Gravitywell
Research
SCENARIO DOSSIER · WHAT-IF
GWW-2026-003
AUGUST 2026
CONDITIONAL SCENARIO · NOT A FORECAST
India Macro · The Rupee

What if the rupee breaks 100?

The rupee reaching 100 is the forward curve, not the scenario. What this dossier prices is the roughly one-in-six path where it arrives in a hurry, because the rupee is the one variable that can damage a dollar investor's India book without a single Indian asset falling in price. Hedge the path, not the number.

P(break) ≈ 12–22%over 12m · vs strict base rate ~8–9% · P(touch) ~70–85%
99.18
the 12-month forward outright — the market already prints it ◐
~80%
the curve's own odds of touching 100 inside a year; the house shades to 70–85% ○
~$522bn
deployable reserves on the desk's adjustment, against a $716.9bn headline ○
44%
of India's $180bn external commercial borrowing carries no hedge ◐
Gravitywell Research · GWW-2026-003Contents

ContentsWhat is in this dossier

PageSectionThe claim it proves
03The ShockThe question in the title is the wrong question; the forward market already prints 99.18
04Executive summaryA break is roughly one-in-six over twelve months, and its damage lands where nothing is marked
05Scenario Quick ReadThe cascade, the hedges and the posture, in sixty seconds
06–10Part I · The SetupTouching 100 carries no information; the defence behind the line is smaller than the headline
11–15Part II · The TransmissionEverything unpriced sits in the second and third order
16–20Part III · The Impact MapA break redistributes inside India before it destroys, and it lands hardest on dollar-funded capital
21–23Part IV · The DynamicsThe dampers are stronger than in any prior episode, which is why the break is unlikely rather than impossible
24–26Part V · The PlaybookNobody should hedge the level; the forward already sells it to you
27Pre-mortem & red-teamThe strongest objection against this dossier survives it
28Probability & assumptionsFive assumptions carry the call; the gold haircut moves it most
29Bottom lineThe rupee will probably touch 100 and probably not break
30–35Methodology · Sources · Glossary · DisclosuresHow it was built, what it rests on, and what it is not
About this series

Gravitywell What-If dossiers study a defined shock the way a risk desk does: conditionally, probability and magnitude held apart, every effect traced to a mechanism and a source, and the strongest case against the argument printed alongside it. A What-If is not a forecast. It sets out what would follow if a stated trigger occurred, weighted by how likely that trigger is, so a reader who allocates capital, underwrites risk or sets policy can decide what to hold now and what to do if it fires.

Conventions

Currency. Dollar figures are US dollars; rupee figures carry ₹, converted at the RBI reference rate on the date cited. Quoting. USD/INR rises when the rupee weakens, so "96.46 to 100" is a fall of about 3.7%. Exhibit frames. An exhibit ruled with a solid line reports observed history or arithmetic on published actuals. One ruled dashed and marked IF is scenario-conditional: model output, true only given activation, never a measurement. Tables carry the same mark. Tiering. Every figure carries its evidence grade: ● primary disclosure, ◐ primary reported through a secondary source or a reconciled press figure, ○ a Gravitywell Research construction or estimate. Sources are numbered and listed on page 31; the superscript beside a figure points to its row. Dates. Full month names in prose; three-letter abbreviations in exhibits and tables where space requires. Point-in-time. Data is as of 25 August 2026 and is not restated silently; a later correction publishes as a dated addendum. Probabilities are stated as bands with a horizon, and are estimates, not measurements.

Gravitywell Research02 / 35
Gravitywell Research · GWW-2026-003The Shock

The ShockThe activation condition

The instrument, before the argument · 25 Aug 2026
USD/INR spot · record low96.4575 · 96.9650 (late May 2026)10
12-month forward premium2.82% annualised9
12-month forward outright99.18
P(rupee touches 100 within 12m)70–85%32
P(the break defined below)12–22%

Read the third and fourth rows together and the question on the cover has answered itself. The forward market will sell you dollars for delivery next August at 99.18, and on that curve the odds of spot touching 100 inside twelve months sit around four in five. A reader who came here to find out whether the rupee gets to 100 can stop now. It very probably does, and few people who price rupees for a living would argue otherwise.

The question worth thirty pages is about manner, not level. A currency can crawl to a round number, smoothed by a central bank that sells a few billion on the bad days, which is what India has done for a decade. Or the crawl can fail and the rate can cover the same ground in weeks. Both paths end at the same number and they are not the same event.

Activation condition — all three, inside a rolling 12-month window

1 · The level holds. The RBI reference rate closes at or above 100.00 and stays there 20 or more consecutive sessions. 2 · The crawl breaks. The episode contains a fall of 6% or more inside a rolling three-month window, against a managed norm near 4% a year. 3 · The defence is spent. Desk-adjusted reserves fall $50bn or more from the episode's start, or the RBI states publicly that it is smoothing volatility only.

Each leg is checkable from a published source: the reference-rate series, the Weekly Statistical Supplement, the Bulletin's forward-book table and the FBIL premia. This is a conditional scenario analysis and not a forecast.

FIG 1.1 · The touch is near-certain, the break is notIFP, unconditional, 12m
severe ~3.5% strict base rate ~8–9% P(break) 12–22% P(touch 100) 70–85% forward outright 99.18 · reflection estimate the faint segment = the uplift this desk adds to the base rate · both bands are unconditional over 12 months 0%50%100%
Sources: forward premium 2.82% at 10 Jul 2026 ◐9; touch band a desk reflection estimate at 4–7% assumed volatility ○32; base rate from three speed-clearing episodes in 35 years ◐25.
Gravitywell Research03 / 35
Gravitywell Research · GWW-2026-003Executive summary

Executive summaryProbably 100. Probably not a break.

In one line · a roughly one-in-six break carries damage that no Indian asset price would show you

A break is about a one-in-six event over the next twelve months, roughly double the strict base rate of 8–9%. The uplift is earned by two facts rather than by mood: the level is only 3.7% away, and the machinery the RBI uses to smooth the path is already heavily committed, with a net short forward position that peaked at $106.6bn in May.2 On the level itself this desk sits with consensus and says so.

What makes the tail worth pricing is where the damage lands. A break can take two to three points of annualised dollar return off an India book without any Indian asset falling in price, because for a global allocator the rupee is the currency the returns are eventually paid in. Funds already underwrite 18–20% in rupees to deliver 15% in dollars,21 and almost none of that gap is hedged at fund level. Three nodes carry the loss, and none is marked where a reader would look.

FIG E.1 · Likelihood and size, in separate panelsIFleft P · right magnitude
PROBABILITY · UNCONDITIONAL, 12 MONTHS MAGNITUDE · USD/INR PEAK, GIVEN ACTIVATION touches 100the breaksevere branch ~78% ~17% ~3.5% 0%50%100% Each bar is a separate unconditional event, not a share of one whole. MildModerateSevere w 35%w 45%w 20% 96100105110115 Weights (w) sum to 100% GIVEN activation; they are not probabilities of the scenario.
Sources: probability panel as FIG 1.1 ○32; weights derived on page 22 from six reference episodes ○.
TABLE E.1 · Three unpriced nodes carry the lossIF
The unpriced nodeMechanismSize
The defence's compositionReserves rise while the rupee sets records because the build is borrowed and revalued: ~$28bn of net new swap-window deposits in eight weeks4, gold near 15% of the stock8$716.9bn headline against ~$522bn deployable ○
The unhedged borrowing book44% of $180bn of external commercial borrowing carries no hedge, and February's amendment widened the pipe without a general hedging mandate1617~$79bn unhedged; a 6% break is a ~$4.7bn mark ○
The return currencyExits strike in rupees and remit in dollars, with negligible fund-level hedging21−0.15 to −0.23× of dollar multiple, 2–3 points of IRR ○
The read

Hedging the level is close to pointless, since the forward already sells it to you. The exposure worth covering is the path. Page 26 prices that honestly and reaches an unfashionable answer: the option tail is not cheap on this desk's own odds, and what is worth buying is the plain forward on money you will genuinely repatriate inside a year. If you run dollar capital against Indian earnings, your hedge roll date is the trigger to watch, not the spot screen.

Gravitywell Research04 / 35
Gravitywell Research · GWW-2026-003Scenario Quick Read

Quick ReadThe cascade in sixty seconds

In one line · the sequence runs through the hedging curve, and that is the part with a price on it

If the break fires, the order of events is fairly predictable even though the outcome is not. Oil widens the trade gap, the rate gives way faster than the RBI wants, and within days the market stops asking where the rupee is going and starts asking what the defence has left. The forward premium answers that question months before the reserve data does, so it sits at the top of the watch list and near the bottom of how most India books are actually positioned.

FIG 2.1 · The sequence, and where the path forksIFday 1 → year 2
Day 1–5 100 trades; RBIsells onshore + NDF Week 1–4 forward premium +80–150bp;corporates rush to hedge Month 2–6 · THE FORK does the near-tenor forward book roll? Mild · crawl resumes, 100–103 Moderate · 103–108, RBI hikes 50–75bp Severe · 108–115, rolls fail, flow measures debated Weights given activation: mild 35% · moderate 45% · severe 20%. Colour: pale = contained, dark = the shock path, crimson = severe.
Source: Gravitywell Research scenario model ○. Timing calibrated on the 2013 episode (four months from first stress to the 68.85 low)24 and 2022 (a defended year-long grind)25.
TABLE Q.1 · Posture by stakeholder, now and on activationIF
If you areWhat to do nowWhat changes if it activates
An allocator with dollar exposure to IndiaPrice the path, not the level: buy 6–12 month out-of-the-money rupee puts or pay the forward premium on the harvest-stage slice of the book while the premium is 2.82%Roll cost jumps first; the marks follow a quarter later. Stop adding unhedged rupee duration
A general partner mid-harvestModel the exit in dollars at 100 and at 106, and tell the LP base which vintages that moves before it moves themThe exit window narrows twice over, on multiple and on translation. Bring forward what is sellable
A corporate with offshore debtFind out today whether your maturity sits in the 44% of the book that is unhedged. Hedge the near tenor, not the far oneHedging becomes expensive exactly when it becomes necessary; the near tenor is where the squeeze lands
A policymakerPublish the adjusted reserve position before someone else constructs it. Credibility is cheaper to keep than to rebuildEvery remaining lever borrows, and each one prices the next one higher (page 23)
What would tell you it is not happening

A Hormuz settlement that returns Brent to the $60s closes the argument outright, because the trade gap is the engine of the whole sequence.11 Short of that, watch for the rate approaching 100 with three-month realised depreciation staying under 4% and the desk-adjusted reserve series flat. That combination is a crawl arriving on schedule, and on these numbers it is the likeliest single outcome of the next twelve months.

Gravitywell Research05 / 35
Part I

The Setup

Touching 100 carries almost no information, because the forward curve already sells it to you. What decides whether the next twelve months are an inconvenience or an event is whether the managed crawl that produced every rupee low of the past decade can absorb this one, and what is actually left behind the line.

