 million, 5.75% below.⁶ Marginally cheaper than its rupee deposit costs, which makes building the book worthwhile. Fifty to seventy-five basis points above its own dollar bond, which makes it worth an NRI's time. Both legs clear.
That is the trade working as designed. It works because SBI funds at T+88.
Step down the curve and it stops working. ICICI printed $1 billion of five-year paper on 23 July at 5.459%, T+100 — its first dollar bond since 2017.⁷ Its FCNR rate is 6.25% above $4 million, 6.00% below.⁸ That is 79 basis points dearer than its own bond, and against a five-year G-Sec it is level with sovereign funding rather than inside it.
The pattern holds across the sector. PNB at 6.60%. HDFC at 6.25% after a 25bp bump on 1 August. Axis, Kotak, Bank of Baroda at 6.00%. AU Small Finance at 7.10%, up from 5.15%.⁹
For most of the system this is not cheap money. It is a deposit-franchise play priced at or above what the sovereign pays and well above the offshore bond market. Banks will take it because the RBI has removed the hedging cost and the deposit base has strategic value — not because the arbitrage is compelling. Expect that to be said out loud before the window closes, particularly by banks still carrying FX risk on the interest leg. The swap covers principal only.
Which raises the question of who captured the three-plus points the RBI gave up.
Pre-announcement FCNR rates sat between 2% and 4%.¹⁰ They now run 5.75% to 7.10%. AU Small Finance moving 195 basis points is the cleanest single-name read.
Not all of that is the swap. CRR/SLR exemption, the withdrawal of the rate ceiling on 17 June and the removal of NRE rate parity all landed at once,¹¹ and the hedging component cannot be isolated from public data. But the direction is not in doubt. A subsidy designed to attract dollars has been competed away to depositors — which is what happens when you remove a cost constraint in a market where banks are chasing the same finite pool of savings.
This is the sharpest break from 2013. Then, the depositor took roughly 3% and the bank kept the subsidy, funding three-year rupees near 7.5% against a sovereign paying north of 8.5% — a full point inside the government curve, fully hedged. In 2026 the depositor takes 6% to 7% and the bank's advantage over sovereign funding has all but vanished.
Same subsidy, different pocket — and on any reasonable view, the better pocket. A concession designed to attract foreign savings reaching the foreign saver, rather than being retained by the intermediary, is what efficient transmission looks like. Competition between banks did that, not policy design. It simply means the banks are now here for franchise and diversification rather than for carry, and their enthusiasm should be read accordingly.
The levered structure survives largely intact — the RBI's June FAQ explicitly permits banks including their overseas branches to lend against these deposits and issue standby letters of credit.¹² Gearing is broadly comparable to 2013, and so are the headline returns: borrow against a cash-collateralised deposit, run it nine or ten times, clear low double digits on your own capital.
What has changed is what low double digits is worth. In 2013 that return sat against a zero front end and a five-year Treasury near 1.5%. It was extraordinary, and it did not need to be explained to anyone. In 2026 the risk-free five-year is 4.37% and hurdle rates across most asset classes sit mid-to-high single digits. The same twelve or thirteen per cent is now merely respectable — and it comes with a three-to-five year lock-in, a one-year hard lock, and collateral pledged against a loan you cannot easily unwind. The trade has not become worse. The world around it has become more competitive.
Providing these swaps at zero cost means absorbing the forward premium. On mids, that is roughly ₹9.3 on the three-year and ₹16.8 on the five-year — call it 3.25% to 3.50% of notional per annum. The announcement itself compressed three-year points by around 50 basis points annualised, but with the dollar bid across Asia there is little scope for that mark-to-market to improve.
No official cost estimate has been published. That is the departure from 2013 that bothers me most.
In 2013 the RBI's own analysis produced a number — ₹10,000 to ₹20,000 crore — set against an estimated ₹1.6 lakh crore in annual import savings from a firmer rupee.¹³ Whatever you thought of the assumptions, the trade-off was on the table and could be argued about. In 2026 there is no figure at all. The only estimates in circulation are external and range too widely to be useful.¹⁴
The accounting compounds this. Dollars taken in under the swap add to headline reserves on receipt. The obligation to return them sits in the net forward position, disclosed separately, currently $103.3 billion short against $707.00 billion of gross reserves.¹⁵ Both numbers are published. Only one gets quoted.
Three things have moved, and all three cut the same way.
The differential has compressed to roughly a third. In 2013, five-year Treasuries around 1.5% against Indian sovereigns above 8.5% gave you a gap near 700 basis points. SBI's August print implies a five-year Treasury at 4.37%; against a five-year G-Sec in the mid-6s, you are at roughly 215. The carry that made the 2013 trade obvious is largely gone.
