Gravitywell Research · What-If
Scenario Dossier · GWW-2026-002

What if private credit cracks?

The first real test of the $1.6 trillion US private credit market would reach India through its lenders, not its loans. The wrapper losses in America are deeply priced already. The commitment freeze that follows is not: a funding winter that meets a refinancing-heavy Indian credit book at its maturity dates, through a channel no Indian instrument prices. Moderate odds, slow mechanics, and a convex payoff: small fixed cost, outsized gain if it hits.

P ≈ 15–25%over 18m · loose base ~20–26% · strict ~7–9%
~$1.6trn
US private credit, end-2024 — the machine under test ●
12.1% v 5%
Q1-26 redemption requests vs the quarterly gate, twelve largest wrappers ◐
6.0%
Fitch default rate, the young index's high — against spreads in their richest decile ◐
71%
foreign share of GIFT City's FY26 fund commitments — the India pipe under test ●
Gravitywell Research Global Credit & Capital Markets Desk · Aug 2026
Medium conviction · Cautious on the tail
Conditional scenario · not a forecast
Gravitywell Research · GWW-2026-002Contents

Scenario DossierContents & conventions

PartSectionPage
The Shock · the activation condition03
Executive scenario summary04
Scenario Quick Read05
I · The Setup§01 Anatomy of the machine · §02 The base rate · §03 The clocks · §04 The calibrations · §05 Why now06–10
II · The Transmission§06 The chain · §07 The wrapper run & the primary strike · §08 The extension machine & the marks · §09 Insurers & the AI-datacentre stack · §10 The India channel11–15
III · The Impact Map§11 By asset · §12 By sector · §13 By geography & stakeholder · §14 Private capital: the LP transmission · §15 Private capital: the India book16–20
IV · The Dynamics§16 Reflexivity · §17 The branches · §18 The policy-reaction function21–23
V · The Playbook§19 Signposts · §20 Positioning by stakeholder · §21 The convex trade24–26
Pre-mortem & red-team · Probability & assumptions · Bottom line27–29
Methodology · Sources · Glossary · Disclosures30–33
About Gravitywell Research

Gravitywell Research publishes independent research for professional investors across public and private markets, and for the risk officers and policymakers who must plan for tails. The What-If series studies defined shocks conditionally: we state the trigger, weigh it against base rates, trace the mechanism, and end in positioning. Every figure is sourced to primary filings or official statistics where they exist, tiered for confidence, and marked to its date.

Conventions & data vintage

Currency: USD, with INR for Indian series. Markets as of 13 Aug 2026; fund data to the latest reported quarter.37 "The wrappers" = the redeemable vehicles through which retail owns private loans: non-traded BDCs, interval and evergreen funds. "Marks" are appraised NAVs; "prices" are traded levels — we never interchange them. Estimates are ranges. Confidence tiers: ● filed/official · ◐ modeled/derived · ○ Gravitywell estimate. Magnitude and probability are separate axes throughout. Reader's key: bp = basis points (100bp = 1 percentage point) · "the tape" = live market prices · a "print" = an actual traded price · "richest decile" / "cycle tights" = spreads this low only ~10% of history · LPs = the investors in a fund; GPs = its managers · "the put" = the market's faith that the central bank steps in · prices like "82" = cents per dollar of stated value.

How to read this dossier

Every effect below carries a mechanism, a dated and tiered magnitude, an order in the chain, and a source. Keep two questions apart as you read: how big this would be, and how likely it is. A vivid cascade is never evidence that the event is probable, and a modest probability is never a reason to skip the insurance arithmetic. This is a conditional study of a defined trigger — not a prediction that private credit fails, and not a claim that the asset class lacks a future.

Gravitywell Research02 / 33
Gravitywell Research · GWW-2026-002The Shock

The ShockThe activation condition

The suspension notice, when it comes, will be three paragraphs long. Picture it: a $14bn perpetual credit fund tells its investors, on a Tuesday in 2027, that withdrawals are paused for good, that the board has approved a liquidating trust, and that cash will follow "as asset realisations permit." By Friday, two rival funds have quietly cut the value of the loans they share with it. A rating agency has pulled four private letter ratings. The queue of withdrawal requests across the wrapper complex has doubled. None of this is invented except the date: every step already has a 2026 precedent on the tape.9

Activation condition — what must be true (all three, inside one rolling six-month window)

1 · A marquee failure: a US direct-lending or perpetual credit vehicle of $10bn or more suspends redemptions outright, converts to a liquidating trust, or defaults on fund-level leverage.
2 · Defaults confirm: the Fitch US private credit default rate reaches ≥8% trailing-twelve-months, with payment defaults and bankruptcies at ≥25% of its events (hard defaults ≥2% — roughly twice today's rate, so forbearance accounting alone cannot fire this leg; the index high is 6.0%, Q2 2026).
3 · The price confirms: US high-yield spreads at ≥550bp — a doubling from August 2026's 271–281bp — with the listed-BDC median price at ≤0.65× reported NAV (net asset value, the fund's own appraisal of a share; 0.80× now).
Any two legs is a watch state: we watch, but the scenario is not active.6

Rewind to now: none of that has happened yet. And the definition matters more than the story. One fund failure on its own is a one-off. A default rate on its own can be definitional noise. A spread move on its own is a macro scare. We require all three together, because together they confirm the thing this dossier actually studies — the moment the funding structure fails, not just the loans. Each leg can be checked from public data (Fitch's index, FRED's spread series, BDC filings), so the scenario is a regime the market either enters or does not. This is a conditional analysis of that regime, not a forecast that it arrives.

~$1.6trn
US private credit, end-2024 (OFR) — the machine under test5
7 of 12
largest wrappers that gated redemptions in Q1 20269
6.0%
Fitch default rate TTM, Q2 2026 — the young index's high6
~1 in 4–5
18-month windows containing a nonbank credit unwind, 1986–2026 (6–8 event counts)34
Probability (magnitude ≠ likelihood)
15–25% / 18m · conf. medium-low
Loose base ~20–26%; strict three-leg retro-test ~7–9%; market-implied ~5–10%. The band is a variant view, ~2–3× the tape — argued on p28. Movers: the Fed path · the fundraising strike · the redemption queue · AI-datacentre credit.34
Gravitywell Research03 / 33
Gravitywell Research · GWW-2026-002Executive scenario summary

Executive summaryThe scenario in one page

We put the odds of activation at 15–25% over 18 months. That is our stated view, roughly 2–3 times what the market implies. The damage, if it comes, is a funding winter, not a crash: listed vehicles down 40–55%, private marks down 10–20% over two to four quarters, new commitments from LPs (the investors in funds) frozen for a year or more.9 The Indian leg is where the mispricing sits. India's tape is calm (financials mid-range, volatility in its cheapest decile, spreads at cycle tights), but no Indian instrument can price this transmission at all. Meanwhile the channel that carries it is visibly under strain: at Apollo's flagship wrapper, requests were met in full at 1.38–1.83% of shares through 2025; in both 2026 quarters the 5% cap was hit. A press-reported offshore/onshore split points down the same channel, but it does not square with the filing — we carry it as an indication, not evidence (page 27).16

FIG E.1 · The two axes — never collapse themprobability | magnitude
PROBABILITY OF ACTIVATION / 18m tape ~5–10% ○ ≈ strict base base ~23% P ≈ 15–25% 0%25%50% a variant view: ~2–3× the tape, below the loose base — argued, not asserted (p28) MAGNITUDE GIVEN ACTIVATION · listed wrappers, peak-to-trough Mild · w 40%Moderate · w 40%Severe · w 20% −25 to −35% −40 to −55% −60%+ · the regulated core is hit weights sum to 100% GIVEN activation
Left: probability ruler — band [15–25%] only; pins at the loose base ~23% ● and the tape/strict cluster ~5–10% ○. Right: conditional wrapper drawdowns by branch, Mar-2020 calibration ◐.34
What the market already prices

The listed wrappers are not just down; they are priced for pain. A 0.74–0.80× median price-to-NAV mathematically embeds roughly 25% cumulative defaults at standard recoveries, tender offers have cleared 20–35% under stated NAV, and the sector lagged the S&P by ~20 points over eighteen months. The US loss leg is deeply priced.31

The read — what it misses

The screen cannot show a commitment that is never made. If the wrappers keep gating, the offshore institutions stuck behind those gates stop writing new commitments everywhere — and India's credit boom is built on exactly that money meeting exactly the wrong book: more than a third of Indian private-credit deployment is refinancing, which fails at maturity dates, quarters after the US event, when the marks still look fine.22

The one trade

Own the convex positions on the Indian leg — small fixed cost, many times that cost back if the scenario lands — while they are cheap: puts on Indian financials with volatility in its cheapest decile, paired with received fixed rates — every rung of the RBI's response playbook is bond-bullish. Skip the crowded expression (shorting listed BDCs is a 2025 trade at 0.80× NAV). With activation odds near their own base rate and premiums priced for none of it, the asymmetry does the work.37

Gravitywell Research04 / 33
Gravitywell Research · GWW-2026-002Scenario Quick Read

Quick readThe cascade in sixty seconds

FIG Q.1 · Gate → strike → stall → converge → India, in five beatsmonth 0 → month 18
Month 0a wrappersuspends Month 1–3queue doubles ·fundraising stops Month 3–6extensions becomedefaults · marks fall Month 6–12LP freeze reachesIndia rollovers FORK · m12–18 Mild 40% · reopen Mod. 40% · winter Sev. 20% · insurers
The clock is 2015–16's, not 2008's: months, not days — no daily margin exists in the chain until the severe branch ◐. Branch weights are conditional on activation.34
If you are…The exposureNow (cheap)If it activates
An allocator / LPcommitments to 2024–26 credit funds; wrapper positionsfinancials puts + receive-OIS; tender wrapper stakes near NAVbid gated books at 80–85 cents on the dollar with locked capital — the Franklin math favours patient buyers
An Indian GP (fund manager) mid-fundraisethe funding winter arrives at your close dateaccelerate closes; broaden past US-fed offshore LPsexpect 2–4 quarter slips; domestic HNI (high-net-worth) capital reprices your terms
A treasurer / NBFC CFOthe rollover jam at your maturity datesladder CP ≤15–20%/month; pre-fund FY28 at today's 50–140bp AAA spreadsdraw committed lines early — 2018 says the window shuts in weeks
A policymakeran offshore funding stop you cannot see in SEBI datapublish the AIF foreign/domestic split; pre-position the SLF-MF templatedeploy the conduit rung at +100bp of spread, not +200 — 2018's lesson on lateness
15–25%
P(activation) / 18m — a variant view, ~2–3× the market-implied ~5–10% ○34
~4%
unconditional severe tail — what the hedge is priced against34
~1–2%
of hedged notional per year — the three-leg book, all legs in37
2–4 qtrs
India's lag behind the US trigger — the cheap-entry window30
In one line

A US wrapper run that never touches an Indian borrower can still shut India's credit-rollover window a year later — that lag, and the absence of any instrument pricing it, is the whole dossier.

