There is a number in this issue that looks like good news and is not. India's private-capital holders realised about ₹37,201 crore in August, on the desk's own count. Block-deal volume hit ₹98,353 crore across 1,459 trades, the highest month of 2026 and the third highest ever recorded.⁵² Read the headline and India's exit problem has been solved.

Read the ledger and 5.3% of that realised value came from a fund actually finishing with a company. The rest was trimming.

That gap — between cash arriving and a position closing — is what this magazine exists to measure. The tape cannot report the distinction, because the tape cannot see it. A block is a block whether it takes out a holding or shaves it, and an offer-for-sale looks the same. It is visible only if somebody classifies every realisation by what happened to the position afterwards. The desk did that to forty-three of them, and publishes the classification so a reader can argue with it.

The honest miss from last month is larger, and it is ours. In July this desk argued that a funding channel had closed, and retired the argument in a stop-press on the strength of one bond order book. August then destroyed it far more comprehensively, with a record month of investment-grade supply absorbed at pace. Conceding on the weaker evidence three weeks before the stronger evidence arrives is not humility. It is a failure of patience, and the Scorecard says so at length.

One absence in this issue is worth more than most of its numbers. The desk swept August — India's annual-report peak — for filed governance signals at companies with a private-capital sponsor on the cap table, and found none. The month produced a full population of going-concern and qualified opinions, and not one of those companies had a sponsor. Either sponsored companies are better governed, or sponsored companies do not file. Our own sweep is evidence for the second, and the Panel prints the absence with its reason rather than a zero.

That is the arrangement. Every figure tiered, every absence explained, every call carrying the condition that would break it. ■

The defining number is 5.3%. India realised about ₹37,201 crore of private-capital value in August, the loudest month of the year, and three sponsors finished with a company. Everything else was a partial sale. The cover story takes the month apart; the short version is that cash arriving and a position closing have stopped being the same event.

Two windows did all of it. Thirty-six realisations were exchange block or bulk trades, and sponsors sold into four of the month's eighteen mainboard listings. Strategic sales to corporates and continuation vehicles recorded nothing above the bar. PRIME Database counted ₹98,353 crore of block deals across 1,459 trades, the highest month of 2026 and third highest on record.⁵² A market can be busy and narrow at once, and India was both.

Venture rebounded on fewer, bigger cheques. Forty-one disclosed rounds at or above ₹25 crore, worth roughly ₹9,120 crore. Twenty-two cleared ₹100 crore. Five rounds at or above ₹500 crore carried about half the month's value, while nineteen rounds below ₹100 crore carried an eighth. Deal count fell year on year even as value held. The concentration is the story, and it mirrors the exit market exactly.

Private credit is invisible, and that is a finding rather than a quiet month. Sweeping more than sixty named lenders one by one returned six publicly announced facilities totalling ₹2,100 crore, not one of which disclosed a coupon or a tenor. Against the institutional run-rate the public record captures perhaps one deal in three. Two of the six facilities were acquisition finance — in the first full year of a rule letting banks fund up to 75% of an acquisition, which is the competitive squeeze arriving in real time.

Foreign money came back to the tape. Foreign investors bought a net ₹29,631 crore of Indian equity in August, the largest monthly equity inflow of 2026 — though the calendar year is still ₹2.24 lakh crore in deficit on that line.⁴² The Nifty 50 still fell 1.24%. The Midcap 150 rose 1.72%, the Smallcap 250 2.52% and the Microcap 250 5.41% — a monotonic breadth divergence, and where most sellable sponsor positions actually sit.

What last issue said would happen. No. 003 named the RBI's 5 August meeting as its first catalyst and put a hold at 75–85%. It held. The issue's other August catalysts were macro and resolved against its thesis; the Scorecard grades them at length, misses first.

Watch this one. On 13 August the statutory auditor of IL&FS Investment Managers — India's oldest listed private-equity manager — resigned with immediate effect, days after signing off the June quarter's limited review.³¹ The filing gave no reason. It is not in the Panel's governance series, because the company's promoter is not a private-capital sponsor and the desk applies its own definition even when the story would be better without it. ■

That transaction is unusual enough in India to be worth naming, because in the whole of August 2026 it happened three times.

The desk logged forty-three realisations at or above its ₹50 crore census bar last month, worth ₹37,200.54 crore, about US$3.89 billion — roughly three and a half times the first half's monthly run-rate and more than twice July's. The tape agrees: PRIME Database counted ₹98,353 crore of block deals across 1,459 trades, the highest month of 2026 and the third highest on record.⁵² By any reading of the headline, India's exit market had its best month of the year.

Now take the same month apart. Of that ₹37,201 crore, ₹1,977.59 crore came from a sponsor that actually finished with a company. Five point three per cent. The other ninety-five went to funds that sold some stock and stayed long.

Brighton Park's Indegene sale was the largest of them.¹⁷ On 26 August, Lightspeed Opportunity Fund II sold out of PhysicsWallah, an education company it had backed at 1.61%, for ₹549.73 crore.³⁸ The next day Alpha Wave India I sold its 7.78% of Aye Finance, a small-business lender, for ₹322.60 crore.³⁹ Tiger Global also left The Viral Fever, an online studio, on 13 August, but that one was off-market and nobody disclosed a price.²⁹

And the exits came through two windows, not one. August was the busiest month of the year in India's primary market: eighteen mainboard listings, sixteen closing their first day above the issue price. Sponsors sold into four. Only LEAP India disclosed a per-seller split — KKR's Vertical Holdings II took ₹1,998.62 crore of a ₹2,000 crore offer-for-sale — so the Bain, TPG, Temasek and Novo rows are logged unpriced rather than apportioned.

The labels matter more than they look. India's control rules record KKR's and Bain's vehicles as promoter sellers rather than investors, though both are private-equity exits in substance. Series built on those labels count the form and not the money, which is one reason the published figure for India's first-half offer-for-sale realisations looks so small. And an offer-for-sale is no more a completed exit than a block: the seller takes a portion and a lock-in on the rest. Not one went to zero.

Everything else was a slice. On 24 August SoftBank's Vision Fund II sold ₹2,887.87 crore of Lenskart and remained a holder.³⁶ Four sessions later Alpha Wave sold ₹1,856.78 crore of the same company and remained a holder.⁴⁴ On 12 August Bain Capital sold ₹2,325.28 crore of Embassy Office Parks REIT and remained a holder.³¹ Twenty-eight of the month's thirty-six block trades are partial sales of that kind, and together they are ₹28,685.17 crore, 77.1% of everything realised. Add the offer-for-sale sellers, none of whom left either, and 82.5% of August's realised value came from a holder that is still long the asset.

The distinction is mechanical, and it is not about size. A negotiated sale ends a holding. A strategic buyer or another sponsor buys the asset, signs on a date the seller helped choose, and takes the whole position. The price is struck privately. The deal can clear in a falling market, because the buyer is buying the business rather than the quote.

A block does none of that. It needs a bid, at size, on a particular day, at a price anchored to the last screen print. So the seller does not hold the exit; the tape grants it. And crucially, a block is divisible in a way a negotiated sale is not. A fund that wants out of a large position and cannot find a buyer for all of it can always sell a bit — and a bit is what August was made of.

There is a respectable case that none of this is a problem. Trimming into strength is what a disciplined manager does, and a fund holding a recent listing at a high multiple should sell into a domestic bid rather than wait for a strategic buyer who may never appear. Every rupee realised in August was cash from identified buyers — mutual funds, insurers, the NPS Trust, sovereign funds, long-only foreigners. None of it was paper. Nor is the market being coy about scale: promoters and funds sold some ₹58,000 crore on the wider measure and the tape absorbed it.⁴¹ A channel that takes that much in four weeks is not broken.

