Analysis only. This version carries the sector reading and the levels at which it would fail. It states no view on any security and recommends no action. Figures, sources and the documented defects are unchanged from the full document.
Sector Research
Everything here is counted. One company publishes the step that matters.
Aurobindo has 728 US approvals standing behind 888 filings. Fortis has 3,418 occupied beds and keeps opening more. Both count things that exist, not money that arrived. Laurus Labs publishes the step in between and Neuland Laboratories only gestures at it, and that step — not the size of anybody’s pipeline — is what the reading below turns on, on one disclosed figure and one impression.
Healthcare As of 12 August 2026
| Figure | What it is |
|---|---|
| 1,881bp | from the latest print to the level at which this reading fails |
| 31.81% | Laurus Labs EBIT margin, Jun 2026, vendor basis, consolidated |
| −14.67% | Syngene revenue growth, Jun 2026, same basis, same quarter |
That is the argument in three numbers, and the first is the weakness rather than the comfort. A retirement level 1,881 basis points below the latest print cannot be crossed by anything a business does in four quarters, so the reading is armed by disclosure and not by a line — and the disclosure runs two names deep, one of which cannot state its own growth rate.
- The two numbers you would reach for first are printed on a basis their peers do not use
- Occupancy here is a capacity reading, not a demand reading — and three operators say so themselves
- Approvals are counted in the hundreds, and the only growth decomposed into price and volume is domestic
- The CDMO revenue already recognised — and the line that would settle this reading is too far away to do the job
Section 1 — The two numbers you would reach for first are printed on a basis their peers do not use
Revenue per occupied bed and network revenue are what a hospital is judged on. In this sector each of them has a large name that reports it on a different basis from everybody else, and neither basis is visible in the number itself.
Fortis Healthcare reports ARPOB per annum. Its hospital business took ARPOB up 3.4% to Rs 2.51 crore per annum in FY26 from Rs 2.42 crore in FY25, and up 2.6% to Rs 2.71 crore per annum in Q1 FY27 after recent acquisitions. Every other operator states the same quantity per occupied bed per day. Rs 2.51 crore a year is about Rs 68,800 a bed-day; Rs 2.71 crore is about Rs 74,200. Converted, Fortis sits just under Medanta and just over Jupiter Life Line — an ordinary metro operator. Unconverted, it shares no scale with the names beside it, and a table that carries both has ranked a reporting convention rather than a business.
Exhibit 1
ARPOB as each company reports it, and on the one basis that compares
Latest print for each name, not all from one quarter — Max Healthcare’s is Q3 FY26 and Yatharth’s FY25 — so the last column compares levels, not a date. The Fortis conversion divides the per-annum figure by the days in a year; the same arithmetic on its FY26 figure returns the Rs 68,800 that print is stated at.
| Name | As reported | Basis | Per occupied bed per day |
|---|---|---|---|
| Max Healthcare | Rs 77,900 | Q3 FY26, per bed-day | Rs 77,900 |
| Medanta | Rs 74,941 | Q1 FY27, per bed-day | Rs 74,941 |
| Fortis Healthcare | Rs 2.71 crore | Q1 FY27, per annum | about Rs 74,200 |
| Jupiter Life Line | Rs 73,500 | Q1 FY27, per bed-day | Rs 73,500 |
| Yatharth | Rs 30,829 | FY25, per bed-day | Rs 30,829 |
| Park Medi World | Rs 30,444 | per bed-day | Rs 30,444 |
| Apollo Hospitals | Rs 178,434 | FY26, average revenue per patient | not comparable |
Apollo is the third denominator and the one that cannot be repaired: its FY26 summary gives 67% occupancy, 628,998 discharges, ALOS of 3.17 days, 7,289 pharmacies and average revenue per patient of Rs 178,434. Converting that to a bed-day figure needs the share of revenue earned by inpatients, which is not disclosed, so it is declined rather than estimated. KIMS shows why the distinction bites: in Q1 FY27 its revenue per operating bed grew 9.7% and its revenue per patient 6.8% — 290 basis points apart, inside one company, in one quarter.