Gravitywell Research05 / 33
Part I

The Setup

The market grew into the one shape that cannot absorb a redemption cycle: locked loans inside redeemable wrappers, held more and more by insurers and retail. And the base rate put roughly 20–26% odds on any 18-month window before this year's tape was even printed.

Part I · The SetupGWW-2026-002

§ 01 · AnatomyThe fragility is the wrapper, not the leverage

In one line · the machine holds illiquid loans in vehicles that promise quarterly liquidity — that mismatch, not fund leverage, is what gets tested

That lag starts with how the machine is built. US private credit reached roughly $1.6trn by end-2024 on the count of the OFR (the US Office of Financial Research), growing ~14% a year for a decade. The loans sit in funds that borrow very little: the median fund runs about 1.0× debt — nothing like the 25–30× of the hedge fund LTCM, or the 2007 conduit chains.5 The weak point moved somewhere else. About $607bn now sits in evergreen funds, mostly private-wealth money admitted at $2,500 minimums. Another ~$550bn sits in ~170 BDCs (business development companies — the fund structure through which US investors, including retail, own portfolios of direct loans); the non-traded ones promise to buy back 5% of NAV each quarter. Five-year loans that cannot be sold, held in funds that promise quarterly exits: every gated fund in history has shared that shape.29

FIG 1.1 · Where the $1.6trn sits — and which shell can run$bn · end-2025/Q1-26
Locked drawdown fundsBDCs (~170 vehicles)Evergreen / interval ~$900bn+ · cannot run — no redemption window ~$550bn · listed + non-traded ~$607bn · 5%/qtr promise the run perimeter: the wrappers + non-traded BDCs shells overlap (non-traded BDCs are also evergreen vehicles); totals bracket the OFR's ~$1.6trn US estimate
Source: OFR Brief 26-02 ●, FSB May 2026 ●, WealthManagement evergreen census ◐. Bars are vehicle populations, not additive to a total.5

The holders changed as fast as the vehicles. Privately placed bonds reached 48.4% of US life insurers' total bonds by end-2025, up from 37.4% five years earlier. On the broadest count, life and annuity insurers hold ~$1.8trn of private credit — a record 46% of their debt holdings. Much of it is graded by private letter ratings from smaller agencies — which the FSB (the Financial Stability Board, the G20's risk watchdog) pairs, in its own words, with "reports of ratings inflation."3 So the same loan now touches a retail redemption promise on one side and an insurer's statutory capital on the other. For an allocator, that is the map of who sells first and who cannot sell at all — and the pages that follow trace the run around exactly that perimeter.

~1.0×
median fund leverage — the core is not the risk5
$607bn
evergreen shell, ~86% private-wealth money29
46%
of life-insurer debt holdings now private — a record20
48.4%
private share of life-insurer bonds, vs 37.4% five years ago20
Gravitywell Research06 / 33
Part I · The SetupGWW-2026-002

§ 02 · Base rateOne window in four, before you look at the tape

In one line · six nonbank credit unwinds in forty years ≈ 20–26% per 18-month window — and none with a private epicenter ever drew a policy put

Has that shape ever actually been tested? Yes — six times in forty years. Credit unwinds that started outside the banks: the Drexel junk-bond bust of 1989–91, LTCM in 1998, the telecom default cycle of 2001–02, the ABCP and CLO chain of 2007–09, the energy-and-Third-Avenue episode of 2015–16, and the COVID seizure of March 2020.34 Six events in forty years is a rate of ~0.15 a year. Run the arithmetic, and about 20% of all 18-month windows contain one — 26% if you also count 1994's mortgage-derivative shock and the loan-fund run of late 2018. So the loose base rate alone sits at, indeed slightly above, the top of our band, while the strict three-leg test and the market-implied read sit well below it. That spread of anchors tells you what this dossier is: the band is argued openly against both (page 28), and the real work below is tracing the channel the odds run through.

FIG 2.1 · The reference class — forty years of nonbank credit unwinds1986 → 2026
19861996200620162026 1990 Drexel/S&Lsystemic 1998 LTCM6 weeks 2002 telecom12.8% dflt · no put 2008 ABCP/CLOsystemic 2016 Third Avenuegate → 7m to normal 2020 COVIDput, same day 2026: thiswindow?
Dot size/colour = resolution: crimson = systemic; amber = contained. 2 of 6 resolved systemic — the naive severe weight the branch table starts from ●.34
The arithmetic, shown

λ = 6/40 = 0.15 events/yr → P(≥1 in 18m) = 1 − e−0.225 ≈ 20.2%; the 8-event count gives 25.9%. In plain terms: if these shocks keep arriving at their historical average rate, any given 18-month stretch has roughly a one-in-five chance of containing one. Two opposing adjustments then apply, and we show both. Strictness cuts it: retro-apply our three-leg test and only ~2 of the 6 events would have activated (~7–9%/18m). The live 2026 markers (§05), two of three legs already part-formed, multiply it back. The 15–25% band nets those honestly wide.34

The no-put record

The "put" is market shorthand for faith that the central bank steps in when prices fall. Of the six, only 1998, 2020 and arguably 2008 drew a rescue, and every rescue attached to instruments a facility could legally buy. No episode with a private, unrated, non-CUSIP epicenter has ever received one. That record leaves the odds of activation untouched; what it fixes is the severe branch's defining constraint in Part IV.35

Gravitywell Research07 / 33
Part I · The SetupGWW-2026-002

§ 03 · The clocksMonths to unwind, one print to reprice

In one line · Third Avenue says the run takes months; March 2020 says the traded wrapper falls ~50% while the marks barely move

If this window is the one, two past episodes set the clock. Third Avenue's Focused Credit Fund ($789m, down 27% on the year) gated on 9 December 2015. High-yield spreads took nine weeks to peak at 887bp, and five more months to settle. Three copycat funds suspended within days of the first, because the gate itself, not the losses, was the news.34 That is the tempo to plan around. Private credit has no daily margin call anywhere in its chain, so the 2022 UK pension (LDI) template, a three-day spiral, is the wrong clock. The right one is 2015: a gate that slowly teaches the market what the assets are worth.

FIG 3.1 · The 2015–16 clock — gate to peak to normal, in monthsHY OAS, bp
900700500300 9 Dec 2015 · the gate 11 Feb 2016 · 887bp peak (+9 wks) Jul 2016 · <500bp (~7m) Dec-15Jan-16Apr-16Jul-16
Source: ICE BofA HY OAS via FRED ● (public history truncated; path reconstructed from BIS QR Mar-2016 ◐). The week after the gate: $13bn of bond-fund outflows, the largest since 1992.34

March 2020 supplies the second hand on the dial: how far the traded shell falls while the marks pretend. In five sessions to 20 March 2020, the listed-BDC complex dropped 28%. By 27 March the sector traded at an average 39% discount to NAV, against 6.6% a month earlier. The wrapper repriced roughly half its value while the appraisals underneath moved by single digits.35 If you hold a wrapper position, that is your loss profile: the mark is not your exit price, and the gap between them closes against you exactly when you want out. What is different this time cuts both ways — today's wrappers pre-cap redemptions, which slows the run; and today's Fed put has no tool for private loans, which removes 2020's same-day rescue.

9 wks
Third Avenue gate → spread peak — the onset tempo34
~7m
gate → spreads back under 500bp — the recovery tempo34
−39%
avg listed-BDC discount to NAV, 27 Mar 2020 — the wrapper gap35
3 days
2022 LDI spiral tempo — the WRONG clock: it needed daily margin35
Gravitywell Research08 / 33
Part I · The SetupGWW-2026-002

§ 04 · The calibrationsWhat a no-put cycle, a funding stop and a gated fund actually cost

In one line · 2002 sizes the no-put default cycle; IL&FS sizes an Indian funding stop; Franklin proves a vehicle can fail while its assets pay 109%

Three more episodes size the damage rather than the clock. The 2001–02 telecom cycle is the honest no-rescue benchmark: 12.8% of high-yield defaulted in one year ($96.9bn across 344 issues), no facility came, and the system absorbed it in nine to twelve months of wide spreads and a shut primary market.34 Defaults alone, even at triple today's level, are survivable. What 2002 did not have was a layer of redeemable wrappers standing between the defaults and the household sector. That difference is why the wrapper pages of Part II, not the default arithmetic, carry this scenario's weight.

EpisodeTriggerPeak damagePolicyTime to normalWhat it calibrates here
2001–02 telecomcapex bust + fraud12.8% HY default rate · ~1,010bp spreadsnone9–12ma no-put system absorbs a full default cycle — slowly
IL&FS · Sep 2018AAA quasi-sovereign defaults on CPNBFC funding +100–110bp · ₹2.4trn of mutual-fund (MF) debt outflows in one monthmarket liquidity (OMOs), then guarantees — late>18mthe India moderate branch: one event reprices a whole funding channel
Franklin · Apr 2020redemption surge into an illiquid booksix schemes gated, ₹25,215cr trappedSLF-MF ₹50,000cr in 4 days~3yrs to full payoutassets returned 109.25% — the vehicle failed anyway; time was the loss
FIG 4.1 · IL&FS, the India calibration — one default, one month, one channelSep–Oct 2018
MF industry AUM, 1 month MF liquid+income outflows, Sept NBFC funding cost, H1 FY19 −₹3.16trn · −12.5% −₹2.4trn · largest since ≥2008 +100–110bp bars scaled within-row (separate units, separate scales); MF exposure to NBFCs was still −30% fifteen months on
Source: AMFI/Business Standard archive ●, Moody's 2018 ◐. DSP's ₹300cr sale of AAA DHFL paper at an 11% yield was the single print that repriced the channel.32

Franklin Templeton's six gated schemes are the closest thing to a controlled experiment on what a frozen credit vehicle is worth. Every scheme eventually paid out more than its gating-day value — ₹27,548cr against ₹25,215cr, or 109.25%. The franchise still spent four years out of Indian debt.32 Hold both facts together, because the scenario needs both: final recoveries can be excellent while the freeze itself does the damage — to a GP mid-fundraise, to a promoter at a maturity date, to an LP who needed this quarter's cash. Time, not credit loss, is how this scenario transmits.

Gravitywell Research09 / 33
Part I · The SetupGWW-2026-002

§ 05 · Why nowRecord defaults, richest-decile spreads — both cannot stay true

In one line · every stress marker is live at once while the price sits at cycle tights — the divergence is the setup, and it resolves one way or the other

Against those calibrations, here is the 2026 tape. Fitch's private credit default rate printed 6.0% in Q2 — the high of an index only launched in 2024. PIK (interest paid in more debt rather than cash) has climbed from 6% to ~10% of BDC loans in three years, rising in nearly every industry at once. Redemption requests at the twelve largest wrappers averaged 12.1% of NAV against 5% gates. BDC fundraising is down 55% on the year. And Q1 2026 was the first net-outflow quarter of the perpetual-wrapper era — a five-year-old era, flipped by collapsing sales against a capped numerator: a strike, not a stampede.1 Now set that row of facts against the price of credit risk. High-yield spreads at 271–281bp sit in the richest decile of their history, and direct-lending spreads have compressed a full point in two years — into the rising PIK. The Boston Fed calls the pair "a puzzle." We call it the setup: either the default indices overstate the cycle, or the price is wrong. The gap between those two answers is roughly this dossier's probability band.

