IKS is a high quality business trading at a high price. At 33 times profit before depreciation and 44 times earnings, the easy money has already been made. But there are four specific things that could still produce a return from here, and each one can be tested against evidence over the next four quarters.

• The company has set a public target of Rs 3,000 crore of profit before depreciation by March 2030. If it hits that number and the shares trade at the multiple whole companies in this sector change hands at, the value works out around Rs 2,570 a share. That is 40% above today. The target needs 29% profit growth a year. The company has just delivered 39% a year for the last two years.

• The TruBridge purchase is being valued by the market as a risk. It may turn out to be the opposite. IKS paid 8.2 times profit for a business it has already improved. The profit stayed the same after IKS cut away the loss-making parts, so the margin went up from 20% to 22.7% before any integration work started.

• IKS has done this exact job once before and the numbers are on public record. When it bought AQuity, group margin fell from 35% to 28.6%. It rebuilt it to 34.2% in two years while cutting the client list from 853 to about 600 and reducing headcount. If it repeats that with TruBridge, the market is currently paying nothing for it.

• The company is in the faster growing half of its industry and nobody in India writes about it. Work bought from outside firms by American doctors and hospitals grows at 12.5% a year. The equivalent work bought by insurers grows at 7%. IKS is entirely on the doctor side.

Set against that, three things have to go right, and they are not certain. Growth after stripping out the falling rupee slowed to 12% last quarter, which is the same rate the market grows at. Staff turnover ran between 44% and 54% a year in the three years before listing and the company has not published the number since. And the two services IKS is best at, coding and note writing, are the two most heavily outsourced already, with underlying markets growing at 0.5% and 6% a year.

01 — How the industry is put together

Six layers, and which ones IKS owns

American healthcare administration is built in layers. Which layer a company sits in tells you most of what you need to know about how much money it makes, how hard it is to displace, and how large it can grow.

The electronic health record is the software a hospital runs on. The appointment is booked in it. The doctor writes the note in it. The order for a scan is placed in it. The charge is generated in it and the bill leaves from it. Every other layer either feeds this software or reads from it.

Replacing it is enormously expensive. A mid-sized hospital changing its health record spends tens of millions of dollars and eighteen months, retrains every clinician, and risks losing money during the switch. This is why the health record is the most defensible position in the whole industry.

The work around the record is everything else. Checking that a patient’s insurance is live. Getting the insurer to approve a procedure before it happens. Writing up the clinical note. Turning that note into billing codes. Assembling and sending the bill. Arguing with the insurer when it refuses to pay. Chasing the patient for their share. This is what IKS has sold since 2006.

The distinction matters for a practical reason. If a supplier’s tools sit outside the health record, the doctor or the billing clerk has to leave the software they work in, do something somewhere else, and come back. Every one of those switches is a point where people stop using the tool. A supplier whose work happens inside the record does not have this problem.

02 — The health record layer

The most concentrated software market in American healthcare

This layer explains why IKS spent 557 million dollars.

• Every large hospital group that chose a new health record in 2025 chose Epic. No other supplier won a decision involving more than three multi-specialty hospitals.

• About 30% of Oracle Health customers say the system is not part of their long term plans. Another 35% are considered at risk of leaving.

• Epic now holds records for 325 million patients. Its patient app has more than 200 million registered accounts.

• The number of hospitals choosing a new health record fell about 40% in 2025 against 2024, and about half against 2023. Hospitals are spending their money on artificial intelligence tools instead, which pay back faster.

Two things follow for IKS.

First, the market for doctors’ clinics looks nothing like the hospital market.

A company selling services into clinics has to connect to many different systems and cannot depend on any one of them. This is why IKS has always sold services rather than software into that part of the market, and why its technology has to work with whatever record the client already has.

Second, small rural hospitals are a separate market with their own two-supplier structure, and IKS has just bought into it. TruBridge and MEDITECH are the two systems that serve small community hospitals, where Epic’s cost and complexity cannot be justified. TruBridge is present in more than 40% of hospitals with fewer than 50 beds, and about 98% of its health record customers are hospitals under 100 beds. Management puts its share of the rural health record market above 30%.