Reading pages
06 – 10
The claim
The defence is smaller than the headline
Part I · The SetupGWW-2026-003

§ 01 · The curveWhat the forward market has already said

In one line · 100 is roughly where the market will sell you dollars, so it cannot also be the surprise

The forward premium is not an opinion. It is the rupee-dollar interest differential, enforced by arbitrage, and at 2.82% for twelve months9 it puts the outright at 99.18 against a spot of 96.46. That is the rate at which a hedge actually strikes, which for anyone deciding what to cover matters more than any house forecast. Treat it as a prediction and you have made a category error; treat it as the price of the thing you are considering buying and it is the most useful figure on the page.

FIG 3.1 · The 100 line is inside the cone by month threeIFUSD/INR · 80% band
100.00 spot 96.46 12m forward 99.18 band reaches 100 · ~month 2–3 1081041009692 now+3m+6m+9m+12m
Source: forward premium 2.82% at 10 Jul 2026 ◐9; the cone is a lognormal 80% band at 5.5% assumed volatility, drawn by Gravitywell Research ○32. Shaded area is where the rupee sits above 100.

From that curve two probabilities fall out, and keeping them apart is the discipline everything here depends on. The chance the rate finishes above 100 in a year is about 41–44%. The chance it touches 100 at any point along the way is roughly twice that, because a barrier is far easier to hit than to end above. This desk shades the touch band down to 70–85% rather than printing the raw 80–88%, for one reason worth stating plainly: rupee forward premia have historically run ahead of what the rupee actually did, so a risk-neutral estimate is biased toward depreciation by construction.

TABLE 1.1 · Narrative consensus, priced consensus, and the house
Whose viewWhat it saysNumber
Street forecastsANZ, Bank of America, Goldman Sachs and Danske see the rate in a range over the next two years; a move to 100 is framed as needing a major oil or dollar shock and is explicitly not a 2026 base case2696–99 / 24m ◐
The forward marketThe only consensus that costs money to hold, and it is more aggressive than the published forecasts999.18 / 12m ◐
This deskAt consensus on the level, and saying so. The departure is on the path: the odds of a break, and what the defence actually containstouch 70–85%; break 12–22% ○
Gravitywell Research06 / 35
Part I · The SetupGWW-2026-003

§ 02 · The outside viewEight episodes, three of them fast

In one line · disorderly rupee episodes are common; ones that break the crawl are not

Since the float, India has been through roughly eight episodes in which the rupee fell 10% or more inside twelve months while reserves were being spent. That is about one every four or five years, and it is the loose base rate for "the rupee had a bad year". The narrower question is how often the speed leg cleared as well, meaning a fall of 6% or more inside a single quarter. On that test the count drops to three, and the arithmetic that follows is the honest anchor for everything downstream.

FIG 4.1 · Narrowing the reference class from "a bad year" to "a break"1991–2026
years observeddisorderly episodescleared the speed leg 35 years since the 1991 float 8 episodes · λ 0.23/yr · ~20% in any 12 months 1991 · 1995–96 · 1998 · 2008 · 2011–12 · 2013 · 2018 · 2022 3 episodes · λ 0.086/yr · ~8–9% in any 12 months 1991 · 2008 · 2013 — the strict base rate this dossier is anchored on Bar length is proportional to the count in each tier; the derived annual rate assumes a Poisson arrival.
Sources: annual and episode magnitudes ◐25; the 2013 path ◐24. Episode selection and the speed test are Gravitywell Research definitions ○.
TABLE 2.1 · Eight disorderly episodes, three of them fast
EpisodePeak moveWhat drove itSpeed leg
1991~19% ◐Balance-of-payments crisis; reserves down to weeks of import cover; a two-step devaluation in Julycleared
1995–96~18% ◐Post-liberalisation adjustment as the managed rate was allowed to find a levelno
1998~13% ◐Asian crisis contagion plus post-nuclear-test sanctionsno
2008−24% ◐25Global funding squeeze and a fast foreign-institutional exit; the worst calendar year in a decadecleared
2011–12−14.4% ◐25Euro crisis, oil near $110, a current account past 4% of GDP. November 2011 alone took 7% off the rupee, a record for one monthno
2013−20% in 4 months ◐24Taper tantrum; the record low of 68.85 on 28 August; answered with a swap windowcleared
2018~−10% ◐Oil to $86, a tightening Fed, emerging-market contagion; the crawl held throughoutno
2022−10%+ ◐25Commodity shock and the fastest Fed cycle in forty years; about $110bn of reserves spent defending itno
Episode selection and the speed test (a fall of 6% or more inside a rolling quarter) are Gravitywell Research definitions. Magnitudes are calendar-year unless a window is stated.

Note what the "no" column contains. In 2018 and 2022 the rupee fell about as far as it did in 2013, but it took a year rather than a quarter, and the difference in outcome for anyone holding rupee assets was enormous. That is the distinction the activation condition is built to capture, and it is why the strict rate of 8–9% rather than the loose 20% is the number argued from here.

Gravitywell Research07 / 35
Part I · The SetupGWW-2026-003

§ 03 · Calibration2013 taught the tool; 2022 taught the grind

In one line · India is running 2013's playbook off 2022's balance sheet, and paying twice the coupon for it

Two episodes calibrate this one. In 2013 the rupee fell about 20% in four months to a record 68.85 on 28 August, and the RBI answered by opening a swap window that pulled in roughly $26bn of foreign-currency deposits at a 3.5% fixed rate, with about 80% of the resulting exposure covered in forwards against a reserve stock of $365bn.24 In 2022 the shock was arguably larger and the response was the opposite: the RBI spent about $110bn of reserves over a year and let the rate grind down 10% without ever losing control of the pace.25 The difference was speed, and speed is what the market charges for.

FIG 5.1 · The same tool, and roughly twice the priceFCNR(B) swap windows
SIZE RAISED · $BN DEPOSIT COUPON PAID · % ON 3–5Y USD 2013 window2026 window $26bn $28bn net new · $52.3bn on the wider count 03060 Two published counts of the 2026 window are irreconcilable; both are printed. 20132026 3.5% 6.0–7.1% 0612 In both windows the RBI, not the bank, carries the hedging cost.
Sources: 2013 window ◐24; 2026 window terms and deposit rates ◐3; system outstanding $60.55bn against $32.56bn on 5 June, and the wider $52.3bn count ◐4. Two different bases are reported for the 2026 total and neither has been reconciled publicly.
TABLE 3.1 · 2013, 2022 and 2026 read on the same rows
201320222026 to date
TriggerFed taper signal, a 4.8% current accountUkraine commodity shock, fastest Fed cycle in 40 yearsHormuz closure, a Fed dissenting toward a hike27
Path−20% in four months−10% over a year, tightly paced−4% and counting, pace controlled so far
Defence usedSwap window, $26bn at 3.5%Spot reserves, about $110bnForwards to a $106.6bn peak, then a swap window23
Reserves at the start$365bnA record stock, then drawn hard$728.5bn record in February1
What it cost later$26bn redeemed in late 2016 without market disruption33Reserves rebuilt over two yearsA 2029–31 maturity at 6–7.1%, still ahead

The bottom row is the one worth sitting with. India has walked this maturity wall before and walked it cleanly, and it is the best evidence against the alarming reading of the 2026 window. It is also the row where 2026 differs most, because the deposits cost roughly twice as much and mature into a period with nothing else competing for the market's attention.

Gravitywell Research08 / 35
Part I · The SetupGWW-2026-003

§ 04 · Why nowThe trigger is not the rupee

In one line · a strait that carries a fifth of its pre-war volume is doing the work, and the EIA does not expect it back before 2027

Crude and liquids moving through the Strait of Hormuz averaged 4.9 million barrels a day in the second quarter of 2026 against 21.6 million in the last quarter of 2025.11 Brent touched $105 in late July and settled near $88 in August, and the EIA does not expect Middle East output near pre-conflict levels before early 2027. Everything in the setup runs off that single number, because India buys more than 85% of its crude abroad and the arithmetic between the barrel and the rupee is mechanical: about $13–16bn on the annual import bill and 30–40 basis points on the current account for every $10 on the average price.13

FIG 6.1 · A supply shock at one end, a deficit at the otherseparate scales
HORMUZ CRUDE + LIQUIDS · MILLION B/D INDIA CURRENT ACCOUNT DEFICIT · % OF GDP 4Q 2025Q2 2026 21.6 4.9 012.525 A 77% fall in throughput. The EIA does not expect Middle East output near pre-conflict levels before early 2027. FY26 actualFY27 · $75–80FY27 · $82–87FY27 · $90–95 0.6% 1.5% 2.0% 2.2% 01.252.5 Crisil's fiscal-2027 scenarios by Brent assumption; its own working assumption is $90–95.
Sources: Hormuz throughput, EIA Short-Term Energy Outlook ●; Brent spot prints via market reporting ◐11; current-account scenarios, Crisil Ratings June 2026 ◐12, with HDFC Bank at 2.1% on $85 ◐. Panels use different units and different scales by design.
TABLE 4.1 · Four pressures, and which one has already moved
PressureWhere it standsWhy it matters here
The trade gapJuly's merchandise deficit was $31.98bn on record imports of $76.2bn, up 17.5% on the year29The flow the rupee has to finance every month
The FedHeld at 3.50–3.75% for a fifth meeting, with three dissents preferring a hike and roughly a one-in-three market probability attached to one27A hike is the dollar leg; India is positioned for cuts
Equity flowsForeign investors have sold about ₹2.4trn of Indian equities in 2026, already past the ₹1.66trn of all of 2025, though July and August turned net buyers22The capital account has been leaning on debt, not equity
The offsetNet invisibles reached $312bn in FY26 against $264bn, with remittances near a record $140bn and a current-account surplus in the March quarter28Genuinely large, and it caps how far rather than how fast
Gravitywell Research09 / 35
Part I · The SetupGWW-2026-003

§ 05 · The defenceWhat $716.9bn actually contains

In one line · reserves rose $24bn in a fortnight while the rupee sat at a record low, and the reason is on the liability side

India's reserves stood at $716.9bn in the week to 14 August, up $9.9bn on the week after $14.1bn the week before, and within reach of February's record $728.5bn.1 Taken at face value that is a central bank with more ammunition than it has ever had. Taken apart, it is a stock in which the fastest-growing component is borrowed, a sixth is metal, and roughly $90–107bn of forward sales are already committed against it. The gap between those two readings is arithmetic anyone can repeat, and page 30 sets out exactly how it was done.

FIG 7.1 · From the headline to what can be sold on a bad Tuesday$bn · axis from 450
450550650750 716.9 −53.8 −23 −90 −28 ~522 Headline14 Aug 2026 Gold at 50%880.5t, ~15% share SDR + IMFnot market ammunition Forward bookpeaked $106.6bn May Swap depositsnet new, 8 weeks Deployabledesk adjustment The gold haircut is a desk judgement, not a disclosure: gold is pledgeable and India pledged it in 1991, but slowly and at a price in credibility.
Sources: headline reserves ◐1; gold tonnage and share ◐8; forward book ◐2, taken at ~$90bn after the unwind reported since May; swap-window deposits ◐4. The decomposition and the haircut are Gravitywell Research constructions ○34, maintained weekly as india-adjusted-fx-ammunition.
TABLE 5.1 · Adequacy read on the headline and on the adjustmentIF
Adequacy, read two waysOn the headlineOn the adjustment
Cover of short-term external debt, residual maturity ~$327bn2.2×1.6×
Cover of total external debt of $762.8bn694%68%
Import cover610.8 months~7.9 months
Adjusted ratios use the deployable figure from FIG 7.1. Short-term external debt on a residual-maturity basis is carried from GWW-2026-002.