FIG 5.1 · The divergence — index-high defaults, richest-decile spreads2022 → Aug 2026
8%5%2% Fitch dflt 6.0% · index high HY OAS 271bp · richest decile 20222024Aug-26 the scissors: risk up, price of risk down
Fitch ● (left, %); HY OAS via FRED ● (right, bp, dashed — separate axes). Amber-line caveats: the index is young (2024-) and PIK entry is its largest event type — PIK and defaults are one signal, not two. Proskauer reads 2.51% falling (p13); the standing anomaly (p27) is spreads-vs-defaults.6
FIG 5.2 · The queue against the gate — redemption requests, largest wrappersQ1-26 % NAV · ADS/BCRED Q2-26 % shares
015%30% 40.7%OTIC 21.9%OCIC 16.8%ADS Q2 11.6%Ares 11.3%Barings ~10%BCRED Q2 8.5%Oaktree the 5% gate — everything above it queues
Source: With Intelligence Q1-26 (% of NAV); ADS/BCRED Q2-26 (% of shares) ◐. Twelve-fund Q1 average 12.1%; 53.4% of requested dollars honoured — the queue mechanics of §07.9
Gravitywell Research10 / 33
Part II

The Transmission

The market prices the wrapper losses in hours. What it cannot see comes in sequence: a fundraising strike that starves the extension machine, marks that then fall into line with traded prices, and an offshore LP balance sheet that carries the freeze to India two to four quarters later.

Part II · The TransmissionGWW-2026-002

§ 06 · The chainFrom a gate in New York to a maturity date in Mumbai

In one line · the first order is priced, the second is half-priced, and everything that touches India is unpriced — own the unpriced rows

The divergence resolves through a chain, and the chain has an order. First come the wrapper run and the fundraising strike — visible, partly priced. Then the effects that need a quarter or two to surface: the extension machine stalls for lack of fresh capital, the marks converge once a liquidating vehicle prints a real price, insurers face their first statutory test, and the AI-datacentre pipeline loses its lender. Then the third order, where India lives: offshore LPs, gated in the US, stop committing everywhere — and a refinancing-heavy Indian credit book finds its bid has gone home.16

FIG 6.1 · The transmission chain — gate to rollover jam1st → 3rd order
1ST ORDER · priced2ND ORDER · half-priced3RD ORDER · unpriced THE WRAPPER RUN 12.1% asks vs 5% gates PRIMARY STRIKE fundraising −55% YoY Extension machine stallsextensions → defaults · 6→8–10% Marks converge to pricesNAV −10/−20% · 2–4 qtrs Insurers + AI-DC stackstatutory tests · $800bn pipeline Offshore LP freezegated abroad → commits nothing INDIA: the rollover jam>35% refi book meets no bid · NBFC +60–110bp no fresh capitala real price printsgated moneyshared lendersthe freeze lands
Source: Gravitywell scenario model ○ on the sourced nodes of §§07–10. Arrows are mechanisms, not correlations; each is argued on its page.3
NodeMechanismMagnitude if activatedAlready priced?
Wrapper lossesdiscounts + tenders clear the gap the marks denylisted −40/−55% (moderate)priced-ish · 0.80× NAV start
Extension machine stallsextensions need fresh capital behind them; the strike removes itdefaults 6.0 → 8–10%half · consensus reads 2.51% and calm
Marks convergea liquidating trust prints; appraisal alpha (2.74%/yr) unwindsNAVs −10/−20% over 2–4 qtrsunpriced
Offshore LP freezegated investors stop committing globally (indication-tier evidence — §10)new EM credit commitments −40/−60%unpriced
India rollover jamrefi-heavy book, offshore-priced bid, maturity-date arrivalIndia PC deployment −50% · NBFC +60–110bpunpriced · VIX 11.7, spreads at tights
Gravitywell Research11 / 33
Part II · The TransmissionGWW-2026-002

§ 07 · The wrapper runA run in slow motion is still a run

In one line · the 5% gate converts a redemption panic into a lengthening queue — and the queue itself is what stops new money arriving

Start at the chain's first box, because you can watch its mechanics today. A gated wrapper cannot suffer a bank-style run; that is the design. But the design creates a ratchet. When NAV falls, the same 5% cap pays out fewer dollars. The unmet queue lengthens. More holders join it before the next window. The fund sells its most liquid assets to pay what it can — which concentrates the illiquid remainder on whoever stays. Woodford — the UK equity manager whose flagship fund froze — ran exactly this sequence in 2019, and the NBER's 2026 modelling of semi-liquid credit funds finds the cash buffers cannot fund repeated 5% quarters.33 One honesty note about the data: requests are a stock, not a flow. Unmet demand must re-file each quarter, so headline request rates carry last quarter's queue inside them. The genuinely clean flow number is gross sales; the honoured fraction (the share of requested dollars the funds actually paid out) tracks the same contaminated stock but shows how binding the gates are. Both anchor the dashboard, each labelled for what it is. Q1 2026 put live numbers on all of this: ~$15bn requested across the twelve largest wrappers, 53.4% honoured, seven funds at their caps.9

FIG 7.1 · The strike — capital formation while the queue builtBDC fundraising
0$15bn$30bn ~$27bnJan–May 2025 $11.9bn · −55%Jan–May 2026 −74% YoYworst month 2026 Q1 2026: first net-outflow quarter in BDC history
Source: Stanger/AltsWire ◐ — a single-vendor series, uncorroborated. Jan–May 2025 level derived from the −55% YoY print. Non-traded gross sales fell 59% YoY in Q1.11

The strike matters more than the queue, and the reason is what the fresh money used to do. New subscriptions are how an evergreen fund pays redemptions without selling. New fund closes are how the industry refinances its own borrowers at maturity. Cut both (fundraising down 55%, the first net-outflow quarter on record) and the system loses its internal refinancing bid just as the 2028 maturity wall comes into view.11 The exits that do clear show you the price of the door: Saba's tender — an open offer to buy other holders out — for a Blue Owl non-traded fund filled at a 34.9% discount to NAV. If you hold wrapper paper, that tender, not the mark, is the bid to assume in stress.

The read

Watch the honoured-fraction against gross sales, not the request rate alone: when the twelve-fund average of honoured dollars drops below half of requests for a second consecutive quarter, the ratchet is self-sustaining and the activation clock starts — that is the wrapper-side trigger in the Part V dashboard.

Gravitywell Research12 / 33
Part II · The TransmissionGWW-2026-002

§ 08 · The extension machineDefaults deferred are defaults financed — until the financing stops

In one line · more than half of Q2's "defaults" were maturity extensions; remove the fresh capital that funds extending, and the 6.0% becomes 8–10%

The strike's first casualty is the machine that has kept the default rate looking manageable. Over half the default events in Fitch's Q2 were maturity extensions, not missed payments. PIK now runs at ~10% of BDC loans and rose at seven of the nine largest non-traded wrappers last quarter — up 42% year-on-year in dollars.6 Extending and PIK-ing are not tricks. They are financing decisions, and they need patient LPs and fresh money behind them. That is why the Fitch–Proskauer gap matters: Fitch's 6.0% counts the extensions, Proskauer's 2.51% largely does not — so which index you believe is really a view on whether the machine keeps running. Neither is a market measure (Proskauer's panel is loans its own lawyers papered; Fitch's is young and half model-implied), so we use the gap as a lesson in definitions, never as a price. Starve the machine, and the gap closes upward.

FIG 8.1 · What a "default" is in 2026 — the machine at workQ2 2026 events
maturity extensions · >half of events payment defaults + distressed exchanges 32 default events · 20 new defaulters · Fitch Q2 2026 each extension consumes lender capacity: dry powder, PIK tolerance, LP patience the strike removes all three → the amber segment migrates right FSB's parallel accounting: outright defaults ~1%; ~5% including selective defaults; higher on distressed-exchange measures
Source: Fitch via Bloomberg 30 Jul 2026 ●, FSB May 2026 ●. Proskauer's 2.51% (falling; sponsor-backed = PE-owned borrowers only) is the same market counted without most of the amber bar.6

Then the marks. Today one loan carries several values at once. The manager's appraisal sits near par (the academic estimate of the appraisal premium is 2.74% a year of phantom alpha). A credit-secondaries bid runs 92–100 cents on the headline, about 95 effective once portfolio discounts and reference-date cash flows are netted. The listed-BDC price implies a 14–20% discount. And where someone actually forced the door, the tender printed at Saba's 34.9% discount.31 A liquidating trust collapses that stack into one printed number, the way Third Avenue's did in 2015 — and every fund holding the same names must then explain the difference to its auditor. Cross-manager gaps on identical loans already exceed five points.31

FIG 8.2 · One loan, four prices — the stack a real print collapsescents on the NAV dollar
60708090100 the mark ~100 secondaries ~95 effective BDC-implied 80–86 Saba tender 65 a liquidating-trust print forces every holder of the shared name toward the left — that is N4, the marks converging
Source: Jefferies/PitchBook secondaries ◐, Mercer BDC discounts ◐, Bloomberg (Saba/Blue Owl tender) ●, Suhonen 2024 (appraisal alpha) ●.30
Gravitywell Research13 / 33
Part II · The TransmissionGWW-2026-002

§ 09 · The regulated holdersWhere the loss would stop being a fund problem

In one line · converged marks meet two amplifiers — insurer statutory capital graded by private letter ratings, and an $800bn AI-datacentre pipeline that assumes the lender never leaves

Converged marks must land on someone's balance sheet, and the two biggest landing zones are what separate the moderate branch from the severe one. The first is insurance. Private paper is now 48.4% of life insurers' bonds. PE-backed insurers control ~$900bn of US insurance liabilities. And much of the collateral is graded by private letter ratings from smaller agencies — a phrase the FSB pairs, in the same paragraph, with "reports of ratings inflation."3 If marks fall far enough to trigger downgrades, the problem stops belonging to fund investors and becomes a statutory capital one — statutory capital being the minimum cushion regulators require an insurer to hold — with forced-selling rules attached. That crossing, the loss moving from voluntary holders to regulated ones, is this dossier's definition of severe. A different regime, not a bigger number.