The second trap is in Max Healthcare’s revenue line, and it is worse because both versions are labelled revenue. On a vendor digest, FY26 revenue is Rs 8,536 crore against Rs 7,184 crore in FY25, up 18.8%, at an EBIT margin of 27.6% against 26.9%. The company reports Network Gross Revenue of Rs 10,538 crore for the same year, up 16%. Two measures under similar names, a gap of Rs 2,002 crore, and no agreement on the growth rate either. Operating EBITDA of Rs 2,638 crore is a different margin depending on which denominator it lands on, and neither figure says which one you hold.
Section 2 — Occupancy here is a capacity reading, not a demand reading — and three operators say so themselves
The occupancy prints in this segment run from 47.2% to 74%, and reading that spread as a demand ranking gets it backwards. Three operators attribute their own number to beds they have just opened, which turns a soft figure into a dated, checkable promise. Read the other two the same way and they cut against the reading rather than for it: Max says its occupancy held up despite an 8% capacity increase, which is the same mechanism claimed as strength, and Fortis’s group number did not move at all while its newest units run near 60%.
- Max Healthcare ran network occupancy of 74% in Q3 FY26 against 75% a year earlier and 77% in the trailing quarter, with operational bed capacity up 8% year on year.
- Medanta held 63% in Q1 FY27 on expanded bed capacity, and 66% excluding Noida — occupied bed days up 21%, inpatient volumes up 28%, outpatient up 34%.
- Jupiter Life Line printed 59.6%, which it states as diluted by the expanded base of Dombivli beds; revenue Rs 411 crore, EBITDA Rs 79.3 crore at a 19.3% margin.
- Park Medi World fell to 56% from 68%, attributed to the step-up in capacity over twelve months, while ARPOB rose 12% to Rs 30,444 from Rs 27,221 and ALOS improved 8% to 5.9 days from 6.4.
- Fortis held group occupancy steady at 69% with occupied beds up about 17% to 3,418 from 2,928 — its new Yeshwanthpur, Greater Noida and Manesar units running at about 60% on installed beds, about 187 of them operationalised.
Three companies making the same excuse would be a warning in most sectors. Here it is a prediction, because the same filings show the ramp completing inside the same networks. Medanta’s 66% excluding Noida against 63% including it is the gradient in one operator. Yatharth’s mature facilities ran 66% occupancy and Rs 34,417 ARPOB against a network at 61% and Rs 30,829 in FY25. Fortis has 14 facilities above 20% operating EBITDA carrying 70% of hospital revenue. KIMS reached break-even at Mahadevapura in under seven months and is EBITDA positive there, expects Electronic City within one or two quarters, and puts its Kerala cluster at single-digit EBITDA margins today, mid-teens this year and 20% to 22% in two to three years — the shape of the curve and roughly when to check it.
The waiting is already visible in margin, at the builders. Fortis grew consolidated revenue 17.5% to Rs 2,545 crore in Q1 FY27 and hospital revenue 19% to Rs 2,187 crore, and its hospital operating EBITDA margin still fell to 21.5% from 22.1%, with the consolidated pre-ESOP margin at 22.3% against 22.6%. Narayana Hrudayalaya’s vendor-basis consolidated EBIT margin reads 18.51% in June 2026 against 20.33% in March. Medanta, diluting less, expanded its EBITDA margin to 26.7% from 25.7%, and Max Healthcare its vendor-digest EBIT margin to 27.6% from 26.9%. Growth in the high teens that does not reach margin is the signature of a build phase, not of a pricing problem.
Two names should not be averaged into that claim. Rainbow Childrens Medicare printed 47.2% occupancy in Q3 FY26 with EBITDA of Rs 147 crore, up 9%, and PAT of Rs 73.9 crore, up 7% — a paediatric business whose bed utilisation is not read the same way. Narayana’s clinics carry approximately 30% of the hospital’s outpatient footfalls and its ALOS was 4.3 days, a funnel that does not resolve into an occupancy number at all. And Park Medi World’s 56% is compared by the company to 68% in Q1 FY26, which dates the print to Q1 FY27, while it is also carried as a Q1 FY26 print: the fall of twelve points survives either reading, the date does not.