None of those adjusted ratios is alarming on its own. India at 1.6 times short-term cover is still better placed than most of its peers and far better placed than it was in 2013. The point is directional: the reserve cover for total external debt has fallen from 100.6% in 2021 to 90.3% today on the published measure,6 and the composition of the stock has been moving the same way underneath it. This is not a novel observation and the desk does not claim it as one, since the adjusted-reserve critique already circulates in commentary. What is missing is a series that maintains it week by week, which is what page 30 describes.

Gravitywell Research10 / 35
Part II

The Transmission

The oil bill and the exchange rate are in the tape before lunch, and repeating them adds nothing. Everything this dossier is for sits one and two steps further on: a defence whose composition nobody marks, a hedging curve that reprices for every borrower at once, and a central bank forced to choose.

Reading pages
11 – 15
The claim
Everything unpriced is second order or later
Part II · The TransmissionGWW-2026-003

§ 06 · The chainWhere the market stops looking

In one line · three of the ten nodes in this chain are priced and the other seven are not

A currency shock propagates faster than most because the rate is a single observable price that everything else is denominated against. That speed is also why the first order is worthless to analyse: an importer's cost, a headline inflation print and a bank-index session all move within hours of the rate, and any screen will tell you so. The nodes below that carry an unpriced tag are the ones that take a quarter or more to show up in anybody's marks, and by then the cost of fixing them has usually moved.

FIG 8.1 · The transmission, and what the tape has already absorbedIF1st → 3rd order
1ST ORDER · PRICED2ND ORDER3RD ORDER · UNPRICED THE SHOCK Hormuz at 4.9 mb/d Brent $88–105 Import bill, current account0.6% → 2.2% of GDP The rate reaches 100already the forward Defence spends forwardscomposition unmarked Hedging curve reprices+150–250bp on the 12m premium Unhedged borrowers marked~$79bn exposed, ~$4.7bn at 6% Dollar returns re-struck−5 to −9 points of MOIC oil billflow gapintervention credibilitytranslationexit maths
Source: Gravitywell Research scenario model ○. Arrows are transmission channels; magnitudes are conditional on activation and sourced on the pages that follow.
TABLE 6.1 · Ten nodes, and the three the tape already pricesIF
NodeOrderPriced?Where it is argued
Oil bill and the trade gap1stpricedPage 12
The level itself1stpricedPage 06
Headline CPI print1stpricedPage 15
Composition of the defence2ndunpricedPages 10 and 13
The hedging curve2ndhalf-pricedPage 13
Unhedged external borrowing2ndunpricedPage 14
The policy reaction2ndhalf-pricedPages 15 and 23
Dollar returns on Indian assets3rdunpricedPage 19
The exit window3rdunpricedPage 20
The 2029–31 deposit maturity3rdunpricedPage 13
Gravitywell Research11 / 35
Part II · The TransmissionGWW-2026-003

§ 07 · First orderThe part already in the tape

In one line · the financing gap is arithmetic, and at $115 crude with a flow reversal it doubles the current account

India's exposure to the barrel is one of the most reliable relationships in emerging-market macro. Every $10 on the average crude price adds roughly $13–16bn to the annual net oil import bill and 30–40 basis points to the current account,13 and at $110–115 for a full fiscal year the increment runs $56–64bn. That is not a forecast, it is division, and it is why the current account is expected to move from 0.6% of GDP in FY26 to about 2.2% in FY27 on Crisil's numbers.12 None of this is news to the market, and none of it is where the argument here lives.

FIG 9.1 · Two things must go wrong together; one already hasIF$bn to find, annual
INCREMENTAL ANNUAL FINANCING NEED vs FY26 BASELINE FPI +$10bnflows flatFPI −$15bnFPI −$30bn Brent $80 $4bn $14bn $29bn $44bn Brent $90 $18bn $28bn $43bn $58bn Brent $100 $32bn $42bn $57bn $72bn Brent $115 $53bn $63bn $78bn $93bn Oil leg = (Brent − $70 FY26 average) ÷ 10 × $14bn. Flow leg = the shortfall against a flat foreign-portfolio year. FY26 actual deficit was $25.2bn. <20 20–40 40–60 60–80 >80 $bn
Sources: oil sensitivity ◐13; FY26 current account of $25.2bn and the FY27 path ◐1228; 2026 foreign-portfolio equity selling of about ₹2.4trn ◐22. Cell arithmetic is a Gravitywell Research construction ○.

Read across the top row and the picture is benign: at $80 crude, even a $30bn flow reversal leaves a manageable $44bn to find against an economy running a record invisibles surplus. Read down the right-hand column and it stops being benign, because $93bn is roughly the entire FY26 current-account deficit three and a half times over. What makes the current position uncomfortable is that one of the two variables has already moved. Crude is not at $80 and is not expected back there soon, so the buffer now depends almost entirely on flows staying orderly.

What the flow picture actually looks like

Foreign investors sold about ₹2.4trn of Indian equities in the first eight months of 2026, more than the whole of 2025, then turned net buyers in July and August.22 Underneath that, debt has been doing the heavy lifting: a record ₹55,518cr came into Indian debt in June alone, more than covering ₹49,340cr of equity selling in the same month, with about ₹21,652cr of it through the fully accessible route that index funds use.23 The capital account is being held up by a channel that did not exist before the JPMorgan index inclusion in June 2024, which is a genuine improvement in the financing picture and a new dependency at the same time.

Gravitywell Research12 / 35
Part II · The TransmissionGWW-2026-003

§ 08 · Second orderThe curve prices the doubt first

In one line · the forward premium is where the market votes on the defence, and it moves months before the reserve data does

When a defended currency comes under real pressure, the reserve release is the last place the market finds out. The first place is the forward premium, because that is where anyone who wants protection has to transact and where the cost of providing it is set. At 2.82% for twelve months the curve currently says the defence is credible and hedging is cheap.9 The mechanism that turns that around is uncomfortably self-reinforcing: as the rate moves, corporates that ran unhedged decide to hedge, their demand lifts the premium, the higher premium raises the cost of the RBI's own forward book, and a more expensive book is a smaller book.

FIG 10.1 · What protection costs, and when it is refinancedIFpremium · calendar
12-MONTH FORWARD PREMIUM · % AND $M PER $1BN HEDGED WHEN THE DEFENCE COMES DUE todayweeks 1–4moderatesevere 2.82% · $28m 3.9% · $39m 5.0% · $50m 6.5% · $65m 0%5%10% now+6m202920302031 $29bn forward bookinside 3 months $28–52bn swap-window deposits mature,3–5 year tenors at 6–7.1% The near block has to be rolled in a market that may not want the other side; the far block was raised to solve the near one.
Sources: current premium ◐9; forward book of $106.6bn at 31 May with about $29bn inside three months ◐2; swap-window terms, size and coupons ◐34. Branch premium levels are Gravitywell Research estimates ○.

The two blocks on the right are the same problem seen twice. The RBI sold dollars forward rather than spot through the first half of 2026, which preserved the headline reserve number and pushed the cost into a roll date, and about $29bn of that book sits inside three months.2 To retire it without spending spot reserves the bank opened a deposit window in June and paid up for the money, which converts a three-month problem into a three-to-five-year one at 6–7.1% instead of the 3.5% it paid in 2013.324 Both moves are defensible and neither is free. What they have in common is that the cost is deferred and the deferral is not visible in the number everyone quotes.

The read

If you want one number to watch rather than twelve, watch the twelve-month forward premium. It is the market's live vote on whether the defence is credible, it reprices before any reserve release, and unlike the spot rate it is not being actively managed. A move through roughly 4.3%, which is 150 basis points above where it sits now, is this dossier's third activation leg arriving in advance of the other two.

Gravitywell Research13 / 35
Part II · The TransmissionGWW-2026-003

§ 09 · Second orderSeventy-nine billion dollars with nothing behind it

In one line · the borrowers least able to absorb a break are the ones least likely to have hedged it

Of India's $180bn of outstanding external commercial borrowing, 44% carried no hedge as of June.16 The regulation permits this by design. Only borrowings with an average maturity under five years must be fully hedged, infrastructure finance companies must cover 70% of those, and everything else is left to the borrower's judgement. February's amendment to the borrowing regulations then widened the pipe considerably, lifting the automatic-route ceiling to the higher of $1bn outstanding or 300% of net worth, from a flat $750m a year, without adding a general hedging requirement.17 More dollars, borrowed on easier terms, with the same optional protection.

FIG 11.1 · What is exposed, and what a break costs its holdersIFthe book · the burden
OUTSTANDING EXTERNAL COMMERCIAL BORROWING · $180BN EXTRA REPAYMENT BURDEN ON THE UNHEDGED $79BN · $BN hedged 56% unhedged 44% ~$101bn~$79bn And the split is not even across the book: large corporates 68% hedged smaller firms 38% hedged at 100.00at 106at 112 $2.9bn $7.8bn $12.7bn 0714 Measured against a 96.46 base: the extra rupees needed to service and repay the unhedged stock, expressed in dollars at the starting rate.
Sources: unhedged share of the $180bn stock, RBI June 2026 ◐16; hedging by firm size, firm-level evidence for FY23 ◐16; regulatory terms ●17. The burden ladder is arithmetic on the unhedged stock ○ and ignores natural hedges in export revenue.

The distribution matters more than the total. Large corporates hedge about 68% of their foreign-currency liabilities and smaller firms about 38%, which means the loss concentrates in exactly the part of the borrower base with the thinnest equity cushion and the least ability to refinance domestically. A $12.7bn aggregate burden at 112 is not a systemic number for a $4trn economy. Distributed the way this book is distributed, it is a mid-market credit event with a currency label on it.

Where this desk has been before

In August, GWW-2026-002 argued that the widened external-borrowing pipe would reprice rather than ration if global private credit seized, and put 0.75–0.85 on it through August 2027. The mechanism here is the same one arriving by a different road. The reprice does not need a credit event abroad, because a currency break delivers it domestically, through the hedging curve, to a book that has grown on the easier terms February granted.

Gravitywell Research14 / 35
Part II · The TransmissionGWW-2026-003

§ 10 · Second orderA price path with no room in it

In one line · the RBI already projects 5.9% for the December quarter, so a modest pass-through puts the print through the band

India's inflation pass-through from the currency is genuinely small, and the reason is structural rather than lucky. Food is about 46% of the consumer basket and India grows most of it, so a weaker rupee reaches the index mainly through gold, edible oils and pulses.15 The RBI's own estimate is that a 5% fall adds roughly 35 basis points to inflation over several months, with other work putting it lower still at 15–25. On any normal price path that is a rounding error. This is not a normal price path.