FIG 9.1 · The quiet migration — private paper into regulated balance sheets% of US life-insurer bonds
30%40%50% 37.4% · 2020 48.4% · end-2025 202020232025 privately placed share of life-industry bonds — the statutory-capital exposure the marks now sit under
Source: S&P Global Market Intelligence 2026 insurance report ●. Broader count: ~$1.8trn of private credit = a record 46% of life & annuity debt holdings.20

The second landing zone is the AI-datacentre stack, and it links this dossier to its sister. Private credit to AI-related borrowers went from ~$3bn originated in 2010 to $40bn+ in 2025, with $200bn+ now outstanding. The FSB counts ~$800bn of private credit inside the $1.5trn of outside capital the 2025–28 buildout needs, structured through off-balance-sheet vehicles like Meta's ~$30bn "Beignet" SPV with Blue Owl.13 Our GWW-2026-001 dossier priced the AI shock and put the credit leg in its severe branch only. The Q1 gates have already overtaken that placement, and this dossier corrects it: here the credit leg is the scenario, and the two shocks share one lender set. A funding stop strands the pipeline even if AI demand holds; an AI disappointment defaults the stock even if funding holds. Either fires the other's signposts — which is why both dashboards carry the same DC-credit row.40

$200bn+
AI-linked private credit outstanding, from ~$3bn/yr originated in 201013
~$800bn
private-credit share of the 2025–28 DC capex stack (FSB)3
34%
of 2025 private-credit deals were AI-related, vs a 17% five-year norm3
~$900bn
US insurance liabilities controlled by PE-backed insurers3
Gravitywell Research14 / 33
Part II · The TransmissionGWW-2026-002

§ 10 · The India channelThe freeze travels on the LP balance sheet, and lands on the refi calendar

In one line · offshore money redeems first in the US and commits ~71% of GIFT City's new capital — the same balance sheet, both ends of the pipe

Now the third order, where the two lender sets become one balance sheet — our thinnest evidence, and we say so. Press reports put offshore redemption requests at Apollo's flagship wrapper near three times the onshore rate — but the split appears nowhere in the fund's filing and does not square with its totals, so we carry it as an indication only (page 27).16 What the filing does verify is the strain itself: requests met in full at 1.38–1.83% of shares through 2025, then the 5% cap hit in both 2026 quarters. The pool behind those wrappers overlaps (by how much, no published series can say) with the pool funding India's credit boom (no AIF split is published; page 30). The pipe itself can be sized: GIFT City took over $5bn of foreign commitments against ~$2bn domestic in FY26 (71% foreign), and the RBI freed up ECB borrowing in February — widening the offshore pipe just as its source began gating.27

FIG 10.1 · One balance sheet, two ends — the LP pipe into Indiathe channel
OFFSHORE LP POOL gated in US wrappers caps binding, queues building GIFT City funnel$5bn FY26 · 71% foreign ECB borrowing routeFeb-26: caps off, NBFCs in INDIA CREDIT BOOK $12.4bn/yr deployed · >35% is refi fails at maturity dates, not today commits (until gated) funds the refi bid the freeze needs no India view at all — liquidity is recalled, not reallocated
Source: Apollo ADS 10-Q ● (the cap series), IFSCA Mar-2026 ● (GIFT), RBI Feb-2026 ECB amendments ●, EY/S&P ◐ (the book).16

What the pipe feeds is a refinancing calendar. More than a third of India's $12.4bn-a-year deployment is refinancing and acquisition finance. The marquee deal, the Shapoorji Pallonji (SP) Group's $3.4bn at 19.75% (zero-coupon: all interest due at maturity), has already needed a covenant waiver. And the national arithmetic carries its own caveat: short-term external debt on residual maturity is 47.3% of reserves and rising, though only 21.6% on original maturity, and the bucket is heavy with sticky NRI deposits and self-liquidating trade credit — which is why it limits the severe branch, not the base case.26 None of this defaults on the day a US wrapper suspends; it fails quietly at maturity dates, quarters later. Three facts currently run the other way: GIFT commitments grew 148% year-on-year through the first gate quarter, FPI (foreign-investor) debt flows are positive three months running, and Kotak's record ₹3,900cr first close was all-domestic. The freeze would show in the GIFT prints only from September 2026 — those prints decide: two falling quarters supports the thesis, continued growth refutes it. The widened ECB pipe may also reprice rather than jam — the mild branch. The impact map prices the lag.

1.8% → 5%
ADS: requests met in full 2025 → cap binding both 2026 qtrs ●16
71%
foreign share of GIFT City's FY26 fund commitments27
>35%
of India private-credit deployment that is refinancing (EY, single-source) — the jam point22
47.3%
short-term external debt / reserves, residual basis (21.6% original) — up from 45.4%18
Gravitywell Research15 / 33
Part III

The Impact Map

The damage is not where the tape looks. The false safe-haven is the "stable" private mark itself. The surprise winners are the domestic Indian funds that buy the offshore book's maturities at a spread. And the deepest losses arrive by calendar, not by contagion.

Part III · The Impact MapGWW-2026-002

§ 11 · By assetThe stable mark is the false safe-haven

In one line · everything traded falls fast and partially recovers; everything appraised falls late and quietly — hold the distinction, it is the whole map

Pricing that lag across assets starts with one split: traded exposures gap at once, and hard; appraised ones move a quarter or two later, by committee. If you take one warning from this map, take this: the "stability" of the second group is its most dangerous line. A private mark in month one of this scenario is not a price. It is a pending negotiation. The heatmap below prices the moderate branch; severe mostly deepens the crimson, except where it flips the sign (duration wins bigger, the rupee loses its defence).9

FIG 11.1 · Impact by asset × horizon — moderate branchscenario-conditional
0–3m3–9m9–18m US listed BDCs / wrappersUS HY spreads (OAS)Private credit marksCredit secondaries bidsIndia: G-sec / receive-OISUSDINR (rupee leg)Gold in INR −30/−45% −40/−55% trough partial recovery 550–700bp 700–1,000bp <500bp (2016 path) −0/−5% · "stable" −10/−20% the late leg lands 95 → 85 80–85 · gated books buyers' market 10Y → 6.5% 6.25–6.50% every rung bond-bullish 95.4 → 96–97 97–99 RBI holds (mod.) record entry ₹152.5k diversifier, not a hedge worst convexity at a record
Gravitywell scenario model ○ on Part I calibrations (2016 spread path ●, Mar-2020 wrapper gap ●, IL&FS India moves ●). Ramp: deep crimson −40%+ · crimson −10/−40% · pink −0/−10% · pale-teal flat/mixed · teal gain. Values printed per cell; severe deepens crimson except duration and the rupee, where it changes the sign.34
The false safe-haven

"Our NAV is stable" will be the most-heard sentence of month one — and the least informative. March 2020's wrappers proved stability of the mark and a 39% traded discount can coexist; the stable line in row three is a queue, not a floor. If you can exit near the mark early, that optionality is worth more than the yield you give up.35

The clean long

Indian duration is the one unambiguous beneficiary across all three branches short of a full FX crisis: the RBI's every plausible response (liquidity, conduits, backstops) pushes yields down, and the carry cost of waiting is minimal with the repo at 5.25% and a neutral stance. Receive-OIS is the anchor leg of the Part V book.37

Gravitywell Research16 / 33
Part III · The Impact MapGWW-2026-002

§ 12 · By sectorSoftware is the epicenter; Indian real estate is the echo

In one line · a fifth of US direct lending sits in software with shared borrowers and no spread differentiation — and India's private-credit book peaked into real estate the same way

Within those asset rows, two sector concentrations decide who defaults first. In the US, it is software: ~$115bn of BDC lending, about a fifth of the total. Sixty percent of software borrowers owe money to seven or more BDCs at once, up from under 10% a decade ago. The five largest BDCs hold 37% of all software loans. And new-issue spreads compressed from ~800bp to ~500bp with, on the measurement of the BIS (the Basel-based bank for central banks), no difference between software and everything else.2 The AI-disruption story gets the headlines, but the structure is the risk: with a shared borrower pool, one honest markdown travels to every lender at once. The same mechanics made 2002's telecom cycle systemic for its financiers. Less than 1% of these loans are behind on payments — and the BIS reads that as PIK delaying the visible strain, not as comfort.2

FIG 12.1 · The software pool — shared borrowers, compressed spreadsBDC lending
Software share of BDC loans Borrowers lending from 7+ BDCs New-issue software spread ~20% · ~$115bn ~60% · was <10% in 2015 ~500bp · from ~800bp in 2017 rows carry separate units (share · share · bp) on separate scales; the middle row is the contagion coefficient — one markdown, every lender
Source: BIS Bulletin 128, 14 Jul 2026 ●. Feb-2026 already previewed it: a software-loan selloff on AI fears crossed from syndicated to private in days (OFR).2

India's echo is real estate. At the boom's hottest, real estate took the largest share of private-credit money: roughly 42% of first-half 2025 volume, priced in the EY survey's own words at the peak of the upcycle, with healthcare and industrials behind it.22 Those are exactly the deals whose maturities meet the frozen pipe of §10. And the borrower who defines the archetype has already flinched: the SP Group's financing arm won lender consent in April to raise its loan-to-value covenant from 34% to 40% after collateral values fell. For the rest (quick-commerce boards, promoter holdcos, mid-tier developers) the practical translation is simple: 2025's 19.75% zero-coupon refinancing becomes a mid-20s conversation, or no conversation.26

~$115bn
BDC software lending — the US epicenter2
~42%
real-estate share of India's H1-25 private-credit volume — "peak of the upcycle"22
34→40%
the SP Group's LTV covenant, waived upward Apr 2026 — the echo's first data point26
19.75%
the 2025 refinancing yield that must itself be refinanced by 202826
Gravitywell Research17 / 33
Part III · The Impact MapGWW-2026-002

§ 13 · By geography & stakeholderAmerica takes the loss; India takes the drought

In one line · the US hit is a credit loss with a recession attached; the India hit is a liquidity drought with fundamentals intact — different damage, different defence

Geographically, the scenario splits into a loss and a drought. The American loss is concrete. Three quarters of private-credit borrowers earn under $100m of EBITDA. A quarter of the middle market already covers interest below 1.0× on the rating agency KBRA's measure. Q2 2026 logged the largest quarterly EBITDA-growth decline on record. A credit crunch for these firms is a real-economy event — worth a mild US recession in the moderate branch.8 India's drought is quieter, because at first nothing defaults: bank NPAs sit at a multi-decade low of 1.8%, NBFC capital passes the RBI's own severe stress test at 20.9%, and the borrowers are mostly fine. What stops is money — new commitments, refinancing bids, exit windows. Droughts kill slowly and selectively. That is why the table below sorts stakeholders by their exposure to time, not to credit.17

FIG 13.1 · The US mid-market's thinning cushioninterest coverage
015%30% 25%ICR <1.0× (KBRA MM) ~10%<1.0× (BDC books) ~20%*construction loans on PIK 1.6×median ICR (falling) *share of construction-sector BDC loans paying PIK, from <5% in 2022 — the sector tell; final bar is a ratio, not a share
Source: KBRA Q2-26 Compendium ●, Boston Fed CPP 26-6 ●, With Intelligence ◐. S&P's harsher cut, via the FSB: 28–41% of middle-market CLO borrowers can't cover interest once tax, working capital and capex count.3
StakeholderWhat actually hits themWhenMagnitude (moderate)
US retail in the wrapperstrapped behind gates; exit via tenders at 65–85 centsnow → 12m−15/−35% realised on exit
US mid-market firmsrefinancing at +300bp or not at all; the crunch3–12mdefaults 8–10%; mild recession
Indian GPs mid-fundraisethe funding winter at their close dates2–4 qtrs laggedcloses slip; ex-mega fundraising −40%+
Indian promoters / developersthe rollover jam at covenant and maturity datesat the calendarrefi +300–500bp or forced asset sales
Indian savers (debt MFs)gates at the margin; swing pricing exists but full swing needs a SEBI dislocation declaration never yet made, and reaches debt MFs, not AIFsevent-drivenbetter-armed than 2020 — armour untested
Indian banksambivalent, not netted: they absorb NBFC funding share (43→45%) and gain loan pricing power — while inheriting concentration risk the FSR already flags6–18mwins the flow, carries the tail — your exposure decides which dominates
Gravitywell Research18 / 33
Part III · The Impact MapGWW-2026-002

§ 14 · Private capitalThe denominator effect without the crash

In one line · gated money is not lost money, but it commits to nothing — and LP behaviour moves two to four quarters before any mark admits why

For the fund-and-LP system, the transmission runs in a strict order, and the marks move last. The classic denominator effect needs a public-market crash to push private allocations over their ceilings. This scenario gets the same freeze without the crash: an LP whose US credit sleeve is gated holds an asset it cannot exit, so the only lever left is the pace of new commitments — and that is cut first, quietly, everywhere. The survey data already shows the lever moving. LPs planning to grow private-credit allocations fell from 42% to 29% in the six months to April, and 54% expect more zombie funds — the same series the disconfirming view on page 27 reads as the freeze already executing in an orderly queue. One datum, two readings, unreconciled.29 India feels all this as a 12–18 month delay between cause and visible effect: US gates this quarter, Indian fund closings slipping two quarters later, Indian marks moving two quarters after that.