Section 3 — Approvals are counted in the hundreds, and the only growth decomposed into price and volume is domestic
A filing count is an input. It says what a company has asked permission to sell, not what it sold, and here the counts are not even dated to the same year.
Aurobindo Pharma has filed 888 US ANDAs and 322 DMFs and holds 728 approvals as of March 2026, with 5,023 dossiers outside the US. Lupin’s cumulative filings are 742 against 584 approvals; Alembic’s 274 against 235; Jubilant Pharmova’s 101 US filings against 63 approvals. Strides Pharma Science had more than 230 filed and more than 215 approved as of March 2025, with 73 commercial products; Marksans more than 350 approved ANDAs and market authorisations; Alkem 176 filed and 145 approved as of 31 March 2024, the date its own disclosure carries. Read as a league table those numbers span three financial years.
Two companies disclose the step from approved to selling, and neither ratio is close to one. Senores Pharmaceuticals went from 30 approved ANDAs in June 2025 to about 58 in June 2026, of which 23 are commercialised and 35 are planned. Strides Pharma Science pairs more than 215 approved ANDAs with 73 commercial products in the US. Those ratios are what every other name above is implicitly claiming and does not state. Gland Pharma comes closest by reporting the flow — 24 ANDAs filed, 28 approvals, 31 launches in FY26 — and a launch is still not a rupee.
The counts are also not internally stable. Aurobindo’s FY26 annual report says the company filed 29 ANDAs during the year in one place and 25 in another, alongside 40 approvals including tentative ones and R&D of Rs 1,590 crore at approximately 5% of revenues. Both numbers are printed here because choosing between them would be inventing one. That is a company’s own headline regulatory statistic disagreeing with itself in a single audited document, and it sets the reliability ceiling for every count above it.
Where money actually shows up, it points three ways on three different bases. Sun Pharma’s US business fell 9.7% to USD 427 million in Q1 FY27, which it attributes to Lenalidomide erosion and additional competition in certain products, even as its India share rose to 8.5% from 8.2% on Pharmarack MAT June 2026. Glenmark’s North America revenue was Rs 10,974 million in the same quarter, up 41.1% — and up 19.8% on the core business, net of deferred out-licensing income on ISB 2001 — a headline less than half of which is the core business. Ajanta Pharma’s US generics grew 49% across FY26 on a few successful launches in the previous 15 months.
The domestic book has the same subtraction problem: Torrent’s vendor-basis consolidated revenue growth reads 56.29% in June 2026 while the company describes FY26 base business revenue growing in the mid-teens, the difference being the controlling stake it acquired in J.B. Chemicals. On price itself there is only commentary: Divi’s says volumes remained stable while pricing continues to reflect competitive market conditions across products and geographies, and Dr Reddy’s says its API and North America generics businesses faced increased competitive intensity.
The one growth rate anybody splits is domestic. Mankind Pharma’s secondary sales per IQVIA grew 12.7% in Q1 FY27 led by volume growth of 4.7%, itself up 220 basis points year on year, against overall revenue of Rs 4,031 crore up 13% and an EBITDA margin up 250 basis points to 26.3%. That is the decomposition every US book above needs and none provides, and it exists only because a paid panel measures the Indian market by volume.
And the sector’s own leading indicator is missing entirely. A US exporter’s largest single swing factor is the classification of the plants that serve it, and not one facility classification, warning letter or inspection observation appears anywhere in the evidence assembled for these names. The only inspection outcome disclosed at all is Concord Biotech completing an ANVISA Brazil inspection at its Limbasi facility — a different regulator, one plant, one company. Which is why no directional reading is offered on the export leg below: the counts can be ranked and the outcome cannot be observed, and ranking the input is how a scorecard answers the wrong question confidently.