This is not a weak software asset. It is the sitting supplier in a part of the market that Epic has chosen not to serve.

03 — The ten steps

What happens between a patient calling and a hospital being paid

Every bill in American healthcare passes through ten steps. Each step is a job someone has to do. Each step loses money. The industry IKS competes in exists because of those losses.

Steps one to three — before the patient arrives

A patient rings up and books a slot. Two problems appear straight away.

• Patients who do not turn up. Between 15% and 30% of booked appointments are missed, depending on the specialty and the type of insurance. An empty slot is pure loss, because the doctor, the room and the nurse are paid whether or not anyone arrives. IKS holds a patent on a system that scores each patient on how likely they are to attend, using three things: whether they know about the appointment, whether they are able to get there, and whether they intend to. It then sends different reminders to different patients and deliberately double books the slots most likely to go empty.

• Wrong information taken at the front desk. The patient’s details and insurance are captured at this first contact. Industry studies consistently find that between 60% and 70% of all rejected bills start here. A wrong policy number typed at step two becomes a rejection at step seven and a write-off at step nine, three to five months later, having consumed staff time at every step in between.

Then comes approval from the insurer. For a growing list of procedures, drugs and scans, the insurer must agree before treatment can go ahead. The burden has become severe. American practices complete 41 to 45 approval requests per doctor per week, each taking roughly 14 minutes of combined doctor and staff time. That is 14 to 16 hours a week for a typical practice. About 35% of doctors now employ staff who do nothing but chase these approvals. Around two thirds say the process has pushed patients towards less effective first treatments or forced extra visits.

This step has no clinical value to the doctor at all. It is the easiest job in the whole chain to justify handing to somebody else, and one of the easiest to automate, because most of it is filling in forms according to insurer rules.

Step four — the doctor sees the patient

Look one level deeper and it gets worse. In general practice, doctors spend 36.2 minutes inside the health record for a visit booked at 30 minutes. The typing takes longer than the appointment. About 2.7 hours a day happens after the clinic has closed, which American doctors call pyjama time.

The most important statistic in this whole report is this one: nearly 85% of doctors say that documentation required purely for billing, not for treating the patient, increases their total typing time.

A large part of a doctor’s unpaid evening work exists to satisfy an insurance company rather than to help a patient. It is waste, it can be measured, and somebody can be paid to take it away. IKS publishes a client result of two hours given back to each doctor per day.

Step five — turning the note into codes

The clinical note becomes standardised billing codes. These codes decide how much the doctor is paid.

• Code too low and the doctor is paid less than the work was worth.

• Code too high and it is fraud, with criminal consequences.

• There is one correct answer and it can be audited afterwards.

This is the highest skill and highest risk job in the chain. It needs certified coders, and the shortage of them in America is a training pipeline problem rather than a hiring problem. India does most of the world’s coding work because it has the deepest bench of certified coders.

Because there is one right answer, this job is the hardest to automate. Management’s own framework, set out on the results call, is worth quoting because it is honest about the limits.

IKS bought a company called ARAI specifically to get knowledge graph technology, so that its systems can show how they arrived at a code rather than simply producing one.

Steps six to eight — sending the bill and fighting the rejection

• Rejection rates rose from 11.8% in 2024 towards 12% to 15% in 2025.

• Rejected hospital inpatient claim amounts rose 12% and outpatient claims 14% in the first three quarters of 2025.

• Every rejection means someone must find the reason, gather documents, write an appeal, resend it and chase it.

• American hospitals spent close to 18 billion dollars in 2025 purely on overturning rejected bills. That figure excludes all the staff time spent earlier trying to prevent them.

IKS bought denial prioritisation technology through an arrangement with Sift Medical Data in March 2023, specifically to move this work from fighting rejections towards preventing them.

Steps nine and ten — getting the money in

Insurers pay their share. Patients owe the rest and increasingly do not pay it. Collection from insured patients has fallen as low as 34.4%, with market averages between 34% and 48%. On amounts above 7,500 dollars, collection can fall to 17%.