FIG 14.1 · The lag structure — behaviour first, marks lastt = activation
tt + 1 qtrt + 2 qtrst + 3 qtrst + 4 qtrs US gates bind commitment pace cutinvisible in any NAV India closes slip rollover bids thinthe jam arrives India marks move the whole India window for cheap positioning is the stretch where behaviour has moved and marks have not
Gravitywell scenario model ○, lag lengths from the 2022–23 markup-to-markdown evidence ◐ and the IL&FS 15-month funding-normalisation record ●.30

The drought meets an exit channel that was already narrowing. Indian IPO exits fell 47% year-on-year in the first half, to $801m across twelve listings, inside a 29% fall in total exit value. So the distributions LPs would recycle into new funds are thinning from both ends at once.23 The secondaries valve exists, but it reprices fast: a record $121bn traded globally in H1, with single-asset continuation vehicles at a 2.9% discount — trophy pricing that holds only while buyers believe the marks. India's own benchmark, ChrysCapital's $700m NSE continuation vehicle, proves the plumbing works. In the scenario, the same plumbing clears 10–20 points lower — and venture tails already clear 20–35 points under water today.30

The read

If you allocate to Indian private funds, watch the US wrappers' honoured-fraction and the Coller intent series rather than any Indian datum. By the time an Indian GP's close date slips, that information is two quarters old and the cheap window is closing.

Gravitywell Research19 / 33
Part III · The Impact MapGWW-2026-002

§ 15 · The India bookA $25bn book, one deal deep, and the buyers who inherit it

In one line · India's private-credit stock is small, concentrated and refi-driven — which is precisely what makes it repriceable rather than unsalvageable

The Indian book itself is the last node, and its small size is its saving grace. Roughly $25bn of private-credit AUM (Moody's July-2026 estimate — SEBI publishes no official line) stands against an NBFC system fifty times larger. Deployment has already slowed to $2.2bn in H1 2026 on the narrower EY-IVCA count. And the concentration is stark: one borrower group, SP, took about a quarter of 2025's record year, and deals above $100m were 9% of the count but 36% of the money.22 Concentrated, refinancing-driven and offshore-priced is a fragile shape for the holders. For the buyers, it is a target-rich one. Domestic funds already took 64% of deal value in H2 2025, domestic AIF commitments compound at 21% a year, and Kotak and EAAA are raising up to $3.5bn between them into exactly this gap.22

FIG 15.1 · The book that must roll — concentration and decelerationIndia private credit
0$6bn$12bn $12.4bnCY2025 amber = the SP deal alone (~25%) $2.2bnH1 2026* The inheritors, already raising: Kotak PC fund III · up to $2.0bnEAAA special opportunities · up to $1.5bndomestic AIF commitments · ₹15.74 lakh cr, +21% YoYdomestic share of H2-25 deal value · 64% *narrower EY-IVCA definition — not comparable; the sign itself may be an artefact of the break
Source: EY Feb-2026 ●, EY-IVCA Jul-2026 ●, Bloomberg (Kotak) ●, SEBI AIF data ●. The offshore lenders it must replace: Ares, Cerberus, Davidson Kempner, Farallon — the SP lender list.22
Casualties · in order of arrival
gatedOffshore-fed GIFT vehicles71% foreign commitments; the funnel narrows first
+300–500bpPromoter/holdco refinancingsthe SP template: waiver, then re-price, then asset sale
−40%+Ex-mega India fundraisingthe funding winter at the close date
−10/−20%Vintage 2024–25 credit markspriced at the peak, marked in the drought
Beneficiaries · the other side
+shareDomestic AIFs / HNI credit fundsbuy the same maturities 200–400bp wider — the non-obvious winner
best since '19Special-situations fundsforced sellers, thin competition, real collateral
+pricingIndian banksreclaim NBFC funding share at better spreads (the §13 ambivalence, unresolved)
cheap entryPatient secondaries buyersthe Franklin math: gated books repaid 109% to those who waited
Gravitywell Research20 / 33
Part IV

The Dynamics

Two self-feeding loops push the cascade forward; two dampers pull it back. Their balance sets the branch weights at 40/40/20 — and severe is a different regime altogether: the loss crossing into balance sheets that are forced, by statute, to act.

Part IV · The DynamicsGWW-2026-002

§ 16 · ReflexivityTwo loops forward, two dampers back

In one line · the queue ratchet and the marks spiral are self-feeding; the locked core and the policy bench are self-limiting — the branch you land in is their arm-wrestle

Whether those winners get their prices depends on two loops. The first is the queue ratchet from §07. Each NAV decline shrinks the dollars a 5% gate pays out; the queue lengthens; more redeemers join; the fund sells its liquid sleeve, which worsens what remains — and around again. The second is the marks spiral from §08. Each real print (a liquidating trust, a tender, a daily-valued Apollo fund) drags nearby marks toward it, and every markdown widens the wrapper discounts that started the queue. Both loops run on information, not leverage. That is why they take months, and why no margin call can stop them halfway.33

FIG 16.1 · The queue ratchet — the loop that needs no leveragecausal order, clockwise
NAV falls the 5% gate honoursfewer dollars queue lengthens · more holders join it liquid sleeve sold ·residual book worsens arithmetic first-mover logic Woodford mechanics the marks follow
Gravitywell scenario model ○; the NBER semi-liquid-fund result ● supplies the arithmetic leg, Woodford 2019 ● the liquid-sleeve leg. The marks spiral (§08) runs concentric with this ring.33
The dampers

Against the loops stand two facts. Most of the market cannot run at all — the locked drawdown core (funds whose investors committed money for the fund's whole life and cannot withdraw it) has no redemption window to gate, so the loops' territory is bounded at the wrapper perimeter. And the yield is real: at 10–12% gross on senior-secured paper, every markdown makes the asset cheaper to patient new capital — distressed fundraising was already up ~180% into 2025. The loops run until price meets that bid.5

What decides the branch

The arm-wrestle has a scoreboard: if the honoured-fraction stabilises above ~60% and fundraising turns before defaults reach 8%, the dampers won and the branch is mild. If the queue is still compounding when the first insurer downgrade lands, the loops reached the regulated core, and the severe mechanics take over: a different regime, argued on the next page.

Gravitywell Research21 / 33
Part IV · The DynamicsGWW-2026-002

§ 17 · The branchesMild reopens, moderate freezes, severe crosses the statutory line

In one line · 40/40/20, argued from how six analogues actually resolved — and severe is defined by channels that simply do not exist below its threshold

Given activation, the path forks three ways, and the weights are argued, not asserted. Two of the six reference episodes ended systemic — a naive 33% severe. We cut that to 20%, for three reasons: the core capital is locked, banks' direct exposure is under 0.5% of assets, and both the US and India enter with policy tools 2008 did not have. We cap mild at 40% rather than higher because 2016, the best mild template, had no $607bn wrapper population and no insurance sector at 46% private paper standing in the blast radius. The weights sum to 100% inside the scenario. Multiplied out, the severe tail is ~4% unconditional over 18 months — and that number, not the vivid branch stories, is what the hedge book on page 26 is priced against.34

Branch · weightWhat has to be trueUS peak readingsThe India readout
Mild · 40%
"2016 with gates"
the queue stabilises; extensions keep clearing; no insurer event; yield buyers arrive at the first real printsOAS 550–700bp · defaults 8–9% · listed wrappers −25/−35% · primary shut 2–3 qtrsa two-quarter fundraising pause · NBFC spreads +40–80bp · no policy action · deployment −30% then resumes
Moderate · 40%
"the funding winter"
multiple wind-downs; the extension machine stalls; marks converge; commitment freeze goes global; mild US recessionOAS 700–1,000bp · defaults 10–12% · listed −40/−55% · secondaries 80–85LP commitments −40/−60% for 12–18m · the rollover jam · NBFC +60–110bp (the IL&FS print is the ceiling: its MF-CP pipe is under half its 2018 share) · Bank Nifty −15/−25% (a risk-off de-rate from June's record ~60,000 — the 2018 pattern, not a loss event) · INR 97–99 · RBI rungs 1–3
Severe · 20%
"the statutory crossing"
private-letter downgrades cascade into insurer capital; forced sales meet a no-bid market; NAV-loan and subscription-line losses reach banks; an AI-DC SPV restructures into the same lender set; the put fails for want of a constructOAS >1,200bp · secondaries 60–70 · listed −60%+ · US recession · correlations → 1FPI out of debt AND equity · the short-term external rollover binds (47.3% of reserves residual · 21.6% original; heavy in sticky NRI deposits + trade credit) · INR >100 despite defence · RBI escalates to guarantee + MF-backstop rungs · India PC marks −20/−30%
Why severe is a regime, not a multiple

Below the threshold, every seller is voluntary and every loop runs on information. Past it, sellers are statutory: an insurer breaching risk-based capital does not weigh sentiment, a bank marking a NAV loan does not wait for the next appraisal cycle, and a no-bid secondaries market prices everyone's collateral at once. In severe, the moderate branch's liquidity is simply absent, because that branch's buyers (insurers, banks' fund-finance desks, yield tourists) have become the sellers. That inversion, not any particular number, is the definition.