The desk's answer is that absorbing is not clearing. A fund that has sold 2.6% of a 9.9% holding has returned cash without returning the company; the rest is still marked, still exposed, still waiting for a bid it does not control. Distributions look like progress while the position that has to be got out of entirely is unchanged. That is a market deferring rather than clearing, and August was decisively the second.

One more number frames the rest. The ten largest transactions were 39.5% of everything the desk recorded this month, and five names carried most of the realised value. A national exit market whose monthly figure is set by ten order books is not reporting the health of a market. It is reporting the depth of ten books.

The strongest evidence sits in what did not happen.

In the whole of August, at or above the desk's bar, there was not one strategic sale to a corporate buyer. Not one continuation vehicle, GP-led secondary, strip sale or LP-interest sale. Six of the ten routes the desk tracks recorded nothing at all, in a month that set records in the two that worked.

The continuation vehicle is the notable absence. Globally it runs near 14% of secondary volume and exists for exactly this situation. A manager who cannot sell outright moves the asset into a new vehicle. Investors choose cash or continuation, and the manager buys time without pretending to have exited. India closed none — though on 27 August Siguler Guff was reported to be raising a US$350–400 million continuation fund for four Indian companies. A formation, not a realisation.

That absence has a consequence the desk prints against its own argument. All of August's realised value is genuinely LP-reachable cash, and not because Indian managers chose to distribute. It is because they have no alternative structure to choose. India's exits are clean because its plumbing is primitive.

For an allocator, a re-up underwrites something new. Not only whether a manager can pick assets, but whether it can hit a tape. The distribution forecast assumed strategic sales and listings. It is being met by drip-fed blocks, and the drip stops when the bid does.

For a general partner, an entry multiple now has to carry the cost of an exit that may only exist in a rising market. The 2021 vintage did not underwrite that. For a founder, the buyer of your next round is pricing its own illiquidity rather than your business.

And for the regulator, the lever is not the listing window. It is the depth of the block market, and the absence of a continuation-fund framework. Those are questions of market structure, which SEBI can act on. Sentiment is not.

Series 13 contradicts the month's headline. Committed per rupee realised — closes over realised exit value, both off the desk's own ledgers — came in at 0.76: India returned about ₹1.32 for every rupee committed. The widely quoted figure implied the opposite, and the difference is perimeter rather than arithmetic. It counts a pan-Asia fund as India. Part IV takes it apart.

Series 10 is the cover, computed mechanically. Sponsor-to-sponsor share of exits: 0.0% by value. That is not quite the same as nothing happening. One sponsor-to-sponsor transaction did print. Kedaara Capital agreed to buy Tynor Orthotics for US$200 million on 24 August, with Lighthouse Funds among the sellers. Neither party disclosed how much of the consideration went to the fund rather than to the promoters. An unattributable value is not a value, so the series carries the row and reads zero. The count is one; the money is unknown.

Series 12 says the month was narrow. The ten largest transactions were 39.5% of all recorded value. Concentration at that level means a national market's monthly figures are being set by a handful of names, and that a single thinning bid moves the aggregate.

Series 4, 5 and 6 are why the census bar exists, being drawn from the whole ledger down to ₹25 crore rather than from the named deals. Median round size was ₹130 crore across 41 rounds. Median step-up is 1.55× on n = 4 — both post-money and prior post-money are public for only four rounds all month — and down-round share reads 25% on the same four. The desk prints n on the face of the series rather than let four cheques pass as a distribution.

Read series 11 against series 10 and the month resolves into a sentence. The public window ran above capacity on a backlog it did not create; the private routes that need a negotiated buyer produced one transaction nobody priced. In August, only sellers already in a queue could choose their moment.

Series 12 and 13 disagree, and the disagreement is the issue. Concentration says ten transactions were two-fifths of everything recorded. Committed per rupee realised says India returned more than it took in. Both are true, and together they describe a market whose aggregate looks healthy because a small number of large, liquid names carried it. Remove those ten and neither series reads the same. The Panel prints both rather than the flattering one.

The five that print an absence. Deployment overhang needs a prior month's stock and this is the first reading, so the series begins next month rather than compounding a number from nothing. Exit velocity needs twelve months of trailing deal value. The private-credit spread cannot be computed because neither of August's two publicly announced facilities disclosed a coupon — the absence is the beat's finding, not the desk's gap. Industry distributions over paid-in reads SEBI's quarterly statistics and is dark two months in three by construction. Governance signals returned zero rows that clear both the filing test and the sponsor test, which the desk defends as a sourced nil rather than a thin sweep.

Coverage this month: 91.0% of logged transactions carried a disclosed value. Every figure above is computed at PANEL_VERSION 1.2.0 and appended, once, to the standing series.

The conversion rate, stated once for the issue. Dollar and euro figures are converted at ₹95.6 to the dollar. FBIL's own mean of published August reference prints is ₹95.47, a difference of 0.14% that moves no figure in this issue materially. Note that 29 August 2026 was a Saturday and carries no reference rate; where a single dated spot is wanted, the last published August print is ₹95.7143 on 25 August.

On what the Panel is for. None of these sixteen readings is worth much on its own. The value is the series: the same definitions, the same ledgers, the same method, twelve times a year. A rival can copy the arithmetic in an afternoon and cannot copy the history. Only the history will say whether August was a regime or a month.

The Clock. The house index family reads the July 2026 vintage — the composites publish a month in arrears, and the vintage is stamped rather than quietly forward-dated. The Capital Cycle Clock puts India at Peak / Late-cycle, momentum 112.96 against fragility 110.47: capital running hot while fragility builds.⁵¹

Two members matter to this issue and they are moving apart. The Capital Formation Index reads 154.09 and rising. The Capital Realisation Index reads 111.94 and falling. Capital is being formed faster than it is being realised, which is this month's cover thesis stated by an engine that was not built to argue it. The Domestic Capital Index fell 2.98 in the month, the sharpest move in the family; Part IV takes that up.

Methodology version v3.2 (2026-07). The desk reads these numbers and does not recompute them. ■

Every block-market realisation the desk logged in August, by sector. Consumer, healthcare, fintech and pharma took two-thirds of the month; each has recent listings and a live domestic bid. Nothing came out of enterprise software, agritech, edtech or climate — the sectors holding India's 2016–2019 vintages, all unlisted, all unreachable by August's one route. ■

The cell that changed is consumer. It took roughly ₹1,150 crore of fresh venture money across eight rounds and returned ₹6,664 crore of realised value — the largest of either column. Those are not the same companies. The money went into early-stage brands; the money came out of a listed eyewear retailer that two sponsors were trimming. A sector can be simultaneously the market's favourite place to deploy and its favourite place to sell, and consumer was both in August.

Healthcare and pharma are the consolidation story, together ₹10,010 crore of realisations against almost no venture inflow. That is a sector being bought rather than built, and it is where the month's M&A actually happened.

Healthcare is the month's clearest single story. It took almost no venture money and returned ₹6,459 crore, and the buyers were foreign. KKR bought Medicover's Indian hospital business outright, TPG sold down Aster DM through the exchange, and a smaller strategic bought a Bengaluru women's hospital. A sector that consolidates while taking no new venture capital is one where the returns are being captured by whoever already owns the assets, not by whoever funds the next entrant.

And the empty column is the finding. Enterprise software, agritech and climate took money in and returned nothing above the census bar. India's 2016–2019 venture vintages are concentrated in exactly those unlisted sectors, and the block-deal window that carried August cannot reach them.

That is the matrix's real use, and it is why the desk prints both columns rather than a ranking. A league table of the month's largest deals would have put consumer, healthcare and fintech at the top and stopped there. The two-column view shows what actually matters to an allocator. The sectors taking capital in are not the sectors giving it back, and the gap between those lists maps where the next four years of distributions have to come from.