As American insurance policies push more cost onto the patient, a growing share of a hospital’s income sits in the hardest bucket to collect. This is why scoring patients on their likelihood to pay, and collecting before treatment rather than after, matters commercially. Money taken before care is delivered is written off at a fraction of the rate of money chased afterwards.

The summary measure for this part of the chain is how many days it takes to get paid. Good performers sit below 30 days. Struggling doctors’ groups run 55 to 75 days, with money concentrated in the 90 to 120 day bucket where insurers start treating bills as uncollectable.

Finally the cash is matched against the bill and checked against the contracted rate. Systematic underpayment against contract is money that simply never arrives, and nobody notices unless somebody is checking.

04 — The economics

What all ten steps cost, and how IKS gets paid

Add up the cost of running all ten steps and you get the number this industry is priced against. It is called cost to collect, and it is expressed as a share of everything the hospital collects.

• The best practice standard set by the Healthcare Financial Management Association is 2% or below.

• Typical hospital performance is 2% to 4%.

• Doctors’ practices commonly run 3% to 6%.

This is the deal in one picture. The supplier asks for a share of the 4.5 million dollars and justifies it by handing back some or all of the 1.5 million saving, plus more money collected in the first place through fewer rejections and faster payment. Both sides can win, which is why paying suppliers on results works in this industry.

IKS publishes results denominated in the client’s money rather than its own hours. One doctors’ group cut the time to get paid from 77 days to 30 and gained more than 5.6 million dollars of cash. A women’s health group called Axia gained 12 million dollars of cash at 96% coding accuracy with fewer coding rejections.

It also sets the ceiling on the industry, and this is the part that gets missed. The supplier’s total available market is cost to collect multiplied by what hospitals collect, and the supplier is contractually obliged to shrink the first number. This is an industry paid by its customers to make its own market smaller. Growth has to come from taking work that is currently done in-house. It cannot come from the pool getting bigger.

Why the work is done from India

India and the Philippines together account for about 59% of all healthcare administration work done offshore. India’s healthcare outsourcing workforce is above 500,000 people. The split between the two countries is not random.

• India takes the process-heavy back office work — coding, charge entry, chasing rejections, posting cash, note writing, analysis. It has the deepest supply of certified coders, the lowest cost, and an overnight production cycle that fits the American working day.

• The Philippines takes work that involves speaking to people — booking appointments, ringing insurers, chasing patients for money. It has native-standard English and evening shifts that overlap with American hours.

IKS runs delivery from Mumbai, Hyderabad, Dallas and Chicago. Each client has an American relationship manager and an Indian delivery manager. That American layer is a real cost that a purely offshore competitor does not carry, and it is one reason IKS wins work that pure cost suppliers do not.

The three ways IKS charges

The company’s offer document is explicit about why this matters. Revenue grows as clients buy other practices, open clinics and hire doctors. Margins improve faster than revenue as software is applied against a price fixed per doctor or per result rather than per hour worked.

This is exactly what the June quarter numbers show: 12% revenue growth after stripping out currency, on a wage bill that grew 4.2%.

Where seasonality comes in

Seasonality on the doctor side of this industry is mild. Companies serving insurers swing much harder, because insurer volumes spike during the autumn enrolment window when members change plans. For IKS, with almost all revenue on repeating contracts priced per doctor or per result rather than per seasonal transaction, the effect shows up in working capital rather than in revenue.

05 — How big each part is

Where the outsourced money actually sits

Only about 13% of the total pool is bought from outside suppliers. That is the growth argument in one number, and it is real. It is also worth asking why it has stayed that low for years. Hospitals keep this work in-house for reasons that have little to do with price: they worry about patient data security, they want control of anything patients see, doctors want their own people, and some hospital boards will not send patient records overseas at all.

This chart contains the single most awkward fact in the whole analysis. The two services IKS is ranked first in are the two that have already been outsourced the most.

• Coding is 29% outsourced already, the highest of any part of the work, and the underlying market grows at 0.5% a year. The offer document itself says the coding market will be roughly unchanged at about 5 billion dollars in 2028. This is a mature market where the fight is on price.

• Note writing is 27% outsourced, and it is the part most exposed to software that listens to the consultation and writes the note automatically. This is where Abridge competes, a company IKS first hired as a supplier, then invested in, and now competes with.