Gravitywell Research22 / 33
Part IV · The DynamicsGWW-2026-002

§ 18 · The policy-reaction functionWashington's put half-fails; Mumbai's ladder mostly works

In one line · the Fed can rescue everything except the thing at the epicenter; the RBI has six pre-built rungs and a 2018 lesson about using them early

Policy bends every branch, and the two benches differ in kind. The American one has a structural gap. In 2020 the Fed's corporate facilities bought CUSIP'd bonds and ETFs — securities with registered IDs, the only kind a facility can legally buy. No tool exists for buying a bilateral, unrated, private loan. So the rescue, when it comes, saves traded credit and leaves the epicenter to clear on its own — splitting the very market it means to calm. Rate cuts help refinancing at the margin, and might dissolve the retail queue; they do not re-open a gated wrapper. That is why "no put for the core" is a load-bearing assumption of the severe branch, not a rhetorical flourish — and why a genuine private-loan facility is one of our tear-it-up conditions.35

FIG 18.1 · The RBI's ladder — six rungs, sized and clockedthe India bench
1 · OMO liquidity — days · ₹10–40,000cr tranches (2018 template)2 · bank→NBFC conduit (TLTRO-style) — weeks · ₹50,000cr, ≥50% to small/mid NBFCs3 · fund backstop (SLF-MF) — 4 days in 2020 · ₹50,000cr against IG paper4 · fiscal guarantee (PCGS) — months · ₹1 lakh cr, 10% first loss, down to BBB+5 · FX defence — continuous · $693bn reserves + FCNR drive; 47.3% ratio is the limit6 · structural (pre-positioned) · conditional swing pricing, side-pockets, AIF dissolution clocks reaches AAA/AA onlythe CP-cliff rung — 2018's actual failure pointthe fastest rung India has ever deployedthe only rung that moves the BBB curvecaps the pace of the rupee's fall, not its levelbuilt, conditional — full swing needs a declaration never yet made; MF-scope only teal = fast/proven · amber = slow/political · crimson = structural
Source: RBI/PIB records 2018–2020 ●, SEBI MF Regulations 2026 ●. Sizes are the deployed precedents, not commitments for the next episode.32

The ladder's caveat is aim, not size. In 2020 an RBI review found only 2.1% of TLTRO money reached the small NBFCs that needed it — banks bought the safest eligible paper and stopped. A rerun that waits for AA spreads at +200bp before acting will again subsidise the strong.32 If you sit near policy, the useful version is timing: rung 2 pulled at +100bp of spread pre-empts the ratchet; pulled at +250bp, it finances the survivors' victory lap. And the newest Indian defence is built but untested. The swing-pricing framework can make a running debt-MF holder pay the cost of running — but full swing switches on only when SEBI declares a dislocation, which it never yet has, and it covers debt MFs, not the AIF vehicles this channel runs through. Count it as a rung on the ladder, not a wall. The dashboard that follows tells you, rung by rung, when each of these gets pulled.

Gravitywell Research23 / 33
Part V

The Playbook

Twelve checkable triggers turn the scenario into a live monitor, and a hedge book turns the monitor into positions — cheapest exactly where no instrument prices the channel, which is the Indian leg.

Part V · The PlaybookGWW-2026-002

§ 19 · SignpostsThe dashboard — current reading, distance to trigger

In one line · every row is checkable from public data; when the US column trips, you have two quarters of India lead time, sometimes four — that is the point of watching

When to pull each lever is a data question, and this page is the data. Two panels: the US trips first and gives the lead time; the India panel confirms arrival with a lag — a tripped US panel beside a quiet India panel is the cheap-entry window from §14, never reassurance.6

United States — the source
Fitch private credit default ratethe activation leg; hard defaults (≥25% of events) must confirm it
now 6.0% · watch ≥7.0 · leg ≥8.0 + hard ≥2%
Top-12 wrapper redemption requeststhe queue; falling NAV compounds it
now 12.1% avg · trigger ≥15% or any outright suspension
Honoured-fraction of requested dollarshow binding the gates are; post-queue-clear dispersion separates credit from plumbing
now 53.4% · trigger <50% twice running
Listed BDC median price / NAVthe traded market's verdict on the marks
now 0.80× · watch ≤0.70 · leg ≤0.65
US HY OASthe price leg; a doubling confirms
now 271–281bp · watch ≥400 · leg ≥550
PIK share of BDC loansthe deferred-strain stock
now ~10% · trigger ≥12%
BDC capital formation, YoYthe primary strike; funds the extension machine
now −55% · trigger −70% or a 2nd net-outflow qtr
India — the arrival
AAA private-NBFC spread over G-secthe funding channel's first gauge
now 50–140bp · trigger >250bp
NBFC 3-month CP yieldthe 2018 failure point, watched directly
now 7.80% · trigger >9.0%
Half-yearly private-credit deploymentthe rollover bid (EY-IVCA basis — definitional break vs the CY25 series)
now $2.2bn/half · trigger <$1.5bn with domestic-only participation
Promoter-financing covenant eventsthe SP template propagating
now 1 (Porteast, Apr-26) · trigger a 2nd large waiver or any forced sale
RBI monthly ECB registrationsdoes the widened pipe reprice or ration?
now ~FY25 run-rate ($61bn/yr) · trigger 3 months <50% of it
GIFT City foreign commitments, quarterlythe funnel — currently +148% YoY, AGAINST the thesis; Sep/Dec-26 prints adjudicate
now +148% YoY · trigger 2 declining qtrs
Gravitywell Research24 / 33
Part V · The PlaybookGWW-2026-002

§ 20 · PositioningWhat to do now, and what to do on the trigger

In one line · the "now" column costs basis points everywhere; the "if-active" column is where 2018's survivors made their decade — sequence them, don't choose

Position by stakeholder, in two tempos. The "now" actions are cheap because the tape currently agrees with the sceptics on page 27. The "if-active" actions are the ones history rewards — but only for holders who did the "now" work first. You cannot buy a gated book at 82 cents on the dollar if your own liquidity is stuck in someone else's queue.32

StakeholderNow — while no instrument prices the channel (0–3m, roll quarterly)If it activates — the trigger actions
Allocator / LP
(India-based)
Buy 6–12m Bank Nifty or Fin-Services puts near the 52-week low strike with volatility in its cheapest decile; receive 1–5y OIS; tender wrapper stakes while bids sit near NAV; cut AA-and-below NBFC paper, whose extra ~100bp does not pay for the gap risk.Deploy into credit secondaries at 80–85 cents on the dollar with genuinely locked capital (the Franklin math paid 109% to patience); add USDINR calls only on a reserve-drain signal — earlier, the RBI is selling you the vol.
Global allocatorCDX HY payer spreads from 271bp; trim evergreen exposure via tenders; do NOT initiate short-BIZD at 0.80× — that trade paid in 2025 and now carries the borrow for half the move.Rotate the payer-spread gains into newly raised direct-lending funds at the wide spreads — the 2016 pattern: the funds priced in the panic were the decade's best.
Indian corporate treasurer / NBFC CFOLadder CP so ≤15–20% rolls in any month; refuse the last 25bp for stretching 3M to 6M; pre-fund FY28 maturities at today's 50–140bp AAA spreads; term out ECB borrowing while the Feb-2026 regime is open and priced.Draw committed bank lines at the first +100bp of spread — 2018's window shut in weeks; assume the offshore NCD bid vanishes before the bank bid does.
Policymaker
(RBI/SEBI-adjacent)
Publish the AIF foreign/domestic commitment split — the data gap is itself the vulnerability; pre-position an SLF-MF-style window for AIF and credit-fund stress; monitor GIFT concentration quarterly.Pull the conduit rung at +100bp of AA spread, not +250 (the 2.1% lesson); pair any liquidity rung with the guarantee rung early — the BBB curve is where the drought becomes defaults.
The unglamorous edge

Across every seat, the highest-expected-value moves are cash-management, not trades: tenor discipline, pre-funding, line-drawing rights read before they are needed. IL&FS's casualties were not the worst credits; they were the best credits with the worst maturity calendars.32

What not to do

Do not short the Indian financial system into this: the banks are the scenario's ambivalent actor, not its victim, and the RBI's ladder is real. The trade is convex insurance on the funding channel plus patience for the assets it strands — never a directional bet that India breaks.

Gravitywell Research25 / 33
Part V · The PlaybookGWW-2026-002

§ 21 · The convex tradePriced insurance on a channel no instrument marks

In one line · ~1–2% of hedged notional a year buys a book that pays 4–10× on the moderate branch — the asymmetry is the position, and it is open now

The book, priced. Three legs, each with a cost, a payoff shape and a reason it is cheap: the market disagrees with the scenario, which is exactly the condition under which insurance is bought well.37

LegStructure & tenorCost (premium)Moderate-branch payoffWhy it is cheap
India equity legBank Nifty / Fin-Services puts, strikes ~10–13% OTM (the 52w low), 6–12m, rolled quarterly~40–80bp of notional per 6m4–8× premium on a −15/−25% moveIndia VIX 11.67 — the cheapest decile; the tape sees no channel
India rates legreceive fixed 1–5y OIS / long 10Y G-sec at 6.78%carry ≈ flat to +ve with repo 5.25%, neutral stance60–90bp of rally = multiples on DV01; pays in every branch short of an FX crisisevery rung of the RBI ladder is bond-bullish and the market prices no rungs
US credit legCDX HY payer spread, 400/600bp, 9–12m — or the institutional expression: pay HY against sold IG protection, beta-weighted ~4:1 (decompression)~35–50bp running6–10× on a doubling of spreads from the richest decilecarry at cycle-tights makes sellers of protection complacent
FIG 21.1 · The asymmetry — cost of the book vs payoff by branchindexed to premium = 1
−1× · no activationP ≈ 75–85% ~1–3× · mild (US leg)~8% uncond. ~5–7× · moderate~8% uncond. ~10×+ · severe~4% uncond. expected value ≈ (0.80 × −1) + (0.08 × 2) + (0.08 × 6) + (0.04 × 10) ≈ +0.2× premium — before the severe tail's fat edge
Gravitywell scenario model ○; payoff multiples are the mid of each leg's range, weighted equally. The mild branch pays through the US leg alone (OAS 550–700bp clears the 400/600 payer spread; the India legs pay from moderate — §17's mild readout moves neither Bank Nifty nor the RBI). Bought for the shape, not the expectation.34
Reading the trades — a plain key

A put is the right to sell at a set level; it profits when the market falls. "10–13% OTM" = that level sits 10–13% below today's, which makes it cheap. Notional = the amount being insured; premium = its cost; "roll" = renew on expiry. Receive-OIS = the side of an interest-rate swap that profits when rates fall (DV01 = the profit per 0.01% move). CDX HY = a traded index of credit default swaps on junk-rated borrowers; a "payer spread" profits when credit spreads widen between two set levels. Long = own; short = bet against; carry = what a position costs, or earns, while you wait; BIZD = the listed fund that tracks BDCs.

Two disciplines keep the book honest. Size it as insurance (1–2% of portfolio premium per year, never a directional view), and let the dashboard, not conviction, decide the rolls: rows tripping = roll up and out; two quiet quarters with the honoured fraction healing = let it decay and be glad. Page 29 states what you are paying for: a ~4% severe tail and a ~8% funding winter, on a tape with no instrument that prices either.