The Dossier picks its sector from this page rather than from interest. This month the flagged cell is infrastructure and data centres — a column with real capital arriving, no realisations at all, and a reported inflow larger than the market that is supposed to contain it. ■

Identical quantity. Identical price. Hero's own announcement put the purchase at ₹17.58 billion, or about US$184 million, and did not name the seller.⁴⁰ The exchange block file names it.⁴⁵

This is the month's most instructive transaction, and almost nobody filed it as a deal. It appeared on the tape as two block trades. Read as a deal, it is a strategic acquirer consolidating its position by buying a sovereign wealth fund's stock at the last screen print.

The consideration mix is the simplest part. All cash, one tranche, settled on the exchange. No earn-out, no deferred payment, no swap, no escrow. A negotiated purchase would carry at least two.

The price is the interesting part, because there is no premium in it. ₹1,480 was the block price. A negotiated purchase of a 2.92% strategic stake would normally carry a premium. The seller delivers something the buyer cannot assemble on the open market without moving the price against itself. Here the buyer paid the screen. The premium a strategic would usually pay went unpaid because the seller was a financial holder that wanted out and was willing to transact at the mark.

The desk states plainly that the creeping-acquisition reading is its own inference from the disclosed holdings and the code, tiered accordingly. What is observed is the quantity, the price, the counterparties and the resulting stake.

Who won. The buyer, comfortably. Hero acquired incremental control of a listed asset at no premium, in one session, with no offer obligation and no conditionality. Its holding is now within touching distance of the one-third threshold that matters for special resolutions.

The seller did fine rather than well. GIC realised ₹1,758 crore at a clean price without having to negotiate. Whether it exited entirely is not established from the file — the classification problem this issue is about. The desk carries the row as a completed sale because it cannot show a residual, not because it can show there is none.

What it signals for the next ten deals like it. Three things.

Sovereign and long-only financial holders are the natural sellers into this structure, because they are indifferent to control and sensitive to execution. India has a great many of them sitting on 2021–2023 vintage positions in recently listed companies.

Strategics with existing stakes above 25% now have a cheap, quiet route to consolidate, and the creeping limit resets every financial year. Expect this to recur in April.

And for the magazine's own cover argument, it is the exception that sharpens the rule. A block moved control here — but only because a strategic buyer happened to want the stock. Where no strategic wants it, the same block market returns a slice and leaves the position open, which is what it did in twenty-eight of the month's other thirty-six trades.

That sequence is unusual. Aviom was taken into insolvency through the route only the regulator can invoke, on the regulator's own application, and what preceded it was not a default. It was the company's own statutory auditors recording discrepancies in the books of account, and a fraud reported to the National Housing Bank.³

What was underwritten. Lending against property to low-income women borrowers, largely outside the big cities. There was a genuine social case and a genuine commercial one. Four private-capital investors backed it, and between them they owned more of the company than its founders did. Nuveen, the asset manager of the American pension giant TIAA, held 35.06%. The Japanese group Gojo & Company held 19.84%, Sabre Partners 9.88% and Capital 4 Development Asia Fund 3.78%. The promoter and her family held 31.4%.¹

What the entry price assumed. That the loan book was the loan book. Every valuation of a lender rests on the assets existing and performing as recorded; growth, funding cost and credit cost are all second-order beside it.

The point at which the thesis stopped being true is not a date on the tape. It is whenever the books stopped describing the assets, necessarily earlier than the day anybody said so — which is why this failure mode is so expensive. The loss is complete before it is visible.

What it cost, and to whom. Sixty-seven secured financial creditors filed claims of ₹1,550.35 crore. Sixty were admitted, for ₹1,328.22 crore.⁵ Those creditors are banks and financial institutions, and behind them sit Indian depositors. The equity ranks behind all of it — the sponsors' ₹256 crore of ordinary and convertible shares, and everything the founders held.

A second observation is worth more than it first looks. The company's credit rating rationale was hosted on the website of Sabre Partners — one of the sponsors on its own cap table.¹ Nothing improper follows: investor sites carry portfolio material routinely. But it shows how thin the independent information layer around an unlisted Indian lender is. Much of what looks like third-party verification is circulated by parties who are not third parties.

And it is not over. Unity Small Finance Bank was reported to have won the company with an upfront bid of about ₹977.5 crore, roughly 71% of admitted claims, which would be a good outcome by Indian standards.⁶ By February 2026 that bid was being described as unsuccessful, and the administrator's most recent list of creditors is dated 26 February 2026. Eighteen months after admission the process is still running. A former director's appeal was rejected by the appellate tribunal.⁷

The company's response. The administrator is a named insolvency professional publishing to the company's own site, and Nuveen, Gojo, Sabre Partners and Capital 4 Development are all institutions with press functions. The desk has not obtained a statement from any of them, and prints that rather than an inference.

Why this one, and not a larger name. Because nothing about the cycle caused it. A fund can lose money to rates, to competition, or to a market that turned, and an allocator will accept all three as the cost of taking risk. What an allocator cannot accept is capital lost to assets that were not there. Four institutional sponsors, a majority of the equity between them, board representation — and the failure still arrived through the audit rather than through the market. ■

Sixteen transactions cleared the desk's census bar, worth about ₹18,138 crore, and three of them were more than half of it. A seventeenth, KKR's purchase of Medicover, is logged to the PE beat.

KKR's purchase of Medicover's Indian hospital business is the month's largest, and it is the month's clearest lesson in perimeter. One transaction carries three defensible values. The seller's own release places an enterprise value of EUR 1.2 billion on 100% of the platform — 24 hospitals, about 4,800 beds. It separately expects gross cash proceeds of about EUR 740 million on the 66.1% Medicover AB held, which is ₹8,128 crore and the only figure the desk can observe directly. And some outlets carried "around ₹10,000 crore" on sources. The ledger records committed consideration rather than enterprise value, so it logs ₹8,128 crore and prints the other two here.¹⁸

Two desks inside this magazine logged that deal on two different perimeters and produced numbers ₹5,000 crore apart. That is what the rule exists to prevent, and it is why every row in the register carries the perimeter it was priced on.

The month's largest is a bank buying into a lender. Bank of America's investment in Jio Credit is carried here at ₹6,612.9 crore, which is the subscribed equity. The headline circulating in August was ₹18,268 crore, and the difference is warrants exercisable over eighteen months. Printing the headline would have inflated the month by roughly the size of the entire real transaction — the sort of error that is invisible until somebody adds the ledger up.

Second, an Indian services firm bought a German one. TCS's purchase of MHP Management- und IT-Beratung is logged at about ₹3,566 crore. Outbound M&A from Indian services companies is the quietest consistent story in this beat and it did not stop in August.

The sector reading is the one the Sector Map flagged. Healthcare and pharma between them produced ₹10,010 crore of realisations and almost no venture inflow all month. Add Medicover and the picture is unambiguous: this is a sector being consolidated by strategic and financial buyers, not one being built by new capital. Kinder Women's Hospital at ₹130 crore is the same trade at a different scale.

And one row that did not clear the bar, deliberately. Wipro's purchase of Dermatouch reads ₹387.5 crore on the enterprise value of 100%, but the committed consideration for the 60% actually bought is about ₹232.5 crore — below the ₹250 crore record bar. It sits in the ledger at the census tier and is not named as a deal. Enterprise value moves transactions across thresholds in both directions, and this beat is where that happens most often. ■

The terms move before the prices do, and in August they moved inside the headline number itself.

Take the four largest venture rounds of the month as they were reported. River Mobility raised US$120 million. Yulu raised US$93 million. Centricity raised ₹280 crore. Third Wave Coffee raised ₹408 crore.

Now take them as they were structured. River Mobility's total bundles venture debt from three lenders, and the split was not disclosed. Yulu's is US$63 million of equity plus US$30 million of debt. Centricity's is ₹230 crore of equity plus ₹50 crore of venture debt. Third Wave's includes an undisclosed secondary leg, so the primary issuance is smaller than the headline by an unknown amount.