FIG 12.1 · The projection leaves 10bp of headroom; a break uses 40IFCPI, % y/y
6.0% upper tolerance band 4.0% target 4.384.705.905.50 6.30 with pass-through Jun 2026 actualQ2 FY27Q3 FY27Q4 FY27 34567 Crimson stubs add 40bp, the pass-through from a 6% three-month fall on the RBI's own 35bp-per-5% estimate.
Sources: RBI Monetary Policy Statement, August 2026, for the FY27 path and the 5.25% repo hold ◐14; June print ◐14; pass-through estimate ◐15. The overlay is a Gravitywell Research calculation ○.

The June print of 4.38% was already the highest in seventeen months, and the RBI's August projection has inflation peaking at 5.9% in the December quarter before easing to 5.5%.14 A 6% three-month fall in the rupee adds about 40 basis points on the bank's own elasticity, which puts the December quarter through the upper tolerance band. That is the point at which a currency question becomes a rate question, because the committee has been explicit that it is guided by headline inflation and has held at 5.25% with a neutral stance for four consecutive meetings while calling itself neither dovish nor hawkish.

The read

The uncomfortable part is the growth number sitting beside it. The RBI raised its FY27 growth forecast to 6.7% in the same statement, so a break would force tightening into an economy that is not obviously slowing, which is the least popular decision a central bank can take and therefore the one most likely to be delayed. Delay is what turns a moderate branch into a severe one, because every month of waiting is another month of the forward book being rolled at a worse price.

Gravitywell Research15 / 35
Part III

The Impact Map

A currency break does not destroy value evenly, it moves it. Dollar-revenue businesses gain what dollar-funded investors lose, the safest-looking asset on the page is the one foreign money holds through an index, and the largest single pool of damage sits in funds that will not mark it for another two quarters.

Reading pages
16 – 20
The claim
It redistributes before it destroys
Part III · The Impact MapGWW-2026-003

§ 11 · Cross-assetThe safest-looking line is the exposed one

In one line · Indian government bonds are the false refuge, because the money holding them arrived through an index and leaves the same way

Set out in one currency the map is unremarkable, and set out in two it is not. Indian equities in rupee terms take a moderate knock and recover; the same equities in dollars take roughly double, and that gap is who is hurt. The line worth pausing on is government debt. It looks defensive, it pays a real yield, and it has been the channel financing India's capital account all year, with a record ₹55,518cr arriving in June alone.23

FIG 13.1 · Total return, by horizon and by currencyIFmoderate branch, %
0–1 month1–6 months6–12 months Indian equities, rupee terms −4% −9% −3% Indian equities, dollar terms −8% −17% −12% Government bonds via FAR, dollar terms −6% −13% −9% Banks and lenders, rupee terms −6% −14% −6% IT services, rupee terms +5% +10% +7% Gold, rupee terms +6% +13% +11% Dollar cash, rupee terms +4% +10% +8% ≤−12 −12 to −6 −6 to 0 0 to +6 >+6% Every cell prints its value; colour is a second encoding, never the only one. Bands are total return, not price change.
Source: Gravitywell Research scenario model, moderate branch ○. Calibrated on the 2013 and 2022 episodes for equity and currency co-movement2425; foreign debt-flow dependence from June 2026 data ◐23. Conditional on activation, not a forecast.
The exposed, least obvious first
−13%FAR government bondsIndex-tracking money sells the bond and the currency in one instruction
−17%Indian equities held in dollarsTranslation doubles a moderate rupee drawdown
−14%Banks and non-bank lendersRate defence, then credit cost from unhedged borrowers
The beneficiaries
+10%IT and services exportersDollar revenue against a rupee cost base
+13%Gold in rupee termsMetal price plus translation; already ~15% of reserves
+10%Remittance-receiving householdsA record ~$140bn of inflows converts at a better rate
Gravitywell Research16 / 35
Part III · The Impact MapGWW-2026-003

§ 12 · SectorsSomebody is being paid for this

In one line · every point the rupee loses adds 30–40 basis points to top-tier IT operating margin, and the tape has been treating that as earnings

Between January and March the rupee fell 3–4% and lifted the operating margins of India's six largest IT firms by somewhere between 40 and 320 basis points year on year, purely because the same dollar revenue converted into more rupees.30 Roughly 60–70% of that revenue is billed in dollars or euros against a cost base paid in rupees, so the rule of thumb is close to mechanical: about 30–40 basis points of margin for each point of depreciation. Quarterly profit growth looks stronger than the underlying business, which is worth remembering when the same firms are held up as evidence that India absorbs currency weakness comfortably.

FIG 14.1 · Margin sensitivity to a 1% fall in the rupeeIFbasis points, midpoint
IT services and captive centres Pharma formulations, export Textiles, gems and jewellery Capital goods, imported content Electronics assembly Airlines +30 to +40 ◐ +20 to +30 ○ +15 to +25 ○ −10 to −18 ○ −20 to −30 ○ −25 to −40 ○ −40bp0+40bp Only the IT band is externally sourced; the other five are desk estimates from revenue and input-cost currency mix, and are marked ○.
Sources: IT margin sensitivity and the Q4 FY26 uplift ◐30. Remaining sectors are Gravitywell Research estimates ○ built from disclosed export-revenue and imported-input shares; they are indicative and should not be used for position sizing on their own.

The redistribution runs deeper than a sector rotation. February's trade agreement cut the United States tariff on Indian goods from about 50% to 18%, leaving India below Vietnam and Bangladesh and well below China, with textiles, gems, pharmaceuticals and engineering goods the main beneficiaries. A weaker rupee stacks on top of that advantage rather than offsetting it, so the export-facing part of the economy would enter a break holding two tailwinds at once. On the other side, airlines and electronics assembly carry dollar input costs against rupee pricing power that is close to nil, and they feel it within a quarter.

Gravitywell Research17 / 35
Part III · The Impact MapGWW-2026-003

§ 13 · StakeholdersWho owes dollars and earns rupees

In one line · the most exposed party on this map is not an Indian company at all

Currency exposure reduces to one question, which is what share of what you receive and what share of what you owe is denominated in dollars. Plot those two shares against each other and the diagonal is the line of natural hedging, where the currency does nothing at all. Everything above it gains from a break and everything below it pays for one, and the further from the line the larger the effect. The corner of that map furthest from safety belongs to a global investor in an Indian fund, whose receipts are entirely in rupees and whose obligations, reporting and benchmark are entirely in dollars.

FIG 15.1 · Dollar receipts against dollar obligationsIF%, indicative
naturally hedged gains from a break pays for a break Remittance-receiving households IT services and captive centres Pharma and specialty export The sovereign · external debt 20.8% of GDP Airlines, electronics assembly Unhedged mid-market borrower A dollar investor in an India fund 100500 dollar shareof receipts 050100 dollar share of obligations and input costs
Source: Gravitywell Research ○, positioned from disclosed export-revenue shares, the unhedged 44% of external commercial borrowing ◐16, external debt at 20.8% of GDP ◐6 and remittances near $140bn ◐28. Positions are indicative, not measured.
TABLE 13.1 · What a break does to each party, and when it showsIF
StakeholderWhat a break doesWhen it shows
HouseholdsRoughly 40bp on inflation from a 6% move, concentrated in gold, edible oils and pulses; remittance recipients gain outright152–4 months
BanksLittle direct currency risk, then credit cost as unhedged borrowers deteriorate; incremental capital rules on unhedged exposures already apply2–4 quarters
Non-bank lendersForeign-currency borrowing has been steadying funding costs, and the RBI flags the exchange-rate vulnerability that comes with it even where most is hedged161–3 quarters
The sovereignExternal debt is 20.8% of GDP and short-term debt is 21.6% of reserves, both worse than three years ago; the fiscal hit runs through the fuel bill61–2 quarters
Global investors in India fundsFull translation loss with no operating offset and, in practice, no hedge (page 19)2–6 quarters
Gravitywell Research18 / 35
Part III · The Impact MapGWW-2026-003

§ 14 · Private capital IThe rupee is the return currency

In one line · a fund can hit its rupee plan in full and still hand its investors a materially worse number

India funds have always carried a currency tax, and the industry prices it openly: to deliver 15% in dollars a manager underwrites closer to 18–20% in rupees.21 What almost nobody does is hedge it. Fund-level currency hedging over a seven-to-ten-year life is expensive, imprecise against an uncertain exit date, and awkward to fund from committed capital, so in practice the exposure is passed straight through to the investor. That arrangement works while depreciation runs at the 4% a year the crawl has delivered. It works considerably less well if the same distance is covered in a quarter, because the exit date does not move to accommodate it.

FIG 16.1 · One rupee outcome, four dollar outcomesIF2021 vintage, 2.2× in rupees
2.20× in rupees — the plan, delivered in full exit at 96.46, todayexit at 100exit at 106exit at 112 1.69× · 11.0% IRR 1.63× · 10.2% IRR 1.54× · 9.0% IRR 1.45× · 7.8% IRR 1.40×1.85×2.30× Axis starts at 1.40× so the differences are visible. Dollar multiple = rupee multiple × entry rate ÷ exit rate; IRR over a five-year hold.
Sources: the 18–20% rupee for 15% dollar underwriting convention ◐21. The ladder is arithmetic on a stylised 2021-vintage position ○ and assumes no fund-level hedge, which is the market convention rather than a universal fact.

The give-up between the first row and the third is 0.15 of a multiple, or a little over two points of annualised dollar return on a five-year hold. That is not catastrophic, and stated as a single number it sounds almost tolerable. It stops sounding tolerable when you notice that it is roughly a fifth of the excess return the asset class exists to deliver, that it lands on vintages already past the point where anything can be done about it, and that it arrives without a single portfolio company missing a budget. Deployment has already been reacting: Indian private equity and venture investment ran $20.5bn across 604 deals in the first half of 2026, 36% below the same period of 2025, with April's $2.7bn attributed in the trade press directly to the currency.18

The counter-argument this desk takes seriously

The India investor base is becoming less dollar-denominated. GWR-2026-IN-005 put Category II alternative-fund domestic capital share on a path to 55–58% by March 2028, and a rupee-funded investor measured in rupees is untouched by any of this. The offsetting fact is that the international pipe is growing faster where it still exists: GIFT City drew more than $5bn of foreign commitments in FY26 against $2bn domestic, on a base of about $39bn.31 The exposure is shrinking as a share and growing in absolute size, and the desk has no clean way to resolve which effect dominates.

Gravitywell Research19 / 35
Part III · The Impact MapGWW-2026-003

§ 15 · Private capital IIThe exit window narrows twice

In one line · exits are struck in rupees and remitted in dollars, so the multiple and the translation move against you together

Realisations were already thinner before any of this. Indian private-capital exits ran $9.4bn across 95 transactions in the first half of 2026, 29% below the $13.1bn of a year earlier, and only $801m of that came through public listings.19 Public issuance is telling the same story: about $5.78bn raised so far in 2026 against $7.32bn at the same point last year, with issuers cutting deal sizes, accepting lower valuations and postponing listings, after records of $22.36bn in 2025 and $20.65bn in 2024.20 A break arrives into a market that has already lost its bid.