Section 4 — The CDMO revenue already recognised — and the line that would settle this reading is too far away to do the job
Laurus Labs and Neuland Laboratories
One of them publishes the number the rest of the contract-manufacturing segment describes instead of disclosing — how much of revenue is commercial supply rather than development work — and the other only gestures at it. Laurus Labs’ small-molecule CDMO division sold Rs 835 crore in Q1 FY27, up 69%, and about 55% of that came from commercial supplies. Group revenue was Rs 2,026 crore, up 29%; gross margin held around 62.7% and EBITDA margin expanded 7 percentage points to 31.8%, with the vendor-basis consolidated EBIT margin at 31.81% in June 2026 against 28.73% in March. Neuland Laboratories states that commercial CMS projects contributed a majority share of revenue and were the primary driver of growth, with EBITDA of Rs 231.1 crore at a 35.5% margin and gross margin at 61.2% against 55.3% a year earlier. “A majority share” is an adjective where Laurus gives a figure. An exporter claiming a majority without stating the ratio would not be credited with the disclosure, and neither should a contract manufacturer: this call rests on one disclosed conversion figure and one impression.
The contrast is in the same quarter and on the same vendor basis, but not in the same business, and that gap is this call’s central weakness. Syngene printed revenue growth of −14.67% and an EBIT margin of 13.48%, and Syngene is essentially all research services. The Laurus margin set against it is consolidated: the CDMO division that this call is actually about sold Rs 835 crore of the group’s Rs 2,026 crore, so 41% of the series carries the argument and the other 59% — whatever it consists of, and the composition is not disclosed here — does not. A margin printed on a basis that does not match the business being judged misleads in exactly the way an ARPOB printed per annum does, and the remedy is the same: name the basis, or do not use the number.
Piramal Pharma Solutions describes 500 plus customers, 15 development and manufacturing facilities and critical-care products in more than 6,000 hospitals, surgical centres and veterinary clinics, and publishes no comparable conversion figure at all. Anthem’s disclosure — CRDMO more than 83% of FY26 revenues against nearly 17% for specialty ingredients — sizes a business rather than dating its recognition.
The argument is retired if Laurus Labs’ consolidated EBIT margin prints below 13.0% for two consecutive quarters, the level carried on this sector’s watch table. Three things are wrong with that line, and all three are stated rather than worked around. It sits 1,881bp below the latest print, so nothing a healthy business does in four quarters reaches it. It is bound to the consolidated series, of which the CDMO division is 41% — the line could survive a total collapse of the business this call is about. And two consecutive quarters is a stricter test than the tracked line, which trips on any single quarter below the level; waiting for confirmation makes an already distant line harder still to reach. The observable that would actually change the view is the commercial-supply share Laurus itself put at about 55%, and no tracked line is bound to it. Neuland, half the named subject, carries no tracked line at all. This call is armed by disclosure and not by a line, and the watch table cannot grade it.
Four things argue against this call, all stated rather than discounted. Neuland cannot state its own growth rate. Total income for Q1 FY27 was Rs 650.1 crore against Rs 300.6 crore a year earlier, described in the same sentence as 16.3% growth; the two figures imply 116.3%. Both are printed here because picking one would be inventing it, and a company that publishes an arithmetic contradiction in its headline is a thin second leg. Customer concentration is quantified by nobody. Neuland’s commentary is the only material on it at all, so whether Rs 835 crore rests on one customer is unknown — for a contract manufacturer, the whole risk. And the customers are inside the sector. Neuland names a strategic collaboration with Gland Pharma in sterile APIs; Sanofi names Emcure in its partnership commentary. When a supplier’s customer is three rows up the same list, one end-market event can be recognised twice.