Strip the debt and secondary legs across the month and the trackers' equity total loses several hundred crore. The desk's ledger carries the equity legs only, which is why its August venture figure and the aggregators' will never reconcile.

Why this is a terms story and not an accounting one. Venture debt sits ahead of equity. When a company raises ₹230 crore of equity and ₹50 crore of debt and the market records ₹280 crore of funding, two things are obscured at once: the equity holders bought less of the company than the number implies, and somebody now has a claim that ranks above them. Both matter at the next financing, and neither is visible in the reported figure.

The other drift is at the letter of the round. Five of the desk's 41 logged rounds carried no new price at all — bridges and extensions, 6.8% of the month. One was explicitly a Series A extension taking a round from US$5.5 million to US$15.5 million at an undisclosed price. A round that triples in size without repricing is a term-sheet event, not a valuation event, and it should never enter a step-up series.

The call. Expect the reported-versus-structured gap to widen through the fourth quarter, because venture debt is the cheapest way to make a round look larger in a market where headline size still signals momentum. The falsifier is a month in which the three largest rounds all disclose their equity and debt legs separately without being asked. ■

SEBI issued eleven circulars in August, plus a long list of consultation papers. Read one by one they are housekeeping. Read together, most of them answer a single question. Who is allowed to put money into a private fund?

The action with a date on it. On 20 August SEBI issued a circular accepting digitally signed powers of attorney for foreign portfolio investor onboarding.^34,35 The regulator's own reasoning is the useful part: notarisation and apostillisation of a power of attorney takes considerable time, and a digital signature removes that step. This is not a headline reform. But it cuts weeks off the process by which foreign money starts trading, and it binds every custodian and every applicant from its effective date.

The proposal that matters more, and does not have a date yet. On 13 August SEBI sought public comment on a review of the accredited-investor framework. By the regulator's own estimate, one proposal could expand the eligible accredited-investor pool to about four lakh. The existing alternative-fund investor base is roughly one lakh.^26,27 A separate limb would let non-resident Indians, overseas citizens and others resident outside India buy into alternative funds with no minimum at all.

What a general partner does differently. Nothing yet, and that is the point of separating an action from a proposal. A consultation paper is a watch item; it binds nobody. But a manager planning a 2027 domestic vehicle should already be sizing a distribution strategy for a retail-adjacent pool that does not yet exist. The lead time on building one runs longer than the consultation period.

The through-line. One regulator spent August trying to enlarge the buyer base for private funds. The other tightened the balance sheets that lend alongside them. Both bind capital. Neither addresses the thing this issue documents, which is what happens when that capital has to come back out. ■

The month belongs to one of them. NIIF Infrastructure Fund II announced a first close of ₹19,000 crore against a ₹30,000 crore target, backed by a Cabinet decision committing an additional ₹30,000 crore of government money to the platform.^49,11 At roughly US$2 billion, that single first close is larger than every other Indian fundraise the desk logged in the month combined.

Accel closed its ninth India-dedicated fund at about ₹5,258 crore, oversubscribed and completed within weeks.²² Bluehill.VC closed a maiden deeptech fund at ₹400 crore against a ₹350 crore target — also oversubscribed.²⁵ Mirae Asset took a ₹1,125 crore first close toward ₹1,800 crore; Piper Serica ₹300 crore toward ₹800 crore; Aum Ventures ₹225 crore toward ₹750 crore.

Two readings follow, and both matter more than the totals.

But the shape of the money has changed. Six vehicles actually took cash. The largest by an order of magnitude is government-backed infrastructure. Three of the other five are early-stage venture funds under ₹1,200 crore, raised substantially from family offices, wealthy individuals and founders. The mid-market institutional growth fund — the vehicle that buys the ₹500-to-2,000 crore positions — is conspicuously absent from the month's closes.

And the thread back to the cover. Every rupee closed in August acquires a distribution obligation somewhere around 2029 to 2033. The route that carried this month's realisations — the exchange block — reaches listed, liquid, recently floated companies. It does not reach an infrastructure asset, an unlisted deeptech company or a growth-stage software business, which is what the month's new money is being raised to buy. Managers taking commitments now are underwriting an exit channel that does not currently exist for the assets they intend to own.

The local call. Deployment from these vehicles will show up from the second quarter of 2027, and the desk expects the gap between fundraising momentum and completed exits to widen before it narrows. The falsifier is a mid-market institutional growth fund above ₹2,000 crore reaching a first close before 31 March 2027. That would say the middle of the market is being funded again, and that this page read a one-month absence as a structure. ■

Consider the position of an allocator asked to re-up this month.

The pitch is strong and mostly true. India's fundraising is at record levels on every published series. A ninth-vintage venture manager closed an oversubscribed fund in weeks.²² The exit tape just printed its loudest month of the year. On the numbers most commonly put in front of an investment committee, the market is functioning.

The problem is that one of those numbers has changed meaning without changing definition.

A distribution model built five years ago assumed realisations came from strategic sales and public offerings — events that end a position, arrive lumpily, and can be forecast as a probability of occurrence times a size. August's realisations were 36 exchange blocks out of 37, and only three closed anything.⁵⁵ That is a different statistical object. It arrives smoothly, in small pieces, and it is bounded by daily traded liquidity in a named stock rather than by the existence of a buyer.

What to actually ask for. Two things, neither of which is in a standard quarterly report.

First, realised proceeds split between positions closed and positions trimmed. A manager returning ₹500 crore from four completed exits is in a different situation from one returning ₹500 crore by shaving six listed holdings, and the reports show both as distributions.

Second, for every listed residual, the traded volume of the underlying stock. A position worth ₹2,000 crore in a name that trades ₹40 crore a day is not a ₹2,000 crore position in any operationally meaningful sense.

The uncomfortable arithmetic. If a 2017-vintage fund's remaining book is concentrated in unlisted companies in sectors that produced no realisations at all last month, then its distribution forecast rests on an event class that had a zero in August. That is not a reason to decline a re-up with a manager whose judgement you trust. It is a reason to stop treating the distribution schedule as a schedule. ■

In August, Accel closed its ninth India-dedicated fund at about US$550 million — roughly ₹5,258 crore — oversubscribed, and within weeks of opening.^22,23 It is the firm's ninth vehicle for this market, which by itself says more than any single close: almost nobody else in Indian venture has raised nine times.

The track record is genuinely strong on the metric that raises funds, and unproven on the metric that matters now. Marks and paper multiples on the 2015-2021 vintages are not the issue. The portfolio has produced several of India's most valuable private companies. The firm has also been an active seller into recent listings, appearing more than once in the month's block tape.⁵⁵

The question is the one this whole issue is about. It is asked without insinuation, because the firm's answer may well be good. What proportion of committed capital has actually been returned, per vintage? And how much of it came from positions the firm has finished with?

What the firm is actually being asked to repeat. Nine India funds is a franchise, and a franchise is a claim about the next decade rather than the last one. Accel India's model has been first cheques into companies that did not yet have a category, held through several rounds, with follow-on capital reserved rather than recycled. That model produced the portfolio; it also concentrates the return in a small number of positions that have to be exited whole. The market described in this issue is one where exiting whole is the thing that stopped happening.

The desk's view. A firm that can close a ninth fund in weeks has demonstrated something real about its franchise. It has not, by that act, demonstrated anything about liquidity, and the two get conflated constantly in coverage of Indian venture. On the evidence of this month — three completed exits nationwide — the market Accel deploys into is one where returning capital is materially harder than raising it. That is a statement about the market, not about the manager.

What would change the desk's reading. A published per-vintage distribution figure from any of India's top-five venture managers. It would be the single most useful disclosure anybody in this market could make, and the first firm to make it will reset the standard for everyone else. ■

The structural break everyone expects in Indian private capital is the arrival of a domestic limited-partner base: family offices, insurers, pension money and corporate treasuries replacing foreign capital as the marginal commitment. August tested that story and gave a split answer.