• The two large pools with real room are billing and collections at 18 billion dollars and only 18% outsourced, and clinical support at 7.7 billion and 11% outsourced. These are exactly where R1 RCM, Omega Healthcare, Access Healthcare and Ensemble concentrate.

• Value-based care grows fastest at 21% a year and IKS has capability there rather than leadership.

An independent check is available. A separate industry study, commissioned by a competitor for its own listing, sizes total American healthcare administration spending at 201 billion dollars in 2023 rising to 259 billion by 2028, with about 22% bought from outside suppliers, making an outsourced pool of roughly 45 billion covering both the insurer side and the doctor side. The definitions do not match the 34 billion above, so neither number should be treated as gospel. But on the split that matters, that study is clear: work bought by doctors and hospitals is less outsourced than work bought by insurers, 19.5% to 21.5% against 22% to 24%, and it grows at 12.5% a year against 7%. IKS is entirely on the faster growing side.

06 — Competition

Every serious rival is privately owned

The right comparison group is companies serving American doctors and hospitals on their billing. Companies serving insurers are a different business with a different buyer, different pricing and different seasonality. Diversified outsourcing companies where healthcare is one division among several are not comparable either. The industry study inside the IKS offer document says this plainly, noting that other recently listed Indian names do not compete directly with IKS and that the listed comparison most often used serves insurers.

Table 1 · How the competitors are built

Table 2 · How they compare on numbers

Two notes on that table. R1's revenue per employee is much higher because it takes over a hospital's entire billing department, including American staff, and charges a share of what the hospital collects rather than a fee per task. TruBridge's figure reflects software licences rather than services.

Table 3 · Strengths and weaknesses

07 — The company

What IKS is, and what its offer document reveals

IKS was founded in 2006. It sells to doctors’ groups and to the physician arms of hospital systems, rather than to hospital billing departments. That is a deliberate difference from R1 and Ensemble, and it is the reason IKS can charge on doctor outcomes, such as hours given back per doctor per day, rather than on cost per bill.

Two purchases define the company’s shape.

• AQuity Solutions, 2023 to 2024. Brought clinical note writing and transcription at scale, taking combined revenue past 330 million dollars and staff past 14,000 serving more than 150,000 doctors. It also brought the data used to train the software, which is why the technology claims here have substance rather than being a licensed product from someone else.

• TruBridge, completed July 2026. A different order of magnitude and a different kind of bet. Section 9 covers it in full.

  

  The client base got much safer and nobody discusses it

  

  

  Almost 100% of revenue sits on repeating contracts. This is not project work dressed up as a platform. It is embedded operating infrastructure billed every month.

• About half of revenue comes from clients buying more than four different services, and another 37% from clients buying two to four. Selling more services to existing clients is not a plan for the future. It is already how most of the money is earned.

• The client list includes Massachusetts General Brigham, Lehigh Valley Health Network, Boston Children’s Health Physicians, Duke Health, McLaren, Providence, Texas Health Care, The GI Alliance and US Digestive Health.

• The Massachusetts General Brigham relationship goes beyond a contract. A joint development agreement with a Boston affiliate produced the prescription renewal product, the data migration tool and the discharge summary product.

One number sits awkwardly against all of that. The sales and marketing team was 51 people in March 2024 and 40 by September 2024, with marketing spending at 0.69% of revenue. A company serving 778 healthcare organisations with forty sales people is either unusually efficient or structurally limited in how fast it can win new names. The offer document says growth comes from expanding inside existing clients and from word of mouth. That works until the existing base is fully sold into.

Staff turnover, the number that stopped being published

• Wages were 75.3%, 73.4% and 67.9% of total costs in FY22, FY23 and FY24. This is overwhelmingly a people business.

• In FY23 the company replaced more than half its staff in a single year.

• Of 12,862 permanent staff in September 2024, 1,019 were doctors. There were a further 666 contract workers.

• Replacing half a workforce every year carries recruitment cost, training cost, lost output while new people learn, and quality risk in exactly the jobs where mistakes are expensive.