FIG 17.1 · Realisations were already falling before the currency did anythingIF$bn
PRIVATE-CAPITAL EXITS, FIRST HALF PUBLIC ISSUANCE PROCEEDS H1 2025H1 20262027 · mild2027 · moderate2027 · severe 13.1 ◐ 9.4 ◐ 8.5 ○ 6.0 ○ 4.0 ○ 0714 CY2024CY20252025 to date2026 to date 20.7 ◐ 22.4 ◐ 7.3 ◐ 5.8 ◐ 01224 Full years above, comparable year-to-date below.
Sources: exits, EY-IVCA first-half 2026 ◐19; issuance proceeds ◐20. The three 2027 bars are scenario estimates conditional on activation ○, not forecasts, and assume the branch definitions on page 22.

The second turn of the screw is the one that does not appear in any of those numbers. An exit is negotiated in rupees against Indian comparables and then remitted to a fund that reports in dollars, so a break lowers the dollar clearing price of a sale at the same moment that higher policy rates are compressing the multiple it clears at. Open-market sales, which were 44% of first-half exit value at $4.1bn across 34 deals, are the most exposed of the routes because they are struck at a screen price on the day.19 Sponsors with the option will reach for continuation vehicles and secondary sales instead, which converts a translation problem into a discount and defers rather than removes it.

Against the house's own record

GWR-2026-IN-003 put full-year 2026 issuance at $11–14bn with 0.55 confidence, which needs $5–8bn in the second half against $5.78bn raised so far. On today's data that call is tracking. A break inside the next two quarters is the cleanest way for it to fail on the low side, and the desk says so now rather than waiting to see.

Gravitywell Research20 / 35
Part IV

The Dynamics

Two loops push the rate away from the crawl and two structural forces pull it back, and the ones pulling back are larger than in any previous Indian episode. That is the honest reason the break is unlikely. It is also why the version that does happen would not look like a bigger version of the version that does not.

Reading pages
21 – 23
The claim
Severe is a different regime, not a larger number
Part IV · The DynamicsGWW-2026-003

§ 16 · ReflexivityThe loop runs through the hedge, not the panic

In one line · the thing that makes a currency move self-feeding here is ordinary risk management, done all at once

Currency spirals are usually described in terms of confidence, which is a way of saying nobody has identified the mechanism. The mechanism in India's case is prosaic. Forty-four per cent of a $180bn external borrowing book sits unhedged because hedging costs 2.82% a year and the crawl has been delivering 4%.169 Move the rate sharply and that calculation inverts for every treasurer at once. Their hedging demand lifts the forward premium, a higher premium raises what the RBI pays to roll its own forward book, and a book that costs more to roll is a defence that has quietly become smaller.

FIG 18.1 · The amplifying loop, in causal orderIFclockwise
1 · The rupee falls faster than the crawl 2 · Unhedged treasurers all decide to hedge in the same week 3 · The forward premium rises 80–250bp 4 · Rolling the RBI's own forward book costs more THE HEDGING LOOP Each step is an ordinary treasury decision. The loop is what happens when several thousand of them are taken in the same fortnight.
Source: Gravitywell Research ○, built on the unhedged share of external borrowing ◐16, the current 12-month premium ◐9 and the RBI forward book ◐2.
TABLE 16.1 · Two amplifiers against two dampersIF
ForceHow it worksSize
Amplifier · the hedging loopAs drawn above: hedging demand raises the premium, which raises the cost of the defence~$79bn of unhedged stock ◐
Amplifier · index flowsForeign holders of accessible-route government bonds lose on the bond and the currency together, and index-tracking money sells both in one instruction~₹75,352cr in since June 2024 ◐23
Damper · the invisiblesServices receipts and remittances are earned in dollars, so a weaker rupee raises their domestic value and pulls repatriation forward$312bn net in FY26 ◐28
Damper · valuationThe real effective rate is already at 100 after a record 108.14 in November 2024, so a break pushes the rupee into undervaluation, where exports and import substitution respond7REER 100.0, April 2026 ◐

Weigh those four and the dampers are heavier, and that is the arithmetic behind a 12–22% band rather than a 40% one. India in 2026 has a larger invisibles cushion than it has ever had, a currency that is no longer expensive on the measure central banks use, and a reserve stock that even after the adjustments on page 10 covers short-term external debt 1.6 times over. None of that prevents a break. It does mean a break has to be driven by something other than the domestic balance sheet, which is why the oil path and the Fed sit at the top of the movers list rather than anything Indian.

Gravitywell Research21 / 35
Part IV · The DynamicsGWW-2026-003

§ 17 · BranchesSevere is a different animal

In one line · the severe branch is not a bigger fall, it is the defence's own hedge turning into a source of stress

Of the eight disorderly episodes since the float, three resolved mildly, four moderately and one, 1991, went to an external programme. That naive split of roughly 37 / 50 / 13 is the starting point for the weights below. This desk moves severe up to 20% for one specific reason and states it rather than burying it: no previous Indian episode was defended with a derivatives book of this size, so no previous episode carried the risk that the defence itself has to be refinanced mid-crisis.

FIG 19.1 · Which branch depends on two things, neither IndianIFmodal outcome
BRENT AVERAGE, NEXT 12 MONTHS × FEDERAL RESERVE PATH Fed cutsFed holdsFed hikes 50bp Brent ~$85Brent ~$100Brent ~$115 no activation no activation mild mild mild moderate moderate moderate severe no activation mild moderate severe Cells are the modal branch, not a probability. The current market pricing sits between the middle and right columns of the middle row.
Source: Gravitywell Research ○, conditioned on the oil sensitivity ◐13 and the Fed's July position, held at 3.50–3.75% with three dissents preferring a hike ◐27.
TABLE 17.1 · The branches, and what is qualitatively different about severeIF
BranchWhat has to be true, and what is different about itPeak rateWeight
MildThe crawl resumes inside two quarters. Adjusted reserves are spent but not exhausted, inflation peaks near 6.0% and the RBI holds. Deployment pauses, exits are delayed rather than repriced100–103~35%
ModerateThe premium reprices 150–250bp, a second deposit window opens at a higher coupon, inflation runs 6.3–6.8% and the RBI hikes 50–75bp into 6.7% growth. The exit window closes for three to four quarters and unhedged borrowers take visible losses103–108~45%
SevereNew channels, not larger numbers. The near-tenor forward book has to be rolled into a market unwilling to take the other side, so the defence's own hedge becomes a source of stress; swap-window deposits face renewal risk, turning the borrowed reserve build into an outflow; index money reverses through the accessible route that had been financing the capital account; and capital-flow measures are debated in public, which reprices India's risk premium rather than its currency and does not reverse when the rate does108–115~20%

Multiply through and the number that matters for anyone buying protection is the unconditional severe tail, at roughly 17% times 20%, or about 3.5% over twelve months. That is the probability the convex trade on page 26 is priced against, and it is a great deal smaller than the headline band. Sizing a hedge off the 12–22% figure would be sizing it off the wrong number.

Gravitywell Research22 / 35
Part IV · The DynamicsGWW-2026-003

§ 18 · PolicyEvery lever left is one that borrows

In one line · the fast tools defer their cost and the cheap tools work slowly, which is why the order they are used in matters

The RBI has handled 2026 with skill and the results show it: the rupee has fallen about 4% in a year with an oil shock running, reserves are near a record, and the governor has been able to say in public that the currency is undervalued and would firm if the conflict eased.714 The question this page asks is not whether the bank is competent. It is what it has left, and what each remaining option costs at what point in the future. Ordered that way the toolkit divides cleanly, and the division explains the sequence of the past six months.

FIG 20.1 · Speed of effect against cost deferred to a later dateIFindicative
Sell spot reserves · 2022, about $110bn Sell forwards · 2026, to a $106.6bn peak Deposit swap window · 2013 and 2026 Raise the policy rate Widen inflow channels further Pledge gold · 1991 Capital-flow measures · severe branch only highlow costdeferred slowimmediate speed of effect on the spot rate The two tools used hardest this year sit in the upper-right quadrant: fast, and paid for later. That is a defensible choice and it is not a free one.
Source: Gravitywell Research ○, positioned from the 2022 reserve drawdown ◐25, the 2026 forward book ◐2 and the swap-window terms ◐3. Axes are ordinal judgements, not measured quantities.
TABLE 18.1 · What each remaining lever costs, and whenIF
LeverWhat it costs, and whenCapacity left
Spot reserve salesVisible immediately in the weekly release, which is the point: it is the honest tool. No deferred cost~$522bn adjusted ○
Forward salesPreserves the headline, moves the cost to the roll date. About $29bn sits inside three months2Already at record use
The deposit windowRaises dollars against a 2029–31 redemption at 6–7.1%, with the RBI carrying the hedge. The 2013 vintage redeemed cleanly in 201633Window closed 31 August; reopening costs more
The policy rateWorks through carry, slowly, at the price of growth the bank has just revised up to 6.7%14Full, and unused so far
Capital-flow measuresFast and, in this desk's view, the most expensive option on the board: it reprices India's risk premium rather than its currency, and that repricing outlasts the episodeSevere branch only
Gravitywell Research23 / 35
Part V

The Playbook

Twelve observables turn all of this into something you can keep on a screen, and the forward premium is the one that moves first. Then the uncomfortable part: run the option maths against this desk's own odds and the tail is not cheap. What follows from that is a different set of instructions from the ones a scenario dossier usually ends with.

Reading pages
24 – 26
The claim
Do not hedge the level; the level is the forward
Part V · The PlaybookGWW-2026-003

§ 19 · SignpostsTwelve things to keep on a screen

In one line · the first four rows move before anything else does, and three of them are published weekly

Every trigger below carries where the reading sits today and how far it has to travel, because a watch list without a baseline is a list of words. Three of the four fastest-moving rows come out of published data on a weekly or monthly cycle, which makes this a dashboard someone can actually maintain rather than a set of principles. The forward premium is first for a reason: it is priced continuously, it is not managed, and it moves before the reserve release that would confirm it.

TABLE 19.1 · Twelve triggers, each with today's reading
IndicatorCycleReading todayTrigger
12-month forward premiumcontinuous2.82%9>4.30% · +148bp to go
3-month realised depreciationdaily~1.5% ○>4% watch · >6% activation leg 2
RBI net short forward bookmonthly$106.6bn at 31 May, unwinding2fails below $75bn by Dec 2026, or back above $110bn
Adjusted ammunition (desk series)weekly~$522bn ○34<$450bn · −$72bn to go
FCNR(B) system outstandingmonthly$60.55bn from $32.56bn on 5 Jun4net redemptions in any month after the window
Hormuz throughputmonthly4.9 mb/d against 21.6 pre-war11stays below 8 mb/d into Q4 2026
Brent, quarterly averagecontinuous~$88, high of $105 on 23 Jul11>$100 · +$12 to go
Foreign debt flow, accessible routemonthlyrecord ₹55,518cr inflow in June23two consecutive months of net outflow
CPI, headlinemonthly4.38% in June; 5.9% projected for Q314any print above 6.0%
Fed funds targetper meeting3.50–3.75%, three dissents to hike27any hike delivered
Real effective exchange ratemonthly100.0 in April, from 108.14 in Nov 20247below 95 · the break has overshot
Public issuance, rolling six monthsmonthly~$5.8bn year to date20below $2.5bn · the exit window has shut
Readings are as of 25 August 2026. Distances to trigger are shown where the metric is continuous.
The read

Rows one to four are the ones that would confirm a break before the rate does, and the fourth is the one nobody else maintains. If you can only watch two, watch the forward premium and the forward book: the first is the market's price for doubting the defence, the second is the quantity of defence that has already been committed. Everything else on this list is confirmation arriving after the decision point has passed.