On the headline, domestic money won decisively. NIIF Infrastructure Fund II's ₹19,000 crore first close rests on a government platform, and the Union Cabinet approved an additional ₹30,000 crore commitment to it.^49,11 That is the state, not the savings pool.

Beneath it, the private domestic base is visible but small. Of the month's other five vehicles that took money, three list family offices, high-net-worth individuals and founders among their investors, and their combined intake is under ₹1,700 crore. Real, growing, and roughly a twelfth of the government's single cheque.

And the largest single named third-party commitment of the month was foreign. CPP Investments committed up to ₹2,070 crore to the NIIF vehicle — a Canadian pension fund, writing into an Indian government platform.⁵³ The share of the vehicle is what makes it legible: against a ₹30,000 crore target, that commitment is about 6.9%, and against the ₹19,000 crore first close it is about 10.9%.

The thread back to the cover. Domestic money is patient in a way foreign money is not, and that matters enormously for a market whose exits take longer than its funds' stated lives. But infrastructure capital raised from a government platform exits through asset monetisation and infrastructure trusts, not through the block-deal window that carried every one of August's realisations. The new domestic money and the working exit route are not connected to each other, and nothing this month suggested they are about to be. ■

The ten largest transactions were 39.5% of all recorded value in the month. Lenskart alone carried ₹4,744.65 crore across two sponsors in five sessions — SoftBank's Vision Fund II on 24 August and Alpha Wave on 28 August.^36,44 One recently listed eyewear retailer was about an eighth of the entire national exit market.

That concentration has a specific cause and it is not coincidence. The block window opens where a domestic institutional bid already exists at size, and that bid exists for a short list of large, liquid, recently listed consumer and financial names. Everything else in India's private-capital book is either unlisted or too thinly traded to absorb a sponsor's position.

The buyer side explains why. Domestic institutions bought a net ₹94,829 crore of Indian equity in August, more than three times what foreign investors put in. That is the bid absorbing the month's blocks, and it is a bid with its own mandate constraints — index weights, sector caps, liquidity rules — which is precisely why it exists for some names and not others. A sponsor does not choose whether the window is open for its holding. The composition of somebody else's portfolio does.

One firm shows what the queue actually looks like from inside. Elevation Capital has realised close to ₹6,000 crore from listed Indian technology over fifteen months, through a mix of offers for sale, blocks and bulk deals. It still holds more than ₹20,000 crore across Meesho, Paytm and Urban Company. That is the clearest public picture anyone has of an Indian venture manager's distribution pipeline. The ratio in it is this issue's argument in one firm. For every rupee returned, more than three are still on the books — in listed names, waiting for the same bid as everyone else.

Concentration is not the same as depth. Ten transactions were 39.5% of everything the desk recorded, and five names carried most of it. A market whose monthly aggregate turns on a handful of order books looks liquid for exactly as long as those books stay open. The month they do not, it looks broken — with nothing about the underlying companies having changed.

The local call. The desk expects September and October realisations to fall well below August's ₹37,201 crore, not because sentiment turns but because the specific names that could absorb a block have now been trimmed. A seller cannot go back to the same book twice in eight weeks without moving the price against itself. Watch whether the next month's realised value comes from new names or from second helpings of the same five. If it is the same five, the constraint is the bid rather than the willingness. ■

Eighteen companies listed on the mainboard. Sixteen of them closed their first day above the issue price. The median first-day move was 25.1% and the mean 27.8%, with Tempsens Instruments up 95.6% and Behari Lal Engineering up 76.2%. Two listings fell: LEAP India by 8.9% and Horizon Industrial Parks by 3.1%.

This is the friendliest primary market India has offered private capital in years, and it makes the rest of this issue harder rather than easier to argue. A hot listing market is precisely the condition under which a sponsor should be able to leave.

Several did sell into it, and the desk's first sweep missed them. Manipal Health Enterprises raised ₹9,277 crore on 5 August with TPG and Temasek vehicles and Novo Holdings selling. Dhoot Transmission listed on 17 August with Bain Capital among the selling shareholders. Shiprocket's ₹2,342 crore issue carried ₹1,242 crore of offer-for-sale, with Temasek, Tribe Capital and Info Edge among the holders. LEAP India's ₹2,400 crore issue was roughly ₹2,000 crore secondary.

But an offer-for-sale is a partial exit by construction. A selling shareholder in an initial offering sells a portion and accepts a lock-in on the rest. TPG and Temasek did not leave Manipal on 5 August; they sold a slice at a price they liked and kept the balance. The route differs from a block in mechanism and in pricing, and not at all in the thing this issue measures — whether the position closed.

Elsewhere in public capital markets, eight issuers raised ₹8,250 crore through qualified institutional placements. No infrastructure or real-estate investment trust issued units at all.

The local call. The desk expects the September and October listing cohort to price more aggressively — a 25% median pop is money left on the table, and bankers adjust. Watch whether the offer-for-sale component of issue sizes rises as they do. If sponsors start taking more of each issue as secondary, the primary market is being used as an exit rather than as a financing, and that is a different market from the one August ran. ■

Foreign portfolio investors bought a net ₹29,631 crore of Indian equity in August, the largest monthly inflow of 2026. They sold ₹1,457 crore of debt, leaving ₹25,492 crore on the combined line.⁵²

Read that against the year and it stops looking like a recovery. Calendar 2026 to 31 August still shows ₹2.24 lakh crore of net foreign equity selling. Two good months — July's ₹20,200 crore and August's — have marked a turn without repairing anything.

The number that matters to this issue is the other one. Domestic institutions bought ₹94,829 crore of equity in August, more than three times the foreign figure. That is the bid that absorbed ₹98,353 crore of block deals in the same weeks, and it is why the exit window was open at all.

That has a consequence the cover story spends four pages on. A bid made of mutual funds, insurers and the pension trust is mandated money: it buys within index weights, sector caps and liquidity rules. It is deep where those rules point it and absent where they do not. A sponsor holding a large, liquid, recently listed consumer name is selling into it. A sponsor holding an unlisted enterprise-software company is not, at any price.

Where the foreign money landed is only half visible. The depository's sector cut for the first fortnight puts financial services top at ₹6,535 crore and telecom bottom at −₹3,322 crore; the second fortnight was unpublished when this issue closed, so the month's full sector map does not exist yet.

Direct investment ran the other way, quietly. The Reserve Bank's balance-of-payments release for the June quarter puts net foreign direct investment at US$6.1 billion, against US$5.2 billion a year earlier. Separately, the easing of the land-border regime brought ₹4,896 crore across 29 projects by 20 August — small in absolute terms, and the first evidence that the change moves money at all.

The local call. The desk expects domestic institutional buying to stay the marginal bid through the fourth quarter, and expects that to keep the block window open for large listed names and shut for everything else. The falsifier is a month in which foreign equity buying exceeds domestic on the net line; that would say the bid has broadened, and with it the range of positions a sponsor can actually sell. ■

It returned six publicly announced facilities, five of them sized, totalling ₹2,100 crore. Inox Clean Energy took ₹1,500 crore of convertible debentures from Motilal Oswal Alternates on 13 August. Kolors Healthcare took ₹215 crore from Piramal Alternatives. Lloyds Engineering Works took ₹200 crore of acquisition financing from Tata Capital, secured on a pledge of shares. Emerald Leisures took ₹105 crore of secured debentures from Lighthouse Canton, and Mintoak Innovations ₹80 crore from BlackSoil, also acquisition finance. River Mobility's debt leg was never sized.^19,30

Not one of them disclosed a coupon or a tenor — zero of twelve price-and-tenor fields. That is why the Panel's private-credit spread series prints an absence rather than a number. A spread over a tenor-matched government bond cannot be computed when neither the price nor the tenor exists.