There are two readings. The kind one is that turnover was falling sharply into the listing and the 14.5% half-year figure suggests a genuine change, which fits the story of fewer, more skilled people doing higher-value work. The sceptical one is that 44.5% was the last full year published, and the company has not given the number since.

It is recoverable. Every listed Indian company files a business responsibility report inside its annual report, and one of its required disclosures is employee turnover. The figure exists for FY25 and FY26 even though no results presentation carries it.

They have done the TruBridge job once before

Every measure that matters got worse in the AQuity year. Margin fell 6.4 points. Return on equity fell 5 points. Cash generated against profit halved from 91.6% to 47.8%. Profit per employee fell.

Then it was rebuilt. The client list was cut from 853 to about 600. Delivery was moved offshore. Headcount came down from 13,241 to 12,889 by June 2026 while revenue nearly doubled. Margin got back to 34.2%.

Management’s plan for TruBridge is the same plan: margin drops to 26% or 27% at first, then rebuilds to the low thirties within a couple of years. That is not a promise. It is a repeat of something already done, on a business of similar relative size, with the numbers filed publicly.

One caution. AQuity left goodwill of 11,738 million rupees and customer relationship intangibles of 5,110 million rupees on the balance sheet, amortizing at about 547 million a year. Against total equity of 13,771 million rupees in September 2024, goodwill alone was about 85% of book value. TruBridge adds more. Comparing this company to others on price against book value is meaningless, and the annual test for whether that goodwill is still worth carrying is worth watching, particularly given the rural exposure.

08 — Recent trading

The growth is smaller than the headline, the efficiency is real

Both halves of that chart matter and they point in opposite directions.

• The concerning half. Growth of 12% after removing currency is the slowest since the company listed, on a business that has averaged around 14%. A rupee that stops falling removes nine points from the reported number. Management’s own stated ambition for the original business is growth “north of 12%”, which is the rate the market itself grows at. The company is telling you it aims to hold its share, not take share.

• The encouraging half. Twelve per cent real revenue growth delivered on a wage bill that grew 4.2% and a headcount that grew 4.8%, from 12,300 people to 12,889. That gap is what produced a thousand basis points of margin improvement in two years. Management stated on the call that the profit growth came from operations, not from currency, in a quarter where currency gains were neutral.

Over the ten years to 2026 the company grew revenue at 28.7% a year and profit at 46.4% a year. Headcount has actually fallen since the AQuity purchase while revenue roughly doubled.

Two footnotes on the June quarter. Return on equity fell from 32.3% to 26.4%, caused not by trading but by writing up the Abridge stake through reserves, which increased the equity base without touching profit. And a 143 crore rupee upfront guarantee payment to a client distorted the cash lines. Adjusted for it, the company converted about 90% of profit into free cash.

09 — TruBridge

What the American filings show about what was bought

TruBridge was a listed American company until July 2026, trading on Nasdaq as TBRG. Its last annual report, filed on 31 March 2026 for the year to December 2025, and its merger document filed on 4 June 2026, answer most of the questions the Indian disclosure leaves open.

The revenue record

• Revenue is 98.3% American. About 5.85 million dollars comes from St Maarten, Turks and Caicos, Anguilla, Canada, England, Australia, the United Arab Emirates and the Netherlands.

• New orders were 82.1 million dollars in 2024 against 80.2 million in 2023. Flat.

• The order book supports the old revenue number, not the reset one. The annual report discloses 332.0 million dollars of repeating revenue contracted for the next twelve months at 31 December 2025, up from 328.0 million a year earlier, plus 6.0 million of one-off system sales.

• Billing services are about 64% of revenue. The health record and patient tools are the rest.

• The target market is about 4,600 American hospitals with under 400 beds, of which about 3,000 have under 100 beds. About 98% of existing health record customers are hospitals with under 100 beds.

This matters for the forecast. A 4% growth assumption is not supported by anything in the record. The model behind this report now uses 1.5%, which is what the company actually delivered. Against that, the contracted order book of 332 million dollars is the strongest evidence that the reset to 300 million is IKS choosing to walk away from unprofitable work rather than a business falling apart.