Gravitywell Research24 / 35
Part V · The PlaybookGWW-2026-003

§ 20 · PositioningWhat to do while nothing is happening

In one line · the useful actions are all about shortening exposure you are not paid to hold, and none of them requires believing this scenario

Most of what follows is worth doing whether or not the break happens, which is the test a recommendation in a scenario dossier ought to pass. If an action only makes sense at a 17% probability, it is a bet dressed as risk management. If it makes sense because you are carrying an exposure you were never compensated for, the probability is beside the point and the dossier has simply told you where to look.

TABLE 20.1 · What to do now, and what changes on activationIF
If you areNow, 0–3 monthsOn activation, hour 1 to week 4
An allocator with dollar exposure to IndiaCover the slice you will actually repatriate inside twelve months with plain forwards at 2.82%. Leave the long-dated book alone: hedging a seven-year exposure with annual rolls is a cost, not a hedge9Stop adding unhedged rupee duration. The roll cost moves first and the marks follow a quarter behind, so the roll quote is your early warning
A general partner mid-harvestRestate the exits you are working on at 100 and 106 and send the difference to your investors before the market does it for you. Bring forward whatever is genuinely sellableMove the marginal exit to a secondary or continuation route and take the discount, because it is smaller and more controllable than the translation
A corporate treasurer with offshore debtEstablish today which of your maturities sits inside the 44% of the external borrowing book that is unhedged, and hedge the near tenor first16The near tenor is where the squeeze lands, and the premium there reprices before the far end. Hedging after that point is buying insurance during the fire
A bank or non-bank risk officerReprice unhedged foreign-currency exposure in the borrower book now. The concentration is in smaller firms, which hedge 38% against 68% for large corporates16Expect the credit cost two to four quarters after the currency move, not with it
A policymakerPublish an adjusted reserve position, because a number the market can reconstruct is a number worth publishing first. Credibility is cheaper to keep than to rebuildUse the slow, honest lever early rather than the fast, deferred one late. The order of the toolkit on page 23 is the whole decision
What this dossier is not telling you to do

Nothing here recommends a directional position on the rupee. The desk's view on the level is the same as the street's and roughly the same as the forward's, so there is no trade in it. Nor is there a recommendation to shift an India allocation, because a 17% probability of a two-to-three-point give-up in annualised dollar return is not a reason to leave an asset class. What there is, on the next page, is an honest calculation of what the tail protection actually costs and whether it is worth buying at today's prices.

Gravitywell Research25 / 35
Part V · The PlaybookGWW-2026-003

§ 21 · The tradeThe convexity is real and it is not free

In one line · a twenty-to-one payoff that returns 0.6 times its cost on this desk's own odds is not a trade, it is a temptation

A twelve-month option to buy dollars at 106 costs roughly 31 basis points of notional and pays about 6.2% if the rate reaches 112, which is a payoff ratio near twenty to one. That is exactly the shape a scenario dossier is supposed to end on, and this one is going to decline to. Run the same structure against the probabilities in this report rather than the ones implied by the curve, and it returns about 0.62 of what it costs.

FIG 21.1 · Excellent payoff shape, odds already pricedIF12m call at 106, % notional
mild · w 35%moderate · w 45%severe · w 20% breakeven 106.3 +6.2% at 112 premium −0.31% 0+3+6+9% 96100104108112115 USD/INR at expiry · shaded zones are branch ranges with their conditional weights, not probabilities of reaching a level
Source: Gravitywell Research ○, priced off the 12-month forward of 99.189 at 5.5% assumed volatility, discounted at a 3.6% dollar rate. Live implied volatility for USD/INR is not obtainable from public sources, so the volatility input is an assumption and the table below states what it would take to change the answer.
TABLE 21.1 · Four structures priced against the desk's own oddsIF
StructureCostPayoff at 112Return on this desk's oddsVerdict
12m call at 1021.08% of notional10.4% · 9.6×0.57× costBuy only below ~4.1% implied volatility
12m call at 1060.31%6.2% · 20.0×0.62× costBuy only below ~4.9% implied volatility
12m forward on the repatriation slice2.82% carryremoves the exposurenot a betDo it regardless of this scenario
Roll short-dated forwardsfront-end carrygains as near premia spikethe direct path expressionThe $29bn roll lands here first2
Premiums are Black-Scholes on the 99.18 forward at 5.5% assumed volatility. Returns use this dossier's branch weights, not risk-neutral ones.
The read

The reason the tail is not cheap is that this desk is less bearish than the curve, and when your view sits inside the market's, buying protection means paying for someone else's larger fear. So the asymmetry to own is not the option. It is the plain forward on money you will actually bring home inside a year, which removes a risk you were never being paid to carry, and the front-end roll, which is where the defence's own refinancing lands. If you can get twelve-month volatility below roughly 4.9%, buy the 106 call and size it against a 3.5% severe tail rather than the 17% headline. Above that, the honest answer is that the market has already thought about this.

Gravitywell Research26 / 35
Pre-mortem & Red-teamGWW-2026-003

FalsificationHow it happened, and why it may not

Pre-mortem · it is October 2027 and the break has happened

Looking back, the sequence was ordinary. The Hormuz talks broke down in late September 2026 and Brent settled above $100 for a quarter, which put the current account past 2.5% of GDP. The Fed delivered its first hike in December and the dollar did the rest. The rupee went through 100 in the second week of January, and the RBI met it the way it had met everything that year, in the forward market. What nobody had modelled was that a book already at record size could not be rolled quietly: the three-month premium went from 2.8% to 5.4% in eleven sessions, and the corporates who had been running unhedged all arrived at the same time. By February the deposit window had to be reopened at 8%. The reserve headline never fell much, which was the odd part, and the commentary kept pointing at it right up until the point where the composition of the number became the story.

The disconfirming view — the strongest case against this dossier

The rupee is not expensive. The real effective exchange rate was 100.0 in April, down from a record 108.14 in November 2024, and the governor has said publicly that the currency is undervalued on both nominal and real measures.7 On the only valuation metric with theory behind it there is nothing left to correct, so a break past 100 would be an overshoot, and overshoots mean-revert. The maturity wall has been walked before. The 2013 deposit vintage matured through late 2016, about $26bn with over $17bn in November alone, roughly 80% pre-covered in forwards, and banks redeemed more than 90% of it with no disturbance to the currency market.33 The offsets have never been larger. Net invisibles reached $312bn in FY26 and the March quarter produced an outright current-account surplus of $7.1bn.28 And the forward has a track record of crying wolf: rupee forward premia have historically run ahead of realised depreciation, so the curve leaned on here has been wrong in the same direction before.

Three of those four cap the severity rather than the odds of activation, so they do not move the band. A large invisibles account determines how far a break travels, not whether an oil shock and a hiking Fed can push the rate through a level in a quarter. The 2016 precedent lowers the weight on the severe branch's deposit-renewal channel and it is priced there, not in the headline. The forward's historical bias is already reflected, since it is the reason the touch band was shaded from 80–88% down to 70–85%.

The objection that stands

The real effective rate does not fit that pattern, and this desk has no answer to it. If the rupee is genuinely at or below fair value, the fundamental case for further depreciation is a flow story rather than a valuation story, and flow stories reverse when the flow does. A Hormuz settlement would reverse this one inside a quarter. The desk carries this to the bottom line unresolved: the single best argument against the 12–22% band is that it is being applied to a currency that is not overvalued, and the only defence is that the 2013 and 2022 episodes also began from a real effective rate near or below 100.

Gravitywell Research27 / 35
Probability & assumptionsGWW-2026-003

The ledgerWhat the call rests on

In one line · the oil path moves the answer more than everything else on this page combined

The band of 12–22% is a product of three conditional legs rather than a single judgement: a 0.35–0.45 chance the level holds for twenty sessions, a 0.40–0.50 chance the move is fast enough given that it happens, and a 0.85–0.90 chance the defence is visibly spent given the speed. Move any of those and the band moves with it. The chart below asks which underlying assumption does the most work, and the answer is not an Indian variable.

FIG 22.1 · What would move the probability, and by how muchIFpp on P(break), base 17%
The oil path · settlement vs $115 Brent The Fed · cuts vs a 50bp hike Gold treatment in adjusted reserves Deposit-renewal behaviour in 2029–31 Inflation pass-through elasticity −7+9 −3+6 −4+3 −2+4 −1+2 5%10%17% base25%30% Teal = the assumption breaking in the direction that lowers the probability. Crimson = the direction that raises it.
Source: Gravitywell Research sensitivity ○. The oil swing is the widest at 16 points, more than five times the pass-through row and roughly twice the Fed row.
TABLE 22.1 · Consensus, the house delta, and what would close it
Consensus and the variantWhat it says
The consensusStreet houses at 96–99 over two years, with 100 framed as needing a major oil or dollar shock and explicitly not a 2026 base case26; the forward market, which is the consensus people pay to hold, sits at 99.189
The house deltaThe desk is at consensus on the level and says so. It departs on two numbers nobody publishes: a touch probability of 70–85%, which is higher than "not a base case" implies, and a break probability of 12–22% against a strict base rate of 8–9%
Why the gap persistsThe level is what every house forecasts and every reader remembers. The path is priced only in the forward premium and in the composition of the reserve stock, and few India allocators mark either
What proves consensus rightThe rate reaches 100 with rolling three-month depreciation under 4% and adjusted reserves flat, which on these numbers is the most likely single outcome; or a Hormuz settlement returning Brent to the $60s
Load-bearing assumptionsGold and the forward book are unavailable for spot defence · swap-window deposits are a dated liability rather than accretion · fund-level currency hedging is negligible · Hormuz does not normalise before early 2027. If the first breaks, activation leg three is far harder to fire and the band belongs lower
Gravitywell Research28 / 35
Bottom lineGWW-2026-003

Bottom lineProbably 100. Probably not a break.

The rupee will very likely trade at 100 in the next twelve months, and that fact carries no information because the forward market has already sold it. What this dossier prices is the roughly one-in-six path where the rate covers that ground in a quarter rather than a year, spends real ammunition getting there, and reprices the cost of hedging for everyone at once. Against a strict base rate of 8–9% the desk puts that at 12–22%, and the uplift is earned by two things: the level is 3.7% away, and the machinery used to smooth it is already committed to a degree no previous Indian episode matched.