Set that against the market's own run-rate. The half-year counted about US$3.5 billion across 102 transactions above US$10 million — roughly seventeen deals a month, at a threshold sitting just below this beat's own bar.³³ So August's public record shows roughly one deal in three, and somewhere between one rupee in ten and two in five — a range rather than a point, because the two series do not share a threshold exactly. The majority of the month's private credit by value was still never announced at all.

A rule change explains the shape of the month. From 1 April 2026 the Reserve Bank has allowed banks to fund up to 75% of an independently assessed acquisition value. The buyer must put in at least a quarter, consolidated leverage is capped at three times, and the acquirer needs a net worth above ₹500 crore and three profitable years. Both of August's verified facilities are acquisition finance. That is the competitive pressure arriving in real time: the business private credit was built to do, banks may now do more cheaply, and the beat's two visible deals sit exactly on the contested ground.

What a price looks like when one is disclosed. The month's own facilities gave none, so the range has to be bracketed from either side. Shapoorji Pallonji's Goswami Infratech refinancing closed on 20 July at a three-year rupee zero-coupon yielding 18.95%, with a US$650 million tranche at 14.5%. That is roughly 1,270 basis points over the matched government tenor. It is also a distressed-adjacent holdco refinancing on pledged stock, so read it as the top of the range. At the other end, the only genuinely private-credit-shaped rupee coupons printed inside the window were 13% and 14%, both below the census bar, or about 660 to 780 basis points. The workable band for the period is 600 to 1,300 basis points, and the desk prints the band rather than a point because that is what the evidence supports.

The local call. Spreads that look unchanged in coupon terms have quietly compressed in spread terms across August, because the risk-free base underneath a three-year facility moved and the quoted rates did not. If the desk is right that most facilities reprice with a lag, the compression shows up as wider quoted coupons in the fourth quarter rather than as anything visible now. The falsifier is simple: a disclosed August-vintage facility priced inside 600 basis points over its matched tenor would break the reading. ■

The usual complaint about private marks is that nobody can check them. For twenty-eight Indian positions last month, that stopped being true.

A partial block is a public price struck by the holder itself. When SoftBank's Vision Fund II sold ₹2,887.87 crore of Lenskart on 24 August, it did not only realise cash — it established, at its own hand, what the shares it kept were worth that day.³⁶ When Alpha Wave sold ₹1,856.78 crore of the same company four sessions later, it did it again.⁴⁴

This is the cleanest mark-discipline event India's private-capital book has had, and it is a by-product rather than a policy. Twenty-eight sponsors, ₹28,685 crore of partial block sales, and in every case a residual position that now has an observable reference price attached to it by the seller's own transaction.

The consequence is uncomfortable, and it runs the other way. The same logic says nothing at all about the unlisted book, which is most of it. August's realisations came from consumer, healthcare, fintech and pharma — sectors with recent listings. Enterprise software, agritech, climate and most of deeptech returned nothing above the census bar, and their marks remain exactly as checkable as they were in July, which is to say not at all.

And the residual is the part nobody has priced. Twenty-eight partial sales left twenty-eight positions still on the books. The sold slice has a printed price; the remainder has the same price attached to it by implication, and that implication only holds while the bid does. A holder that sold 2.6% of a 9.9% position at ₹1,480 is carrying the other 7.3% at ₹1,480. That mark was set by a trade which, by its own logic, could not have absorbed seven times the volume.

There is a second reading, and it is less comfortable. A partial block is also a signal about what the seller thinks it can get. A manager that sells a tenth of a position rather than a third is telling you what depth it found when it asked. Those enquiries are not public. The size of the printed trade is their only trace.

The local call. Expect the next round of reported net asset values to show unusually tight agreement with public comparables in the listed names, and unchanged carrying values everywhere else. That divergence is not evidence of discipline. It is evidence of where a price happened to be available, and a reader should discount any headline about improved mark accuracy accordingly. ■

India's data centres and allied sectors are reported to have attracted US$33.3 billion across 25 deals in the first half of 2026.¹² In the same half, all Indian private equity and venture capital — every buyout, every growth cheque, every seed round, across 604 transactions — came to US$20.5 billion.

One sector, on twenty-five transactions, apparently took sixty per cent more capital than the entire private-capital market it belongs to.

The two series are not counting the same thing, and the difference is one word. The private-capital series counts transactions that closed. The data-centre series counts, in its own description, commitments and investments. A commitment is an intention with a press release attached, and one of them is almost the whole number. Blackstone-backed AirTrunk announced a US$30 billion investment programme in India. That is a multi-year capital-expenditure pledge, not a transaction that settled in the half.^13,8

Strip it and roughly US$3.3 billion of actual deal value sits behind the twenty-five transactions. That is a real, substantial, fast-growing sector. It is not larger than Indian private capital, and the sentence saying it is has been repeated across the market all year.

The unit economics follow from that, and they are not software economics. A data centre is a long-lived physical asset, and three things set its returns. The cost and reliability of power. The term and credit quality of the contracts underneath it. And the cost of the capital that built it. Nothing about it compounds the way a venture portfolio does. It is a spread business, and the spread is between a contracted revenue stream and a funding cost that August moved: the five-year government yield rose 15.9 basis points inside the month, and every rupee of project debt in this sector prices off that curve.

The bear case, which the desk finds credible. Three things have to hold at once for the announced pledges to become deployed capital.

Power has to arrive on schedule, at a price, in the states where the land is. Contracted demand has to be signed by counterparties who will still exist in a decade, which in practice means a small number of hyperscale customers with concentrated bargaining power. And the funding cost has to stay inside the contracted spread for the life of an asset measured in decades rather than years.

A shortfall in any one converts a commitment into a smaller commitment. That is the ordinary way these announcements resolve, and it is why the desk treats a pledge and a deal as different objects rather than adding them together.

And the exit math is this issue's own argument, in its sharpest form. A data centre platform is the hardest thing in Indian private capital to sell through the route that carried August. It is unlisted. It is enormous. Its natural buyers are a short list of global infrastructure funds and sovereign vehicles, which is a negotiated market rather than an exchange one. The block window that returned ₹28,685 crore of partial realisations last month cannot touch it at any price.

So the sector taking the largest share of India's announced capital is the sector with the least access to the country's only functioning exit channel. The routes that would serve it — a strategic sale, a sponsor-to-sponsor transaction, an infrastructure trust listing — recorded, between them, nothing above the desk's bar in August.

What would change the reading. A single closed, priced, disclosed data-centre transaction above US$1 billion in the second half of 2026 would establish that the negotiated market for these assets functions at scale in India. An infrastructure trust listing carrying data-centre assets would do it more decisively. Neither happened in August, and until one does, the sector's headline number measures ambition rather than capital. ■

The month's evidence points one way. India can raise capital and deploy it; what it cannot reliably do is finish. So the desk's positioning is about liquidity structure rather than sector selection.

Where we would deploy. Into managers whose remaining book is listed or listable within thirty-six months, and into the secondary and continuation infrastructure that does not yet exist here. Into private credit, where the base rate has moved and the quoted coupons have not caught up. And into domestic mid-market growth, which is the part of the fundraising market that emptied out in August and is therefore the part with least competition for assets.

Where we would not. Into any 2016–2019 vintage extension whose distribution plan rests on strategic sales, without a named buyer. Into unlisted assets in sectors that produced no realisations at all last month. And into a fund whose reported distributions are not split between positions closed and positions trimmed.

Every invalidation above is stated inclusive, on a named series, at a named close. That is a house rule adopted this issue, and it exists because last month's Trade 5 dodged its own stop by one basis point on a threshold written as "below" rather than "at or below". The Scorecard sets out what that cost.

Five risks, most dangerous first.