Investment was being cut before the sale

There are two ways to read this. The kind reading is discipline, after the Centriq platform was discontinued and the post-acute care business was sold. The sceptical reading is that a health record supplier cutting product spending, in a market where Epic is winning every large decision, is starving the asset. Either way, IKS has bought a software business whose development budget has been shrinking, and has to decide whether to reverse that.

The merger document sets out the sequence in its own words, and it is worth reading slowly.

• 26.25 dollars is a 14.7% premium to the 22.88 dollar close on 22 April 2026, the last trading day before the merger agreement.

• It is a 49% premium to the 17.59 dollar close on 16 March 2026, which was the day before TruBridge announced it would not file its 2025 annual report on time.

• It is an 87.5% premium to the unaffected price of 14.00 dollars on 30 March 2026, the day before the company disclosed in that late annual report that it had been running a review of strategic options.

How the deal was put together

• Total funds needed were about 572 million dollars gross, against debt commitments of up to 670 million dollars. Not all of it was drawn.

• The lenders were Citigroup Global Markets Asia, Citibank, Deutsche Bank Singapore and JPMorgan Hong Kong. All Asian branches, with the Indian listed parent guaranteeing the debt.

• A separate approval from Indian shareholders was required within 60 days of signing, alongside the American vote.

• TruBridge would pay 12.3 million dollars if it walked away. IKS would pay 24.6 million dollars if it failed to complete. The outside deadline was 23 October 2026.

• Solomon Partners gave the fairness opinion on 23 April 2026. Holders of about 27% of shares, being Pinetree Capital, L6 Holdings and Ocho Investments, signed agreements to vote in favour.

• The vote on 7 July 2026 was 11,305,399 in favour against 8,818 opposed, with no shareholder claiming appraisal rights. That is 99.75% in favour, with nobody disputing the price.

• Performance share awards granted to TruBridge management in 2024 vested at the maximum 200% of target on the change of control. Awards from 2025 and 2026 vested at 100%. Executives also held severance agreements, accelerating cash awards and retention bonuses.

The annual report also confirms why the business was available. The board had been running a formal review with outside financial and legal advisers, considering the sale of all or part of the company, a joint venture, share buybacks or investing for growth. IKS bought a company that had put itself up for sale.

The India connection was already there

• TruBridge had over 3,500 staff at the end of 2025, roughly half in America and half in India.

• More than 1,800 were in India, mostly around Mumbai, working on the billing services business.

• That workforce came with the October 2023 purchase of Viewgol, a clinic-focused billing business.

• A dedicated India capability centre opened in February 2026.

This changes the integration story. IKS is not moving an all-American business offshore from a standing start. TruBridge had already built an Indian delivery arm in the same city where IKS runs its own operations. The margin plan has a running start.

One personnel link is worth noting. TruBridge appointed Mike Daughton as chief business officer in October 2025. He spent 2011 to 2023 in senior roles at AQuity Solutions, including chief commercial officer, where he led that business through its sale to 3M and then to IKS Health. Six months after his appointment, TruBridge agreed to sell itself to IKS.

What the filing says about the customers

That last pair is the sharpest thing in the filing, and it cuts both ways.

• For the customer, it is a squeeze. Medicare Advantage now covers about half of rural elderly patients, pays those hospitals about 90% of the traditional rate, and demands more prior approvals. On top of that come the Medicaid cuts. More than 40% of American rural hospitals already lose money, and 417 are judged at risk of closing.

• For a billing supplier, it is demand. More approval requests, more rejections, more appeals. That is exactly the work IKS sells.

The same policy that threatens the customer base increases the amount of work per customer. Which effect wins is the central open question on this purchase, and it can only be answered by watching TruBridge’s customer count, not its revenue. A rural hospital that closes does not move to a competitor. It disappears, and the revenue never returns to anyone.

Running the model at 10% annual customer loss instead of 3% removes roughly 196 crore rupees of profit by FY30, from the same cash flow that is meant to be paying down the debt.