Touch 100
70–85%
The market's own arithmetic, not a house view
The break
12–22%
Level held, crawl broken, defence visibly spent
Severe, unconditional
~3.5%
The number a hedge should be sized against

The damage, if it lands, lands somewhere the tape does not show you. Headline reserves rose $24bn in the fortnight to 14 August while the rupee sat within half a rupee of its record low, because roughly $28bn of borrowed deposits arrived in the same eight weeks and about a sixth of the stock is metal. Forty-four per cent of a $180bn external borrowing book carries no hedge, concentrated in the firms that hedge least. And a global investor in an Indian fund sits at the most exposed point on the whole map, receiving rupees and reporting in dollars, giving up two to three points of annualised return in a moderate break without a single portfolio company missing a number.

One objection to all of this survives the report intact. The real effective exchange rate is 100.0 and the governor says the rupee is undervalued, which means there is no valuation gap left to close and a break would be an overshoot. The desk has no answer beyond noting that 2013 and 2022 also started from a real rate near fair value. That objection deserves to sit on the last page rather than be tidied away three pages earlier.

The position

Do not buy the level, because the level is the forward. Do not buy the tail on faith either: at 5.5% assumed volatility the twelve-month option struck at 106 returns 0.62 of its cost against this desk's own odds, and only turns positive below roughly 4.9%. What is worth doing at any probability is removing exposure you were never paid to carry, which means plain forwards on the money you will genuinely repatriate inside a year and hedging the near tenor of an offshore book before the premium reprices. The asymmetry here is not in an option payoff. It is that a 2.82% carry buys certainty on a translation risk that costs two to three points of annual return when it goes wrong, and the people carrying it mostly do not know they are.

Gravitywell Research29 / 35
MethodologyGWW-2026-003

MethodologyHow this was built

Scope

In scope: the USD/INR exchange rate and its transmission into India's external balance, domestic prices, policy rates, listed markets and private capital, over a twelve-month horizon from 25 August 2026. Out of scope: cross-rates other than the dollar, the structure of India's payment system, and any security-level recommendation. The dossier studies a defined shock conditionally; it does not forecast the exchange rate.

How the key figures were derived

The touch probability. The 12-month forward outright of 99.18 comes from a published forward premium of 2.82% applied to spot. The terminal probability of finishing above 100 is a lognormal calculation at 4–7% assumed volatility, giving 41–44%. The barrier probability of touching 100 at any point is estimated by reflection at roughly twice the terminal figure, giving 80–88% risk-neutral. The house band of 70–85% shades that down for the documented tendency of rupee forward premia to exceed realised depreciation. Live implied volatility is not obtainable from public sources; the volatility input is therefore an assumption and every figure derived from it is tiered ○.
The break probability. Decomposed into three conditional legs (0.35–0.45 × 0.40–0.50 × 0.85–0.90) rather than asserted, giving 12–22%. The strict base rate of 8–9% counts three speed-clearing episodes in 35 years.
Adjusted ammunition. Headline reserves less 50% of gold, less the SDR and IMF reserve position, less the net short forward book, less swap-window deposits. The gold haircut is a desk judgement: gold is pledgeable and was pledged in 1991, but slowly and at a cost in credibility, so excluding it entirely overstates the case and counting it fully understates it.
Option pricing. Black-Scholes on the forward at 5.5% assumed volatility, discounted at 3.6%. Expected values use this dossier's own branch weights rather than risk-neutral ones, so the answer differs from the market's.

Source tiers and point-in-time policy

● filed or official · ◐ derived, reconciled, or primary reached through a secondary outlet · ○ a Gravitywell Research estimate or construction. Where a primary release was read through press reporting rather than the original document, it is tiered ◐ and not ●. Two published counts of the 2026 swap window ($27.99bn of net new system balances against $52.3bn reported under the facility) could not be reconciled from public sources, and both are printed rather than one being chosen. Figures are as of 25 August 2026 and are not silently restated; a material correction publishes as a dated erratum.

Dataset minted with this dossier

india-adjusted-fx-ammunition — a weekly series decomposing headline RBI reserves into gold, the SDR and IMF position, the net short forward book and swap-window borrowed dollars, leaving deployable spot ammunition, expressed against short-term external debt on a residual-maturity basis. Definition, sources, method, vintage policy, cadence and known gaps are in the dataset's own README. The adjusted-reserve critique itself is not original to this desk and is not claimed as such; the maintained weekly series is what the desk contributes.

Gravitywell Research30 / 35
MethodologyGWW-2026-003

Methodology IIConfidence, review and data

Confidence rationale, stated as Medium

The transmission mechanisms rest on primary or official data and the main figures reconcile across independent sources. The probability estimates do not: they depend on an assumed volatility that could not be verified, on a reference class of eight episodes which is small, and on a gold haircut that is a judgement. Directionally confident, numerically approximate.

Red-team procedure, and where it fell short of house standard

House practice is a three-panellist adversarial review run independently of the author, before drafting and again at the publish gate. This session was unable to run either panel, so both the pre-drafting red team and the publish-gate review were conducted by the same analyst who wrote the dossier, structured against the same briefs but not independently sourced. Shared blind spots are therefore uncontrolled, and this is disclosed rather than glossed. Six material changes came out of it and are recorded in the commission file: the forward was reframed from an expectation to a hedgeable rate, the touch band was shaded down, gold was haircut rather than excluded, the 2016 redemption precedent was printed at full strength, the private-capital node was sized rather than asserted, and the real-effective-rate objection was carried to the bottom line unresolved.

Prior calls on adjacent terrain, and how they are tracking
CallWhat we saidStatus at 25 Aug 2026
GWW-2026-002 · Aug 2026The widened external-borrowing pipe reprices rather than rations, p 0.75–0.85 to Aug 2027Open. This dossier argues the reprice arrives through the currency rather than the spread
GWW-2026-002 · Aug 2026Placed USD/INR 97–99 in its moderate branch and above 100 only in severeMis-placed. Spot is 96.5 and the forward prints 99.18 with no credit seizure at all
GWR-2026-IN-003 · Aug 2026CY2026 public issuance of $11–14bn, p 0.55Tracking. $5.78bn raised year to date needs $5–8bn in the second half
Known gaps

Three things would materially improve this analysis and none was obtainable publicly. Live USD/INR implied volatility would replace the assumption every probability and option price here rests on. A current reading of the net short forward position, rather than the 31 May figure, would fix the largest input to the adjusted-reserve series. A measured estimate of fund-level hedging by India managers would convert the private-capital section from a convention-based argument into a sized one. Each is flagged where it is used.

Gravitywell Research31 / 35
Source registerGWW-2026-003

Source registerRows 1 to 17

Every figure in this dossier carries a superscript pointing to a row below. Tier glyphs follow the house policy: ● filed or official, ◐ derived or a primary release reached through a secondary outlet, ○ a Gravitywell Research estimate or construction. Where a number could not be reconciled across sources, both readings are printed rather than one being chosen.

#TierSource
1RBI Weekly Statistical Supplement via Business Standard, 21 Aug 2026 - FX reserves $716.907bn (wk to 14 Aug), +$9.905bn; prior week +$14.136bn to $707.002bn; FCA $581.851bn; record $728.494bn (wk to 27 Feb 2026)
2Bloomberg, 30 Apr / 10 Jul 2026 + Business Standard 22 Jul 2026 - RBI net short dollar forward book $67.8bn (Jan-26) -> record $103bn (Mar-26) -> record $106.6bn (31 May 2026); ~$29bn in <3m tenors; unwind under way using FCNR(B) inflows
3RBI press release 5 Jun 2026 + Business Standard 9 Jun / 17 Aug 2026 - FCNR(B) USD-INR swap window: fresh/renewed 3-5y deposits booked 8 Jun-30 Sep 2026, deadline advanced to 31 Aug 2026 (swaps executed by 11 Sep); RBI bears the hedging cost
4System FCNR(B) outstanding $60.55bn vs $32.56bn on 5 Jun 2026 = $27.99bn net new in eight weeks (bank/RBI data via NRI Affairs, Aug 2026). Separately reported as $52.3bn 'under the facility' by 13 Aug (includes renewals) - CONFLICT, both printed
5RBI, NRI deposits outstanding $165.96bn end-May 2026: FCNR(B) $34.04bn, NRE $98.39bn (vs $101.86bn May-25); FY26 NRI deposit inflow $14.4bn, Apr-Feb flow -24% to $11.04bn
6RBI external debt release, 29 Jun 2026 - external debt $762.8bn end-Mar 2026, 20.8% of GDP (19.8% prior yr); short-term (original maturity) 19.6% of total (18.3%); ST debt/reserves 21.6% (20.1%); reserves cover 90.3% of external debt vs 100.6% in 2021; import cover 10.8 months
7RBI Bulletin - REER 100.0 (Apr 2026) from 101.7 (Mar); 101.49 (Mar), 102.37 (Feb); record 108.14 (Nov 2024). Governor Malhotra: rupee undervalued on nominal and REER terms
8RBI Annual Report 2025-26 - gold 880.52 tonnes end-Mar 2026, 680.05t held domestically; gold share of reserves ~15%, up from <7% a decade ago
9Capera USD/INR forward curve - 12-month forward premium 2.82% annualised, 10 Jul 2026 (implies 1y outright ~99.2 from spot ~96.5)
10Reuters / InvestingLive, Aug 2026 - USD/INR ~96.4575 intraday, RBI selling dollars onshore and offshore; record low 96.9650 late May 2026; weakest major Asian currency for much of 2026
11EIA Short-Term Energy Outlook (throughput and the 2026 Brent average, primary) plus Al Jazeera/CNBC market reporting (spot prints, press), Jun-Aug 2026 - Hormuz crude+liquids throughput 4.9 mb/d in Q2 2026 vs 21.6 mb/d in 4Q2025; Brent spot high $105/b on 23 Jul, settled $87.72 on 10 Aug; Middle East output not expected near pre-conflict levels until early 2027; EIA sees Brent averaging $87 in 2026
12Crisil Ratings, Jun 2026 - India CAD to widen to 2.2% of GDP in FY27 from 0.6% in FY26; base case 1.5% at $75-80 Brent; 2.0% at $82-87; Brent assumed $90-95 average this fiscal (+32% YoY). HDFC Bank (May 2026): 2.1% at $85
13ICRA / market rule of thumb - every $10/bbl on average crude adds $13-14bn (ICRA) to $14-16bn to net oil imports and 30-40bp to CAD; at $110-115 FY27 the increment is $56-64bn; India imports >85% of crude
14RBI Monetary Policy Statement, Aug 2026 (Governor Sanjay Malhotra) - repo held 5.25% for a fourth consecutive meeting, neutral stance; FY27 CPI 5.0% (Q2 4.7%, Q3 5.9%, Q4 5.5%), core ~4.3%; FY27 GDP raised to 6.7%; 'neither dovish nor hawkish'; rupee could firm if geopolitics de-escalate
15RBI estimate via press - a 5% rupee depreciation adds ~35bp to CPI over several months; alternative estimates 15-25bp annualised; food is ~46% of the CPI basket with near-full domestic self-sufficiency, concentrating pass-through in gold, edible oils and pulses
16RBI (Jun 2026) - of $180bn outstanding ECBs, 44% unhedged; SME FX liabilities 38% hedged vs 68% for large corporates (FY23 firm-level study, Goel/SenGupta/Das 2025)
17Foreign Exchange Management (Borrowing and Lending) (First Amendment) Regulations, 2026, effective Feb 2026 - automatic-route ECB cap raised to the higher of $1bn outstanding or 300% of net worth (from $750m/yr); no general hedging mandate; 100% hedging only for average maturity <5y, 70% for infrastructure finance companies
Gravitywell Research32 / 35
Source registerGWW-2026-003