HIGH, thesis-killer — the trimming is deliberate and it works. If managers are selectively trimming listed winners while holding the compounders, the desk has mistaken discipline for distress, and the completed-exit share will rise on its own as positions run down. This is the strongest argument against everything in this issue and it is given a full page in Against the Grain.

HIGH — the perimeter defeats the measurement. The desk's ledger carries a REIT unit sale and infrastructure trust stock that a pure-play series would exclude. If the correct perimeter is narrower, several ratios in this issue move, and the direction is not obvious.

MEDIUM — the block market keeps absorbing. August took ₹98,353 crore of blocks market-wide without breaking. If domestic institutional demand keeps growing at its recent pace, the conditionality argument weakens every quarter, because the tape's capacity to absorb stops being the binding constraint.

WATCH — the September and October tape. A strong quarter in midcaps would reopen the block window for exactly the names that were trimmed in August, and a second helping from the same order books would make the desk's Call 3 look early. ■

Eight named, resolvable propositions. Each carries a band rather than a point, a basis the desk is reasoning from, and a date by which it settles. All eight enter the calls file and are graded in a future issue whether or not they flatter the desk.

On market anchors. House standard is to print a liquid market's price beside the desk's band where one exists. For the rate row, the overnight index swap curve is the natural anchor and moved materially after the 19 August minutes.¹⁶ For the other seven there is no traded instrument on the proposition, and the desk prints no anchor rather than an unverified one — which is the same choice the prior issue made and the same reason.

Resolution rules. A month means a calendar month, measured on the desk's own ledger at the census bar, published in the issue covering it. "Publicly announced" means a filing, an exchange disclosure or a party's own release — not a report of one. Any row unresolved at its date is graded as not having occurred. Every threshold above is inclusive.

The two bands the desk is least comfortable with. The continuation-vehicle row at 25–35% is a forecast about a market structure forming, and structures form slowly until they do not; the desk would not argue hard with anyone at 15% or 45%. The realised-value row at 55–65% is the closest thing here to a pure liquidity call. It assumes the same names cannot be sold twice in a quarter. That is an assumption about depth, not willingness. ■

F2 — hyperscaler order-book coverage returns above 4× with no fall in long-end yields. Fired, and then kept firing. August delivered a record month of US investment-grade supply, the most issuers since January, and the long end rose rather than fell.²¹

F4 — credit never confirms; high-yield spreads stay inside 300bp. Fired, and not marginally. The ICE BofA US high-yield option-adjusted spread went from 285bp on 31 July to a low of 260bp on 28 August, closing the month at 263bp.⁵⁰ Twenty-two basis points tighter. It never came within fifteen of the trigger.

The corroborating witness is the one the prior issue built itself around. From the Jackson Hole podium on 28 August, the Chairman said credit spreads on corporate bonds and leveraged loans were near the low ends of their historical ranges, and that issuance had been quite strong.⁴⁷

The miss inside the miss. No. 003 retired its own thesis in a stop-press on 6 August, on one bond deal's peak demand, and wrote three paragraphs apologising for the ambiguity in its own threshold. It did not need to. The clean refutation — record supply, absorbed — arrived over the following three weeks. Conceding on the weaker evidence when the stronger evidence is three weeks out is its own error, and it is the one this desk most wants back.

Trade 5 is a coin-edge and the desk rules against itself. The stated invalidation was "high-yield spreads below 260bp with coverage restored above 4×". Coverage was restored on 6 August. The spread printed exactly 260bp on 28 August, and not below it. On the letter, the invalidation did not fire. On the substance, a trade that entered at 284bp and ground 25bp against itself for four weeks was dead. It grades as a miss.

What resolved for the desk. One odds row matured inside the window and it landed. The RBI held the repo rate at 5.25% on 5 August, unanimously, at a stated band of 75–85%. The desk also marked down the probability of a pulled jumbo bond deal at publication, from 20–30% to 10–15%, and August vindicated the mark-down.

The two bands that should move. The September rate-hike row was set at 15–25% before the Chair described inflation as accelerating and named prices as the predominant focus.⁴⁷ Payrolls fell 23,000 in July and core inflation cooled to 2.5%, and the long end still made a nineteen-year high.^20,24 A labour market that is contracting could not pull yields down. That band is too low. The high-yield 350bp row, at 20–30%, is too high against a spread moving the other way. ■

Take the number the whole month quoted. India's private-capital fundraising in the first half of 2026 was reported at US$21.2 billion across 48 vehicles, roughly double the prior year, and the year-to-date figure passed US$23.7 billion — a record on any published series. It was read everywhere as proof that global capital had chosen India.

The single largest component of that total is Bain Capital's Asia Fund VI, at US$10.5 billion. It is not an India fund. It closed on 17 May 2026 and it deploys across Australia, Japan, India, South Korea and China. Of the headline, roughly US$9.1 billion is external commitment; the balance is the manager's own partners and employees, who are collectively the fund's largest investor. India is one market of five, and the fund publishes no country allocation.

So India-attributable first-half fundraising runs somewhere between about US$10.7 billion and the full US$21.2 billion, and nobody outside Bain Capital can narrow it. The desk publishes the range and the reason rather than a split it cannot source.

That has a consequence the cover has to absorb. The ratio of capital committed to capital returned is the cleanest way to state an exit problem, and on the reported figures it looks catastrophic — a market taking in more than twice what it gives back. On the desk's own ledgers, computing closes against realisations for August alone, the ratio is 0.76. India returned more than it committed last month.

Which is the better argument against this issue than anything else on this page. If August returned ₹1.32 for every rupee committed, the exit machine is not jammed. It is working.

The desk's reply is that the ratio measures cash, and cash is exactly what a partial block produces without closing anything. A market can post a flattering committed-per-realised number for years while the positions underneath it age, because trimming generates the numerator without touching the problem. That is why this issue leads on 5.3% rather than on 0.76.

The reply is not a refutation and the desk will not pretend otherwise. If the 2016–2019 vintages clear over four quarters at anything like August's pace, this page was right and the cover was early. Both numbers are in the Panel and the series will settle it.

If this desk is wrong about August, it will most likely be because the market it measured was not the market it thought it was measuring. ■

This page is a device. The questions are the desk's own, put as sharply as the desk can manage against its own argument. Nobody is being quoted and no outside person is involved.

Skeptic — You have written forty pages on a month when Indian funds returned ₹37,201 crore in cash. That is a good month. What exactly is the complaint?

The complaint is composition. Cash returned and positions closed have separated, and only one of them shortens a queue. Three sponsors finished with a company in August, out of forty-three realisations. The rest sold part of something and remain exposed to the same asset, the same tape and the same eventual need to get out.

Skeptic — Trimming a listed winner into institutional demand is textbook portfolio management. You are describing competence as pathology.

That is the best objection on this page and the desk concedes most of it. Selling into strength is correct. What is not competence is how few alternatives there were. No strategic sale to a corporate buyer, no continuation vehicle, and exactly one sponsor-to-sponsor deal whose size nobody disclosed — in a month with record blocks and the busiest listing calendar of the year. A manager choosing to trim is disciplined. A market where trimming is the only available action is constrained, and the two look identical in the distribution report.

Skeptic — Then your real claim is about the unlisted book, not the listed one.

It is, and the Sector Map says so. Everything that realised value in August was already listed, or was listing that month. Enterprise software, agritech, climate and most of deeptech returned nothing above the census bar, and those are where the 2016–2019 vintages sit. The listed book is fine. The problem is that the route which works for it does not exist for anything else.

Skeptic — Your own Panel says India returned more than it committed last month. Series 13, 0.76. Does that not refute you?

It refutes the crude version, and the desk prints it rather than bury it. A market can post a flattering committed-per-realised ratio for years while its positions age, because trimming generates cash without closing anything. But if that ratio holds below 1.0 for four consecutive quarters, the desk was wrong and the Panel will show it.

Skeptic — One month. You are building a regime out of thirty-one days.