Who TruBridge said its competitors were

Note what is absent. R1 RCM, Omega, Access Healthcare and GeBBS appear nowhere in TruBridge's competitor list. TruBridge competes against clearing houses and software firms on billing, and against health record suppliers on the clinical side. IKS's competitors and TruBridge's competitors barely overlap. That is an argument that this purchase genuinely widens the business rather than simply buying more of the same.

10 — Debt and valuation

What the current price already assumes

The debt is manageable

• The loan is priced at the American benchmark rate plus 2.75%, stepping down to plus 1.75% as debt falls.

• Because net debt before the deal was only 265 crore rupees, the starting position is 2.8 times profit, inside management’s own stated ceiling of about 3 times from day one.

• Interest cover starts at about 5 times and rises above 15 times by 2030.

• A two percentage point rise in interest rates still leaves cover above 4 times in the first year.

• Debt is not the risk in this business.

The one thing not in the public record is the exact covenant levels in the loan agreement. Those are assumptions in the model and should be replaced when the annual report is published.

The price against what similar companies have sold for

• R1 RCM, the largest independent billing company in America, handling over a trillion dollars of hospital billing, was taken private in 2024 for 8.9 billion dollars. That worked out at roughly 14.5 times its 2023 profit before depreciation, in the largest private equity buyout of a healthcare technology company on record.

• GeBBS sold in 2025 at 17 times profit and 4.3 times sales, described at the time as the top of the range for this kind of business.

• IKS trades at 33 times.

Part of that gap is earned. IKS earns a higher margin than R1 did. It converts more of its profit into cash. It carries less debt. It produces 71% more revenue per employee than the closest listed comparison. Its client base is far less concentrated. And it operates in the faster growing half of the market. A premium is deserved.

Whether it deserves more than double what the sector leader itself changed hands for is the question.

The company’s own target is the most useful test

Management has put a specific, dated, public number on the record, which is rare for an Indian mid-sized company. It is 3,000 crore rupees of profit before depreciation by March 2030, achieved without issuing new shares beyond employee options, with debt back to where it was before the purchase.

• On that number, today’s enterprise value is 12.1 times 2030 profit. That is a reasonable price for a business of this quality, if the target lands.

• Working backwards instead: apply the multiple that similar companies actually sell for and solve for the profit the current price needs. The answer is about 2,584 crore rupees, which is 86% of the company’s own target.

• So the price does not require more than management has promised. It requires most of it, with little room for the integration to slip, the rural customer base to shrink, or the rupee to stop falling.

• If the full target is delivered and the shares then trade at the transaction multiple, the value works out around Rs 2,570, or 40% above today.

The caution is that the 38.8% already delivered came from a single high margin, asset-light business absorbing a purchase it had already digested. The 29.2% still needed must come from a combined business carrying a lower margin acquisition, a debt load that did not exist before, and integration work that management says will take several quarters simply to understand.

The range of outcomes

11 — What to watch

Eight things that will settle this over four quarters

Each of these is observable, each is published or can be asked for, and each one moves the answer.

1. Growth after removing the currency effect, not the rupee headline. Above 14% and the good outcome opens up. Below 10% for two quarters in a row and the case is broken. Remember the number is stated openly in the quarterly presentation, so there is no excuse for using the rupee figure.Quarterly investor presentation

2. Staff turnover. The last published figure was 44.5% for the year to March 2024. It has to be in the business responsibility section of the annual report, because that disclosure is mandatory. In a business where wages are 70% of costs and the entire margin story rests on revenue growing faster than headcount, this is the most important operating number the company has.FY25 and FY26 annual reports

3. Revenue split by service. Coding is 29% outsourced already with an underlying market growing at 0.5%. Note writing is 27% outsourced growing at 6%. If those two are a large share of revenue, the growth ceiling is lower than the headline market size suggests. The company sizes the market segments in its offer document but has never disclosed its own split.Not currently published. Ask.

4. The wage bill against real revenue growth. The gap between 4.2% and 12% is the business model working. Watch whether it closes.Headcount disclosure and management commentary

5. TruBridge customer count, not TruBridge revenue. Rural hospitals that close show up in the count first. It also feeds the annual goodwill test.Not currently published. Ask on every call.