Source registerRows 18 to 34

#TierSource
18EY-IVCA, Jul 2026 - India PE/VC 1H2026 $20.5bn across 604 deals, -36% YoY ($31.8bn) and -29% vs 2H25; deals -18%; pure-play PE/VC $14.2bn (-29%); real estate & infrastructure $6.3bn (-47%). April 2026 alone $2.7bn across 83 deals, halved, attributed in press to rupee depreciation near INR 96
19EY-IVCA / Business Standard, Jul 2026 - PE/VC exits $9.4bn across 95 transactions in 1H2026, -29% from $13.1bn in 1H2025; open-market exits $4.1bn across 34 deals (44% of value); IPO exits only $801m in 1H2026
20Bloomberg, 3 Aug 2026 - India IPO proceeds ~$5.78bn YTD 2026 vs $7.32bn a year earlier; record $22.36bn in 2025 and $20.65bn in 2024; issuers cutting deal sizes, accepting lower valuations, delaying listings
21Outlook Business, 2026 - to earn a 15% dollar IRR India funds underwrite closer to 18-20% in rupee terms; currency is explicit in PE underwriting and exit multiples, more diluted in VC where target returns are 30-35%; domestic-revenue businesses absorb the depreciation in entry valuation
22Business Standard / NSDL, Aug 2026 - FPIs net sellers of ~Rs 2.4trn of Indian equities in 2026 YTD, exceeding the Rs 1.66trn full-year 2025 outflow; Rs 20,199cr bought in July and Rs 16,621-23,544cr in August to date - a two-month reversal
23NSDL via Business Standard, Jun 2026 - record Rs 55,518cr ($5.8bn) FPI debt inflow in June 2026 (general limit Rs 30,620cr, FAR Rs 21,652cr), more than offsetting Rs 49,340cr of equity outflow; FAR securities have drawn ~Rs 75,352cr since JPMorgan GBI-EM inclusion (28 Jun 2024)
24RBI / press, 2013 calibration - rupee record low 68.85 on 28 Aug 2013; RBI mobilised ~$26bn via the FCNR(B) 3-year swap window at 3.5% fixed, covering ~80% in forwards, against $365bn of reserves; rupee -11 to -12% over calendar 2013
25Historical annual INR moves - 2008 -24% (worst in a decade), 2011 -14.4% (including a record -7% in November alone), 2013 -11 to -12%, 2022 -10%+ (worst since 2013); long-run average ~4.5%/yr since 1991
26Market/press consensus, 2026 - ANZ, Bank of America, Goldman Sachs and Danske see USD/INR 96-99 over the next two years; the depreciation norm re-based from ~4%/yr to a 4-10% range averaging ~6.5%; a move to 100 framed as possible only on a major oil or dollar shock and not a base case for 2026
27Federal Reserve, Jul 2026 - fed funds held at 3.50-3.75% for a fifth consecutive meeting with three dissents preferring a 25bp HIKE; markets assigned roughly a one-in-three probability to a hike; no cuts delivered in 2026
28RBI, Q4 FY26 balance of payments - current account SURPLUS $7.1bn (0.7% of GDP) in Jan-Mar 2026 on services and remittances; FY26 full-year CAD $25.2bn (0.6% of GDP); net invisibles $312bn vs $264bn; net services receipts $60.4bn in Q4 vs $53.3bn; remittances a record ~$140bn
29Ministry of Commerce, Jul 2026 - merchandise trade deficit $31.98bn in July 2026 vs $27.5bn a year earlier; imports a record $76.2bn, +17.5% YoY, on higher oil, fuel and coal prices
30Brokerage estimates via mstock, 2026 - a 3-4% rupee move in Q4 FY26 lifted top-six Indian IT EBIT margins 40-320bp YoY; ~1% INR depreciation adds ~30-40bp to operating margin; 60-70% of top-tier IT revenue is USD/EUR denominated
31IFSCA / press, 2026 - GIFT City IFSC AIF commitments ~$39.09bn as of Mar 2026 across 200+ fund management entities and 350+ schemes; FY26 drew >$5bn foreign commitments against $2bn domestic
32Desk construction - reflection-principle estimate of P(touch 100) from the 12m forward outright (source 9) at 4-7% assumed vol; terminal probability 41-44%, barrier probability 80-88%. Implied vol not obtainable from public sources; the vol range is a desk assumption and the estimate is tiered accordingly.
33Business Standard / Gulf News, Sep 2016-Feb 2017 - the 2013 FCNR(B) vintage matured Sep-Nov 2016: ~$26bn, over $17bn in November alone; ~80% pre-covered in forwards; banks redeemed >90% with no FX-market volatility; RBI injected Rs 10,000cr via OMO; the period coincided with demonetisation
34Desk series - india-adjusted-fx-ammunition: headline $716.9bn less 50% of gold (~$53.8bn of a ~15% share), less SDR + IMF reserve position (~$23bn), less the net short forward book (~$90bn, from the $106.6bn 31-May peak, unwind in progress), less FCNR(B) swap-window borrowed dollars (~$28bn net new) = ~$522bn. Gold haircut is a disclosed desk judgement, not a disclosure.
Gravitywell Research33 / 35
Glossary & exhibitsGWW-2026-003

ReferenceTerms and exhibits

Glossary
Forward premium
The annualised gap between forward and spot, set by the interest-rate difference between the currencies. An arbitrage relation rather than a forecast, but the rate at which a hedge transacts.
Forward outright
Spot adjusted by the premium: the level at which dollars can be bought today for a future delivery. Here, 99.18 for August 2027.
Risk-neutral vs real-world probability
A probability implied by market prices against one a forecaster believes. They differ by the compensation demanded for bearing risk, so an option can look cheap on payoff and dear on odds.
Reflection estimate
The chance a price touches a level at any point before a date rather than finishing beyond it. Near the forward it is roughly twice the finishing probability.
Implied volatility
The volatility an option's price implies. It is the input this dossier could not observe and therefore assumed.
REER
Real effective exchange rate: the currency against a trade-weighted basket, adjusted for relative inflation. Above 100 the currency is expensive against its 2015–16 base, below 100 it is cheap.
FCNR(B)
Foreign Currency Non-Resident (Bank) deposits: dollar deposits placed with Indian banks by non-residents, on which the depositor bears no rupee risk.
The swap window
A facility under which the RBI takes the dollars a bank raises through such deposits and returns them at maturity, absorbing the hedging cost.
External commercial borrowing
Foreign-currency debt raised abroad by an Indian company. Written in full here because its abbreviation collides with the European Central Bank.
FAR
Fully Accessible Route: the class of Indian government bonds open to foreign investors without limit, and the class included in global bond indices since June 2024.
Residual maturity
Debt measured by time left to run rather than original tenor. Short-term debt is much larger on this basis.
Pass-through
How much of a currency move reaches consumer prices. Small in India, because food dominates the basket and is grown at home.
Invisibles
The non-goods current account: services, remittances and investment income. Large enough to offset most of India's goods deficit.
MOIC
Multiple on invested capital: what a fund returns per unit invested, before annualising.
Continuation vehicle
A new fund raised by a manager to buy assets from its own older fund, giving investors liquidity without a third-party sale.
Open-market exit
A sponsor selling listed shares on the exchange rather than by block or strategic sale. It clears at the screen price that day.
List of exhibits · figures below; tables are numbered TABLE n.n in the body
FIG 1.1Two probabilities on one rulerFIG 13.1Total return by horizon and currency
FIG E.1Likelihood and size in separate panelsFIG 14.1Operating-margin sensitivity by sector
FIG 2.1The sequence, and where the path forksFIG 15.1Dollar receipts against dollar obligations
FIG 3.1The 100 line inside the forward coneFIG 16.1One rupee outcome, four dollar outcomes
FIG 4.1Narrowing the reference classFIG 17.1Realisations and public issuance
FIG 5.1The same tool, thirteen years apartFIG 18.1The amplifying loop, in causal order
FIG 6.1Supply shock and current accountFIG 19.1Which branch, by oil and the Fed
FIG 7.1From headline reserves to deployableFIG 20.1Speed of effect against deferred cost
FIG 8.1The transmission and what is pricedFIG 21.1Payoff shape against the odds in the price
FIG 9.1Incremental annual financing needFIG 22.1What would move the probability
FIG 10.1Cost of protection and the calendar
FIG 11.1The unhedged book and its burden
FIG 12.1The projection against the tolerance band
Gravitywell Research34 / 35
Disclosures & governanceGWW-2026-003

DisclosuresGovernance and terms

The series disclaimer

This is a conditional scenario analysis, not a forecast or prediction. It sets out what would follow if a stated activation condition occurred, weighted by how likely that condition is. Nothing in it is a view on where the rate will be on any date.

Certification

The Gravitywell Research desk responsible for this report certifies that the views expressed accurately reflect its independent judgement about the subjects and instruments discussed, and that no part of its compensation was, is, or will be tied to the specific recommendations or views expressed herein.

Positioning & conflicts

As of the publication date, Gravitywell Research and its analysts do not hold positions in the securities, currencies or assets discussed, and the firm has no advisory, banking or commercial relationship with any institution named. This report was not commissioned or reviewed by any party named in it.

House view scale

Conviction is High, Medium or Low, set by coverage depth, source tier and how far the thesis has been stress-tested; Stance is Constructive, Neutral or Cautious. This dossier is Medium conviction, Cautious stance: cautious on the asymmetry of an unhedged return currency, not on the direction of the rate, where the desk sits with consensus.

Distribution

Prepared for readers who allocate capital, underwrite risk, or set policy. Not for general retail distribution, nor for readers who lack the expertise to assess the assumptions. Recipients may not redistribute without attribution, and availability in some jurisdictions may be restricted.

Errata and permalink

Material errors are corrected in a dated erratum and noted in the next edition; the permalink always serves the current version. Figures are marked to 25 August 2026 and are not silently restated. Two published counts of the 2026 swap window could not be reconciled and both are printed. The probability estimates rest on an assumed volatility input that could not be verified publicly, disclosed on the methodology page and tiered ○ wherever used.

Independence of review. House practice requires an adversarial review run independently of the author. That was not possible for this issue: both the pre-drafting red team and the publish-gate review were run by the writing analyst against the same briefs. Any reader relying on the probability estimates should weigh that.

Gravitywell Research35 / 35
Gravitywell
Research
SCENARIO DOSSIER · WHAT-IF
GWW-2026-003
AUGUST 2026
CONDITIONAL SCENARIO · NOT A FORECAST

The rupee will probably touch 100, because the forward curve already prints 99.18. It will probably not break. Do not hedge the level; cover the money you will actually bring home inside a year.

India Macro · The Rupee
© 2026 Gravitywell Researchgravitywellresearch.xyz