Fair, and the second point the desk concedes. August is one observation. What makes it more than an anecdote is that the half-year data points the same way on two independent perimeters, and that the house index family has realisation falling while formation rises. But a single month cannot establish a regime, and the falsifiers in the commission are written to settle it over four quarters rather than to be argued about now.

Skeptic — You keep saying the exchange is procyclical, but August's tape was down and the blocks still cleared. Does that not settle it against you?

Partly, and it is the sharpest version of the objection. The Nifty fell 1.24% and ₹98,353 crore of blocks cleared anyway. What the month actually shows is a split: large caps fell and mid, small and microcaps rose monotonically as capitalisation dropped. The blocks cleared where the tape was strong. The desk's claim is not that a falling index shuts the window everywhere at once — it is that the window is open per capitalisation band, and the largest positions sit in the band that fell.

Skeptic — What would actually change your mind?

Two continuation vehicles above ₹500 crore, or a month where completed exits clear 15% of realised value. Either would say the plumbing is being built. Both are in The Odds with a band and a date. ■

5–7 October · RBI Monetary Policy Committee. The August minutes read hawkish and the market moved from pricing a pause to pricing an increase.¹⁶ A hold with an unchanged stance leaves the private-credit base where August left it. An increase repices every floating facility written this year and widens the gap between quoted coupons and matched-tenor government yields, which is the desk's Call 4.

Late September · EY-IVCA August roundup. The first independent aggregate for the month this issue has measured from the bottom up. The desk's ₹37,201 crore is a floor built from 43 verified rows; the published figure will use a different perimeter and a different bar. Where the two diverge, the desk will print both and explain the gap rather than quietly adopt the larger number.

30 September · the September block tape. The most informative event of the quarter, and the cleanest test of this issue's argument. If September's realised value comes from names that did not appear in August, the block window is deeper than the desk thinks. If it comes from second helpings of Lenskart, Groww, Meesho and the same handful of order books, then depth is the binding constraint and Call 3 is on track.

Through the fourth quarter · SEBI's accredited-investor consultation. Comment closed on a proposal that could take the eligible pool from roughly one lakh to four.²⁶ Adoption substantially as proposed is in The Odds at 50–60%. It is the fastest available route to a domestic buyer base for private secondaries, which is the structure India most conspicuously lacks.

31 December · the year's fundraising total. India's published figure will set another record and will again include pan-Asia vehicles. Part IV shows what that does to the number. The desk will publish its own India-dedicated total alongside it, computed the same way each month, and the two will not agree.

Through 30 June 2027 · the continuation-vehicle question. The single structural gap this issue identifies. Not one printed in August; global volume runs near 14% of secondaries. If India's first one above ₹500 crore appears, it will be reported as a deal. It should be read as infrastructure.

The page prints the ten largest. The published file carries every logged transaction at or above each beat's census bar, with date, parties, beat, sector, size, structure and source. ■

Months in market runs from the vehicle's SEBI registration date to 31 August 2026, applied uniformly. A blank in the raised column means the manager is publicly raising and has disclosed no figure — the honest state of the record rather than a gap the desk could have closed.

Read rows eight and sixteen together. Gaja Capital is raising a ₹1,250 crore secondaries fund and White Whale Ventures a ₹250 crore one. This issue argues at length that India lacks the structure that lets a position change hands without a strategic buyer. Two managers are raising capital to build exactly that, and neither had closed by month end. It is the most encouraging fact in the issue and the desk prints it against its own thesis. ■

The separation is the point. A circular with an effective date changes what a general partner or a lender may do from that date. A consultation paper changes nothing until it is adopted, and several of August's are consequential enough that treating them as decided would be a serious error. Both are tracked; only the first is counted in the coverage gate. ■

Democratizing Private Markets: Equilibrium Predictions · NBER working paper⁴³

The argument. If retail and near-retail capital is admitted to private markets at scale, the effects are not confined to who owns the assets. Expected returns, fee structures and the mix of what gets funded all move, because the marginal buyer changes.

The flaw. It is an equilibrium model, and equilibrium arrives on its own schedule. It has less to say about the transition, which is the part a regulator writing a rule in 2026 actually has to manage — and the transition is where the accidents happen.

The usable takeaway. If SEBI's accredited-investor expansion is adopted substantially as proposed, the first-order effect on Indian managers is not a bigger pool at the same terms. It is pressure on terms. A manager modelling a 2028 domestic vehicle at today's fee load is modelling the wrong vehicle.

The Fragility of Semi-Liquid Private Credit Funds · NBER working paper¹⁰

The argument. Funds offering periodic liquidity against illiquid credit assets carry a structural mismatch. The redemption gate is the mechanism that makes them work and also the mechanism that makes them fail, because the option to redeem is worth most exactly when everybody wants it.

The flaw. The evidence base is overwhelmingly American, and the US market has both a deeper secondary bid and a different investor mix. Transplanting the conclusion to India requires an assumption the paper does not make.

The usable takeaway. India's private-credit vehicles are mostly closed-ended today, so the fragility described here is not yet the local problem. It becomes one the moment a semi-liquid structure is marketed to a widened accredited pool — which is precisely the combination August's consultation would permit.

The verdict. Take the first for what it says about fees and the second for what it says about structure. Neither is about India, and that is the point worth making to a regulator: the empirical work on democratising private markets is being done elsewhere, on other people's data, while the rule is being drafted here. ■

The Lexicon accretes. A term defined here is used in the issues that follow, not defined again.

Trim. A sale that reduces a position without closing it. August produced 37 realisations above the desk's bar. Thirty-six crossed the exchange as blocks, and only three closed anything, which leaves thirty-four trims.⁵⁵ The word earns its place because a distribution model built on exits is forecasting a different event from the one that actually happened.

Block. A large trade negotiated away from the screen and crossed on the exchange in a dedicated window, at or near the prevailing price. August's block volume passed ₹1 lakh crore.⁵² A block is a route rather than a valuation event. It tells you a holder sold. It does not tell you what the asset is worth to a buyer who wanted all of it.

Sponsor-to-sponsor. One financial sponsor selling to another. It is the exit that keeps an asset inside private capital instead of handing it to a strategic or to the public market. This issue reads it as zero by value, because the single transaction that qualified went unpriced.

Creeping acquisition. Stake-building by an acquirer already large enough to disclose, inside the headroom the takeover code allows across a financial year, so no open offer follows. The desk uses the term as a reading of observed holdings against the code. It is never a filing that says so, and this issue tiers it accordingly.

Residual. What is left on the books after a partial sale. Every trim creates one. A residual in a thinly traded listed name is not the same asset as the mark carried against it, which is the question the Marks puts to a valuation committee.

Record bar · census bar. The desk's two thresholds. The record bar is what the desk names in print. The census bar is what it counts into the Panel. They are set apart deliberately, so that a month's narrative and a month's statistics cannot be tuned by one decision.

Nil. A zero the desk has sourced. A source that refused a fetcher is not a nil, and the distinction is the whole difference between a finding and a gap. ■

None this month.

What belongs here. A correction is a factual error in a published issue: a wrong figure, a transaction attributed to the wrong party, a misstated date or perimeter. Each one is numbered and dated, and it names the issue and the page it corrects. Corrections are never quietly amended into the record. The published number stands, and the correction sits beside it, because a reader who took a decision on the original is entitled to see what they acted on.

What does not belong here. A call that turned out wrong is not a correction. Calls are graded in the Scorecard, misses first, and a forecast that failed is a forecast that was honestly wrong rather than a fact that was wrong. A restatement of a Panel series is not a correction either. That publishes as a dated methodology action carrying a version bump, and the superseded figures stay readable in the standing series.

Why the heading prints empty. A corrections column that appears only when there is something to confess is a column a reader cannot trust, because its absence carries no information. This one runs every month. An empty month is a claim the desk is making, and it is gradeable like any other. ■