The fourth is stronger than the first three. Those three are objections about what cannot be verified. This one is about how the business itself could deteriorate, and it is reasoning from the disclosed figures rather than a finding any filing states. Commercial supply is supply of a named product to a named customer, so it lasts as long as that product does — which makes Rs 835 crore growing 69% with about 55% already commercial the good news and the exposure in one number. The more of the book that has converted, the more of it is bound to particular products whose end dates nobody discloses. Nothing in the evidence tests this either way, and that is the point. No filing quoted here states a customer share, a molecule count, or a contract end date for any contract manufacturer in the segment, so a book that is 55% commercial and one that is 55% commercial to a single expiring product read identically from outside. The disclosure that would separate them is the same one section 3 says nobody publishes. Nothing in the evidence says Laurus is next, and nothing in it says Laurus is not — the disclosure that would settle it, customer and molecule concentration, is the same one section 3 says nobody publishes.
Exhibit 2
The tracked retirement levels, and how far the latest print sits from each
Vendor-basis EBIT margin including other income, consolidated. Nine of the fifteen tracked lines; the other six are two revenue-growth lines — Lal PathLabs above 25% year on year, Biocon below zero — and four macro lines: USD/INR below 92 or above 98, Brent below 75 or above 100 dollars a barrel. Prints are not all from one quarter: Apollo’s is March 2026, the rest June 2026, so the distances compare as distances and not as a snapshot. One difference from the tracked lines themselves: those trip on a single print by their stated date, while an argument here is retired only on the second consecutive one — the level is the same, the patience is not.
| Name | Latest print | Level | Direction | Distance |
|---|---|---|---|---|
| Lal PathLabs | 33.68% Jun 2026 | above 35.0% | up | 132bp |
| Apollo Hospitals | 15.99% Mar 2026 | below 13.0% | down | 299bp |
| Sun Pharma | 30.81% Jun 2026 | below 27.0% | down | 381bp |
| Mankind Pharma | 27.09% Jun 2026 | below 23.0% | down | 409bp |
| Poly Medicure | 28.56% Jun 2026 | below 24.0% | down | 456bp |
| Biocon | 20.23% Jun 2026 | above 25.5% | up | 527bp |
| Narayana Hrudayalaya | 18.51% Jun 2026 | above 25.0% | up | 649bp |
| Divi’s Laboratories | 41.95% Jun 2026 | below 30.0% | down | 1,195bp |
| Laurus Labs | 31.81% Jun 2026 | below 13.0% | down | 1,881bp |
Read that column before trusting any argument attached to it. Six of the fifteen lines run upward — they are confirmations, and a slate read as one-way misreads it. The nearest line of all is one of those: Lal PathLabs sits 132bp below a level whose crossing would validate rather than kill, having moved from 29.01% in March to 33.68% in June on 18.87% revenue growth. At the other end, Divi’s is 1,195bp above a level that sits just inside the bottom of a thirteen-quarter range running 29.95% to 41.95% — the 30% line sits above that 29.95% low rather than below it, so the series has already been there, and a line described as out of reach has in fact been crossed within the lookback. Two levels also exist in two versions — Divi’s at 30% and at 26%, Biocon at 25.5% and at 28% — and the nearer of each pair is used above, because the further one cannot be reached by anything the business would do. Four segments carry no bound line at all: MNC subsidiaries, single-speciality care, discovery R&D and diversified holdings. Nothing said about those four could be refuted by a print, which is why nothing is said about them.
The macro lines deserve the same scepticism. Brent printed 79.45 on 5 August and 89.22 on 12 August; USD/INR printed 95.40 and then 95.32 over the same window. Levels at 75 and 100 dollars a barrel straddle a series that moved nearly half the way to one of them inside a week, and a four-quarter horizon on a daily series that volatile is a coin toss dressed as a trigger.
What is deliberately not covered, and what would settle it
- Plant-level USFDA status — no facility classification or warning letter for any exporter’s plants, which is the largest single risk to the US book. The FDA’s own inspection-classification and approvals data would settle it: public, structured, free, not held. US generic price erosion — no series at any price, so no US growth print here is split into price and volume. Indian market growth by therapy — IQVIA and AWACS are paid and absent, so every market-share figure here (Sun’s 8.5%, Glenmark 13th at 2.37% with 11 brands in the top 300, Mankind’s 15.2% prescription share) is company-stated and reported as such, never as a series.