Gravitywell Research26 / 33
Pre-mortem & Red-teamGWW-2026-002

FalsificationWhy it happened — and the case that it never does

In one line · the pre-mortem writes itself from 2026's tape; two of the disconfirming objections survive this report unanswered
Pre-mortem · assume it happened

It started when the queue stopped clearing. The wrappers had gated politely through 2026 until a mid-sized perpetual fund, its honoured-fraction sinking, chose a liquidating trust over more pretending. The trust printed at 81 cents on the dollar; within a month auditors were asking why identical loans sat at 98, and the weak fundraising stopped entirely. The extension machine ran out of fresh capital, and the default index did the rest. Offshore institutions, gated in dollars, cut commitments everywhere; eleven months later a first Indian promoter missed a rollover with nobody in Mumbai having sold anything. The only surprise, in hindsight, was the calendar: everyone had watched the queue; almost nobody had watched the honoured-fraction.

The disconfirming view — the strongest case it does NOT happen

Six objections, strongest first. (1) Rotation, not run: retail leaving 8% credit for 7% annuities while the Fed holds; cuts dissolve the queue into an allocation story. (2) The freeze is already executing, orderly: LP intent fell 42%→29% in six months, BofA called the Q2 request peak, early Q3 reads agree — the design working. (3) Borrowers look cyclical: Proskauer 2.51% and falling, non-accruals ~2%, hard defaults ~1%. (4) The India channel runs the OTHER way: GIFT +148% YoY through the first gate quarter, FPI debt inflows three months running, a record all-domestic Kotak close — and the one direct offshore-seller statistic failed verification against the filing (iCapital reads the remainder as wealth-channel plumbing). (5) The core cannot run: the $1.6trn is mostly locked drawdown capital at ~1.0× leverage. (6) India's insulation is layered: domestic AIF money +21%/yr, NBFCs re-banked, the RBI ladder proven — swing pricing built, though untested and MF-scoped.7

Reconciliation, and its limits. Objections three, five and six cap severity, not odds — they are why mild carries 40% and why the India readout is a drought, not a crisis. The fourth is why the India leg is a load-bearing assumption, not a finding — the September GIFT prints adjudicate. The first two we decline to reconcile: if the queue is a rotation, or the freeze is executing in an orderly queue, our band is too high, full stop. They stand.

Standing anomalies — carried, not resolved

The divergence itself. Fitch's index-high default rate and richest-decile spreads cannot both be true of the same market; we cannot adjudicate which is wrong. The offshore-first indication. Press accounts put offshore requests near 3× onshore at the flagship wrapper; the filing neither discloses nor reconciles the split. Either the channel tell exists (each explanation gives the India leg a different half-life) or it is reporting noise; published data cannot adjudicate. Both survive unresolved.

Tear-it-up conditions

We retract if: Fitch prints under 4.5% for two quarters with requests uncapped and the listed median over 0.95× NAV; a Fed/Treasury facility accepts private loans as collateral (severe collapses into moderate); or BDC fundraising grows two quarters running.

Gravitywell Research27 / 33
Probability & assumptionsGWW-2026-002

The ledgerWhat the 15–25% rests on

In one line · consensus is printed beside the band; the Fed path moves P most, the no-put assumption moves the tail
FIG S.1 · The probability, drawn honestlyP(activation) / 18m
tape+strict ~5–10% ○ loose base ~23% P ≈ 15–25% / 18m the band sits below the loose base and ~2–3× above the tape — a stated variant, not a stealth one 0%50%100%
Band drawn over 15–25% only; pins at the loose class ~23% ● and the tape/strict cluster ~5–10% ○◐ — a disclosed variant; the ledger below prices the gap.34
Consensus & the variantStated, cited, priced
The consensusStreet: defaults peak ~8%, "significant but not systemic" (Morgan Stanley); BofA sees defaults easing and called the Q2 request peak; every major regulator has published the mechanism (FSB · IMF · Fed · RBI). Market-implied: index credit prices the seizure at ~5–10%/18m (a two-state read: today's 271bp spread treated as a blend of a calm price and a crisis price ○); BDC prices at 0.74–0.80× NAV embed ~25% cumulative defaults ◐; Saba tenders cleared 20–35% under NAV.12
The house delta15–25%/18m on the three-leg seizure (~2–3× the market-implied read), plus an India funding-channel transmission no traded instrument prices.
Why the mispricingGates turn a run into a quiet queue, extensions turn defaults into "amendments," marks lag prices by quarters — index spreads can sit in their richest decile while the funding structure erodes. And no Indian instrument spans an 18-month offshore-commitment risk: silence, not judgement.
What proves consensus rightQ3-26 requests falling again with the honoured fraction healing · hard-default share under ~20% of Fitch events · GIFT still growing through Dec-26 · BDC fundraising turning. Any two retire the variant.
Load-bearing assumptionStatusIf it breaks
No US facility reaches private, non-CUSIP loans in timeload-bearing · no construct exists todaysevere collapses into moderate; tail ~4% → ~2%
Evergreen behaviour persists (buffers can't fund repeated 5% quarters)load-bearing · NBER-modelled, live in the datathe ratchet stalls; P falls toward the base rate's floor
Marks converge to traded prices within 2–4 quarters of a real printload-bearing · Third Avenue/2020 precedentthe freeze stays contained to fundraising; India leg halves
The LP-identity assumption: the offshore money redeeming from US wrappers overlaps materially with the pool funding India (no published series sizes the overlap)load-bearing · the India leg's lead time and cheapness edge hang on itmagnitudes survive as generic risk-off; the lead time and cheapness thesis do not
FIG S.2 · The tornado — what moves P(activation), in pointsswing on the 20% mid
Fed path (cuts early vs holds) Rotation-vs-run resolution Extension-machine capacity AI-DC credit performance −8pp / +6pp−7pp / +5pp−5pp / +4pp−4pp / +3.5pp ← lowers P · 20% mid · raises P → · top driver ~1.9× the bottom; the no-put assumption moves the TAIL, not P
Gravitywell scenario model ○. Swings are the desk's judgement of each driver's resolution range against the 20% mid — argued in §§05–09, not fitted.34
Gravitywell Research28 / 33
Bottom lineGWW-2026-002

Bottom lineThe weighted answer

Private credit's first real test would reach India through its lenders, not its loans — and the arithmetic of that sentence is now on the table. A 15–25% probability over eighteen months of a three-leg seizure. Inside it: a 40% mild path that reopens within a year; a 40% funding winter that costs India a fundraising vintage and 60–110 basis points of NBFC spread; and a 20% statutory crossing that no facility is built to stop — a ~4% unconditional tail. Against that, the Indian tape offers financials volatility in the cheapest tenth of its range, credit spreads at cycle tights, and a rates market pricing none of the RBI's ladder. The mispricing is structural: the channel, the offshore LP balance sheet, appears on no Indian screen.37

15–25%
P(activation) / 18m — a variant view, ~2–3× the tape (~5–10% ○)34
~4%
unconditional severe tail — what the hedge is priced against34
~1–2%
of hedged notional per year — the three-leg book, all legs in37
4–10×
moderate-branch payoff multiple, by leg37

What would change our mind is written down. Two clean quarters of falling defaults, uncapped gates and healed fundraising retire the scenario. A private-loan facility guts its tail. And if the queue proves to be a rate rotation, our band was too high — that objection rides with the dossier, unresolved. What does not change our mind is the absence of Indian symptoms, because the absence of symptoms is the scenario: the drought arrives by calendar, on the maturity dates, quarters after the cause. Two anomalies ride with the dossier exactly as they are, because we cannot explain them: an index-high default rate beside richest-decile spreads, and the press-reported offshore-first skew the filing neither discloses nor reconciles.

The house record on adjacent terrain: our GWW-2026-001 dossier on the AI bubble placed the private-credit leg in its severe branch only, at a 20% conditional weight. The gates arrived in the base path within two quarters — we were early on the equity mechanism and late on the credit one, and this dossier is the correction. The two scenarios now share a lender set, a signpost row, and a lesson: the financing plumbing moves before the asset story does.40

The position

A roughly 8% chance of a funding winter and a 4% chance of a statutory crossing, against an Indian tape with no instrument that prices either: buy the financials puts while volatility is a gift, take the receive-fixed side of rates — the side that profits when the RBI's every response pushes rates down — and hold the US payer spread against the doubling — and let a dashboard, not a mood, decide the rolls. Insurance is cheapest where the channel is invisible, and this channel is invisible by construction.

Gravitywell Research29 / 33
MethodologyGWW-2026-002

MethodologyHow this dossier was built

The scenario method

The What-If series works one fixed sequence: define the shock as an observable activation condition; anchor the base rate in a counted reference class; assign probability as a band; trace transmission by order, one mechanism per link; map impact across assets, sectors, geographies, stakeholders and private capital; branch the outcome with argued weights; build a signpost dashboard with triggers; run a pre-mortem and a disconfirming view; end in positioning. Magnitude and probability never share a number.

Coverage & scope

In scope: US private credit on the FSB's narrow definition (~$1.5–2.0trn global, ~$1.6trn US), its wrapper layer, its insurer and AI-datacentre holdings, and the transmission into Indian private capital — AIF credit, NBFC funding, promoter finance, PE-VC fundraising and exits. Out of scope: European private credit, US bank solvency, single names. Activation legs are confirmable from the Fitch index, FRED OAS, and fund filings.

Source policy, verification & known gaps

Sources are numbered, tiered (● filed/official · ◐ modeled/derived · ○ Gravitywell estimate) and dated; primary regulators (Fed, Boston Fed, BIS, FSB, OFR, IMF, RBI, SEBI, IFSCA) anchor every load-bearing figure. Honest gaps, disclosed rather than filled: RBI FSR June-2026 figures are secondary-verified only, flagged ◐; SEBI publishes no private-credit sub-line and no foreign/domestic split, so no "foreign share of India private credit" number appears; India OIS and the USDINR forward premium are unverified to Aug-2026 and absent from the hedge arithmetic; the 2015–16 spread path is reconstructed from BIS documentation. Conflicting sources (Fitch vs Proskauer; two India AAA-spread series) are shown, carried, never averaged. v1.1 discloses: the twelve-wrapper panel is survivorship-selected; Moody's $25bn India AUM sits uneasily beside $12.4bn/yr deployment (refi churn); the v1.0 ADS split was withdrawn after failing 10-Q reconciliation.

Confidence rationale — Medium (P-confidence: medium-low)

US-side stress data is primary and current; the reference class countable; the India channel evidenced by the binding repurchase cap and GIFT composition. Held down: two anomalies unresolved, rotation-vs-run undecidable, key India calibrations single-episode. v1.1: band 20–30%→15–25%, leg 2 hardened, one press statistic withdrawn (redteam/thesis-verdict.md).

Point-in-time & revision

Markets as of 13 Aug 2026; fund data to the latest quarter; figures carry their own dates. Nothing is silently restated: errors correct via dated errata; the permalink serves the current version. The p24 dashboard, not this document, is the live surface.

House view scale

Conviction (High/Medium/Low) is set by coverage depth, source tier and stress-testing; Stance (Constructive/Neutral/Cautious) is the house view's direction. This dossier: Medium conviction · Cautious on the tail — the thesis is the asymmetry of the hedge, not a prediction of the crack.