Leverage costs money now. Zero front-end dollar rates were doing enormous work in 2013. They are not available.
Policy divergence points the wrong way. US inflation is the more pressing problem today. Indian CPI ran 4.38% in June with July tracking near 4.50%, and the RBI has held at 5.25%.¹⁶ A more hawkish Fed is a higher-probability path than a more hawkish RBI. If the differential narrows further from the US end, the structural case for rupee appreciation weakens.
Put those together and the FCNR window reads less like a strategy for currency recovery than an instrument for slowing depreciation. That is a defensible objective. It is not the one the 2013 comparison implies, and it should be named for what it is.
Reserves were $707.00 billion on 7 August against $681.61 billion in the week to 5 June — roughly $25 billion over the life of the scheme.¹⁵ Against $52.3 billion of inflows, between a third and a half has not shown up in the stock. A current account under tariff pressure, equity outflows near $13.7 billion, continued spot intervention, valuation effects and deposits not yet swapped will account for it in some combination.
The harder question is how much of the $52.3 billion is new external money. Eligibility under the circular turns on instrument and tenor, not on provenance. Deposits can be funded from savings held offshore, from balances already sitting with Indian banks' overseas branches, from maturing deposits rolled forward, or from borrowed dollars whose origin is not disclosed. Which of those is genuinely additive to India's external position is not something the reporting is designed to reveal.
The definitive test — the change in FCNR(B) stock against reported gross mobilisation — requires RBI's monthly NRI deposit series, which runs about two months behind. July's numbers arrive around late September, after the window has closed. A scheme whose central claim is that it brought in foreign capital cannot be evaluated on that claim until after it has finished running.
The RBI operates three concessional windows and prices them differently. FCNR(B) gets the swap free. ECB and OFCB pay a fixed 1.5% per annum, compounded semi-annually.¹⁷ Against a market hedging cost near 3%, that is roughly double the subsidy per dollar on FCNR.
On 14 August the RBI closed the expensive window five weeks early and left the cheaper ones running to year-end.
That does not read as a judgement about dollar sufficiency. It reads as cost management. There was a liquidity dimension too — bank deposits rose ₹11 trillion across three fortnights to 31 July, to a record ₹269.4 trillion¹⁸ — and flooding the system with rupees faster than it absorbs them is its own problem. But the binding constraint was never the dollars. It was the cost of warehousing the currency risk against them.
The June package was built to support the currency without raising domestic rates. That constraint selected the instrument, and it is worth asking why the constraint was there.
With CPI at 4.38% and the repo at 5.25%, the real policy rate is under a point — thin for a deficit economy under external pressure. Lifting local rupee rates would have raised real yields in sympathy with US real yields and supported the currency directly rather than through an inflow subsidy, at no cost to the central bank's balance sheet.
The problem is reversibility. A rate hike is a monetary policy action with economy-wide reach and a slow exit. If external pressure abates — the dollar turns, tariffs are renegotiated, the Fed pivots — you are left holding higher domestic rates, a higher cost of government borrowing, and a growth drag that takes several meetings to unwind. A subsidised deposit window is self-liquidating by design: it opens, it closes, and the balance sheet cost is borne by one institution rather than every borrower in the economy. Raising dollar deposit rates directly, and paying for it out of the RBI's book, targets the external problem without committing domestic policy to a stance that may be wrong in six months.
That trade-off is the more plausible explanation for the choice than any view about which instrument works better on the currency.
Sequencing is a separate question. In 2013 the window followed rate action. Launching this subsidy after a hike would have compressed the basis between offshore INR cross-currency swaps and local rates — plausibly to zero or through it — letting banks raise the same rupees via FCNR dollars materially inside sovereign funding rather than level with it. Objectively better economics for banks and a smaller bill for the RBI, but an economy-wide impact of higher local rates and a still higher cost of government borrowing.
The window has delivered on reserves and given the RBI the option of rolling its short forward book out three to five years. Kicking the can is a legitimate choice when the alternative is disorderly.
But India's external liabilities have shifted toward subsidised medium-term debt maturing 2029 through 2031, on terms whose cost has not been quantified publicly, in a rate environment far less generous than the one that made 2013 work. The currency has not responded. And the test of how much of this was genuinely new money arrives only after the scheme has closed.
The criticism worth making is not that the RBI was wrong to act. It is that a programme of this size should arrive with a published cost estimate — as the 2013 version did — so the trade-off can be debated in public rather than inferred from forward points and closure dates. On the evidence of 14 August, what it bought was time.
Figures as of 16 August 2026. Analysis, not investment advice.