- NLEM and DPCO exposure — no dated price-control series, so no view on the domestic branded book’s realisation risk, precisely where Mankind’s 26.3% EBITDA margin and Torrent’s 33.22% vendor-basis EBIT margin would be tested. Hospital payer mix — the cash, insurance and government-scheme split is not disclosed, so the ARPOB gradient in section 1 is described and not underwritten. Valuation and positioning — no multiples and no consensus estimates, so a view on price would be a view on nothing.
One boundary is worth naming, because a reader will otherwise assume it was averaged in. Divi’s earned a 35.0% EBIT margin on the FY26 annual digest against Aarti Drugs at 12.1%, on June 2026 revenue growth of 24.32% and 19.09% — comparable growth, margins nearly three times apart, one segment label. Any single API figure averages two businesses.
Provenance. Written from the same evidence base as the sector’s data set, with its conclusions withheld from the author — so the argument here is an independent read of the evidence rather than a paraphrase of an existing view. Figures computed rather than quoted (the basis-point distances, the Fortis bed-day conversion, the Rs 2,002 crore revenue gap) are declared as computed on the root element.
Sources and conventions. Every figure above is drawn from a maintained evidence base for this sector and is reproducible from it. The latest available print is used for each name, and where prints in one table are from different quarters the caption says so. The basis is labelled on every print — consolidated or standalone, vendor digest or company-reported, per annum or per occupied bed per day or per patient. EBIT margins called vendor basis include other income and are the vendor’s computation; EBITDA margins are the company’s own. Rupee amounts are stated in the unit the reporting company uses, crore or million or billion, and the unit is named each time; where a source gives a figure without naming its currency, no currency is supplied. Two disagreements inside the sources are printed rather than resolved: Aurobindo’s 29 and 25 ANDA filings, and Neuland’s 16.3% and 116.3% growth. Data as of 12 August 2026. The sector has 123 active members. 26 of them were read at length — 21% — and the figures below rest on quoted filings from 41. Those are two different measurements and both are printed because neither alone is honest: 26 is how many companies were read at length, 41 is how many are quoted somewhere in the evidence, and the gap is names reached through a single filing or a structured financial rather than a full read. Two coverage figures exist for this work and they disagree. The reading record counts 26 read at length; a note in the same evidence base says 92 of 123. The 26 is used because it is the figure that enumerates — every unread name is listed by ticker and the count follows from that list — while the 92 is stated without a list and cannot be reconciled with the 18 dropped for budget. The disagreement is unresolved, and a reader should know it exists rather than be handed the lower number as though nothing competed with it. Of the rest, 18 were dropped as the smallest by market capitalisation to fit the reading budget, and KPL, KMC Shil and 3B Blackbio hold no readable characters across their filings and are read as zero rather than as absent. For the remainder no reason is recorded, and inventing one here would be worse than saying so. What follows from that is the only rule that matters here — no count on this page is a statement about the sector. When a section says three operators disclose something, it means three of the companies quoted, not three of 123, and the rest of the sector has not been asked. The levels at which each argument is retired are stated in advance and tracked to their dates. Written without knowledge of any portfolio. Nothing here is a recommendation to buy or sell. Ten known defects are documented rather than repaired, and here is where they are: two in the masthead (the number strip compares a margin against a growth rate under “same basis”; the headline’s count contradicts section 3’s), two in section 1 (“every other operator” is refuted by the next paragraph; the exhibit ranks two hospital tiers in one column), five in section 4 (Syngene’s business is described from outside the evidence; the call names two companies while only one meets the disclosure test it is made on; the seven-point margin expansion names no comparator period; the call carries no calendar date and retires on a stricter test than the tracked line; and one clause is printed twice and misattributed), and one in the boundary list, which declares fewer gaps than the evidence does. Two smaller ones are recorded alongside them. Locating them is the point: a reader who cannot tell a documented defect from an undocumented one has not been told anything. The reasoning that left them standing is recorded alongside them. None of the ten is part of the standard being set — several are failure classes this page argues against elsewhere, which is why they are named rather than quietly mended.