Gravitywell Research30 / 33
Source registerGWW-2026-002

SourcesThe register

#Source · date · tier
1Boston Fed, Fillat/Shen/Wang, Early Warning Signals in Private Credit?, CPP 26-6, 5 Aug 2026 ●
2BIS Bulletin 128, AI disruption in private credit, 14 Jul 2026 ●
3FSB, Vulnerabilities in Private Credit, 6 May 2026 ●
4Federal Reserve, Financial Stability Report, May 2026 ●
5OFR Brief 26-02, Measuring Counterparty Exposures to Private Credit, 12 Mar 2026 ●
6Fitch US Private Credit Default Index, via Bloomberg, 30 Jul 2026 ◐
7Proskauer Private Credit Default Index, Q2 2026 ◐
8KBRA, Q2 2026 Middle Market Compendium, 28 Jul 2026 ◐
9With Intelligence, BDC portfolio & redemption data, Q1 2026 ◐
10PitchBook: BCRED Q2 letter; BofA redemption forecasts, Jun 2026 ◐
11Stanger / AltsWire, BDC capital formation, through May 2026 ◐
12Morgan Stanley via Bloomberg, default path, 16 Mar 2026 ◐
13BIS Bulletin 120, Financing the AI boom, Jan 2026 ●
14IMF GFSR, Apr 2024 ch.2 + Apr 2026 ●
15ICE BofA US HY OAS via FRED, 271bp at 6 Aug 2026 ●
16Apollo Debt Solutions BDC 10-Q (30 Jun 2026) ● + CNBC/InvestmentNews, 23 Jun 2026 ◐
17RBI, Financial Stability Report, June 2026 (via secondary summaries) ◐
18RBI, India's External Debt at end-March 2026, 29 Jun 2026 ●
19RBI (Investment in AIF) Directions 2025, 29 Jul 2025 ●
20S&P Global Market Intelligence, 2026 US Insurance Investments report ◐
#Source · date · tier
21SEBI: AIF consultation 30 Jun 2026; MF Regulations 2026 (swing pricing) ●
22EY India Private Credit Report H2-2025, Feb 2026 ◐
23EY-IVCA, PE/VC H1 2026 roundup, Jul 2026 ◐
24Chambers, Private Credit 2026 — India chapter ◐
25CRISIL Ratings, NBFC funding mix, Apr 2026 ◐
26Business Standard: SP Group deal 16 May 2025; Porteast waiver Apr 2026 ◐
27IFSCA, GIFT-IFSC fund management data, Mar 2026 ●
28McKinsey, India's private markets: the global LP view, 11 Mar 2026 ◐
29Coller Capital, Global Private Capital Barometer, Summer 2026 ◐
30Jefferies Secondary Market Review 2025; PitchBook credit secondaries 2026 ◐
31Mercer Capital BDC discounts 2026 ◐; Suhonen (2024) appraisal alpha ●
32Business Standard archive: IL&FS 2018, DHFL, Franklin 2020, RBI/PIB facility records ◐
33Fang/Goldstein/Zeng, NBER WP 35385, 2026 ●; FCA Woodford record ●
34Reference-class data: Altman/NYU 2002 ●, NY Fed HY history ●, FDIC S&L chronology ●, BIS QR Mar 2016 ●, SEC Third Avenue filings ●
35Fed FEDS notes: SMCCF (Oct 2020) ●, ABCP (2009-36) ●; BoE QB 2023 LDI ●; BDC Mar-2020 pricing ◐
36AMFI Jul 2026 ●; SEBI AIF commitment data Dec 2025 ●; RBI MPC 5 Aug 2026 ●
37Market readings 10–13 Aug 2026: Fed H.10 USDINR ●, India VIX ◐, NSE indices ●, RBI WSS reserves ●
38Preqin (via With Intelligence) $2.1trn Q1 2026 ◐; FSB/OFR sizing ●
39S&P Global, India's private credit market is coming of age, 2026 ◐
40DC-SPV records: Bisnow (Meta/Blue Owl) ◐, press aggregation of Oracle/xAI vehicles ◐; Gravitywell GWW-2026-001 ○
Tier policy

● filed/official: the source of record, opened directly. ◐ modeled/derived: computed from primary inputs, or a credible secondary carrying a primary number we have not opened at source. ○ Gravitywell estimate: a desk judgement, always also disclosed in the methodology. Consultancy reports, broker research and press are never ●, whatever they cite. The data pack (private-credit-unwind-data.csv) carries every exhibit's series, values and dates, keyed to this register's numbering. The Part V signpost panel is maintained as a living desk dataset (data/private-credit-signposts — quarterly cadence, append-only), so the dashboard outlives this edition.

Gravitywell Research31 / 33
GlossaryGWW-2026-002

GlossaryTerms, as this dossier uses them

Private creditnonbank bilateral lending to companies — loans negotiated directly between a fund and a borrower, never listed, rarely rated publicly.
BDCbusiness development company — the US fund structure (listed or non-traded) through which outside investors, including retail, own portfolios of direct loans.
The wrappersthis dossier's term for the redeemable vehicles: non-traded BDCs, interval and evergreen funds — the layer where a run can happen.
The gatesthe 5%-of-NAV quarterly caps on wrapper redemptions; "gating" = enforcing them pro-rata.
PIKpayment-in-kind — interest paid by issuing more debt instead of cash; defers the strain, grows the claim.
Amend-and-extendrestructuring a maturing loan by pushing its maturity out, usually with sweeteners; "the extension machine" is this practice at industry scale.
OASoption-adjusted spread — the extra yield over Treasuries that compensates credit risk; this dossier's price-of-risk gauge.
NAV / marks vs pricesNAV is a fund's appraised net asset value; "marks" are those appraisals, "prices" are what trades — the gap between them is a recurring character here.
Liquidating trustthe vehicle a suspending fund moves assets into for orderly wind-down — it converts marks into prints.
NAV loan / subscription linebank credit to a fund secured on its portfolio (NAV) or its LPs' unfunded commitments (subscription) — the channel that connects fund losses to banks.
Private letter ratinga non-public credit rating issued to a single investor, often by a smaller agency — the grade much insurer-held private paper carries.
AIFAlternative Investment Fund — India's regulated private-fund structure; Category II houses most private credit.
GIFT CityIndia's offshore-rules financial centre (IFSC) — the dollar-denominated funnel through which foreign capital funds Indian credit.
ECB borrowingexternal commercial borrowing — Indian entities borrowing offshore under RBI rules (liberalised Feb 2026); no relation to the European Central Bank.
NBFC / CPnon-banking financial company — India's non-deposit lenders; CP = the short commercial paper that funds them, the 2018 fault line.
FCNR(B)foreign-currency deposits from non-resident Indians — an RBI tool for raising dollar funding without selling reserves.
OIS / receive-OISovernight indexed swap; "receiving" fixed profits when rates fall — the way this dossier expresses the RBI-response trade.
DPIdistributions to paid-in capital — the cash LPs have actually received; the fuel for new commitments.
Denominator effectwhen one part of a portfolio falls or freezes, the others breach their allocation ceilings mechanically — forcing commitment cuts with no view taken.
Continuation vehicle / secondariesthe resale market for fund stakes and assets; the scenario's price-discovery venue and, at a discount, its exit.
Axis discipline

Throughout: "de-rate / drawdown / markdown / −X%" describe magnitude; "odds / base rate / probability / P" describe likelihood; "weight" appears only for branch weights, which are conditional on activation and sum to 100% inside the scenario. Where one sentence carries both axes, both words appear.

List of exhibits

E.1 the two axes (p04) · Q.1 the cascade in five beats (p05) · 1.1 where the $1.6trn sits (p06) · 2.1 the reference class (p07) · 3.1 the 2015–16 clock (p08) · 4.1 IL&FS, the India calibration (p09) · 5.1 the divergence (p10) · 5.2 the queue against the gate (p10) · 6.1 the transmission chain (p11) · 7.1 the strike (p12) · 8.1 what a "default" is in 2026 (p13) · 8.2 one loan, four prices (p13) · 9.1 the quiet migration (p14) · 10.1 one balance sheet, two ends (p15) · 11.1 impact by asset × horizon (p16) · 12.1 the software pool (p17) · 13.1 the thinning cushion (p18) · 14.1 the lag structure (p19) · 15.1 the book that must roll (p20) · 16.1 the queue ratchet (p21) · 18.1 the RBI's ladder (p23) · 21.1 the asymmetry (p26) · S.1 the probability, drawn honestly (p28) · S.2 the tornado (p28). Data pack: private-credit-unwind-data.csv, keyed to these numbers.

Gravitywell Research32 / 33
Disclosures & governanceGWW-2026-002

DisclosuresCertification, conflicts & governance

Certification

The Gravitywell Research desk responsible for this report certifies that the views expressed accurately reflect its independent judgement about the subjects and securities discussed, and that no part of its compensation was, is, or will be directly or indirectly 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 or assets discussed. Gravitywell Research has no advisory, banking, or commercial relationship with the entities named. This report was not commissioned or reviewed by any issuer named.

Scenario disclaimer

This is a conditional scenario analysis, not a forecast or prediction. It studies what would follow if a defined activation condition were met, weighted by an explicit probability band. It is not a statement that the condition will be met, not a rating on any fund or issuer named, and not investment advice. The hedging structures discussed are illustrations of payoff shapes, not recommendations sized to any reader's circumstances.

Prior-calls scorecard

GWW-2026-001 · Jul 2026 — "the AI bubble bursts": placed the private-credit contagion leg in the severe branch only (20% conditional weight). The wrapper gates and the first net-outflow quarter arrived in the base path within two quarters. Verdict: too conservative on the credit leg — corrected here, where the credit leg is the scenario. The equity-side call remains open and unscored.

Governance

Errata: material errors are corrected in a dated erratum appended to this report and noted in the next edition; the permalink serves the current version. Permalink: gravitywellresearch.xyz/whatif/private-credit-unwind. Contact: [email protected]. Version: GWW-2026-002 · v1.1 · as of 14 Aug 2026 — v1.1 same-day revision: band 20–30%→15–25%, leg 2 hardened, one press statistic withdrawn (Methodology; redteam/thesis-verdict.md).

Distribution

For professional and institutional investors across public and private markets, sophisticated individual investors, and policymakers. Not for general retail distribution, nor for readers who lack the expertise to assess the assumptions. Intended recipients may not redistribute without attribution. Availability of this research in some jurisdictions may be restricted; recipients are responsible for their local rules. Figures are marked to the dates shown and may be revised; while drawn from sources believed reliable, accuracy is not guaranteed.

Gravitywell Research33 / 33
Gravitywell Research · What-If
What if private credit cracks?

A 15–25% probability over eighteen months that the $1.6trn market's funding structure fails — and a transmission into India that runs through the offshore LP balance sheet, arriving at the refinancing calendar quarters later, priced by no Indian instrument at all. A ~4% severe tail the hedge book is built against: a convex position, not a call.

GWW-2026-002 · Aug 2026 · Gravitywell Research Global Credit & Capital Markets Desk
Conditional scenario analysis. Not a forecast, prediction, or investment advice.
Independent research; not regulated ratings, not an offer or solicitation. Figures marked to the dates shown.
© 2026 Gravitywell Research.