6. Named client wins at hospitals running Epic. The Epic relationship is the largest unquantified strategic risk in this business. Continued wins in Epic environments are the evidence that it is holding.Quarterly client win disclosures

7. Combined margin getting back to the low thirties. The AQuity recovery took two years from a 6.4 point fall. Hold management to the same clock. They have said 26% to 27% at first and low thirties within a couple of years.Quarterly, from September 2026 onwards

8. Whether the 2030 profit target is repeated or quietly dropped. A specific public number is a governance asset while it stands and a warning sign if it disappears.Quarterly investor presentation

^{Sources. Inventurus Knowledge Solutions prospectus dated 16 December 2024, 601 pages, including the key performance indicator tables, the commissioned industry report, the business section and the restated financial statements. IKS quarterly investor presentation and earnings call of 6 August 2026, and results releases for FY25 and FY26. TruBridge annual report on Form 10-K for the year ended 31 December 2025, filed 31 March 2026. TruBridge merger proxy statements filed 21 May and 4 June 2026 and the related current reports. KLAS health record market share data for American acute care and ambulatory settings, end 2025. Transaction disclosures for R1 RCM and GeBBS. Everest Group industry data as reproduced in a competitor’s November 2024 offer document. Industry benchmarks from the Healthcare Financial Management Association, the American Medical Association, the Medical Group Management Association and Black Book Research. Healthcare outsourcing delivery cost benchmarks from published industry surveys.}

^{A companion financial model accompanies this report. It contains 528 linked formulas with no calculation errors, models the original business and TruBridge separately, splits revenue growth into real growth and currency, carries a loan-by-loan debt schedule on the disclosed pricing, and includes a tab that tests the model against every number management has publicly committed to.}

^{Prepared for personal research use. This is not investment advice.}

When a company’s existing business is not growing and margins are under pressure, it bets on a genuinely new one. The only question: can you underwrite the incremental return? Four historical examples, the Waaree verdict, and the maths that separates compounding from destruction.

The framework is built on a specific premise. A company’s existing business has stopped growing. Margins are under pressure. Management responds by investing capex into a genuinely new business — a new technology, a new product category, a new business model — not more capacity of the same thing. The capex takes years to come online. The stock consolidates. Should you play this cycle, or walk away?

Most investors answer by looking at demand visibility. That is necessary but not sufficient. The decisive variable is whether you can underwrite the incremental ROCE — put a number on the return the new capital will earn, and check whether that number clears the cost of capital. If you can, the consolidation is an accumulation window. If you cannot, the capex narrative is a trap.

· · ·

Four Historical Transitions: Stagnant Core, Genuinely New Business

Four Indian companies went through this exact cycle in recent years. Each had a genuinely stagnant or margin-pressured core. Each invested in a genuinely different business. Two succeeded. One is mixed. One is partial — the new business took four years of losses before turning profitable.

KPIT Technologies (Commodity IT services → Automotive embedded software): +550%. Until 2018, KPIT ran two businesses: a commodity IT services arm and a smaller automotive engineering arm. The IT services business was larger, lower-margin, and commoditised. KPIT merged the IT arm with Birlasoft (January 2019) and listed the residual pure-play automotive engineering entity. Even post-demerger, the automotive core was hit hard: FY21 revenue declined to ₹2,050 crore, net profit fell to ₹141 crore. The new business — automotive embedded software for electric powertrains, autonomous driving, and digital cockpits — is categorically different from body engineering or IT outsourcing. It is middleware/OS-level software for software-defined vehicles, a domain where KPIT took a deliberate leadership bet (BMW scalable autonomous-driving platform, ZF partnership). The capex was R&D and talent, not plant. ROCE went from ~14% (FY21) to ~39% (FY25). The stock went from ₹34 (March 2020 low) to ₹1,929 (July 2024 peak), before correcting to ~₹615 in 2026 on auto-OEM R&D softening. The consolidation phase of 2020–21 was the accumulation window.

KPIT — Why It Worked

The demerger removed the structurally weak business. The focused automotive-software rebuild rode the CASE (connected, autonomous, shared, electric) super-cycle with genuine domain depth. Incremental ROCE was underwritable from the start: the business was asset-light, the customer contracts were visible (BMW, ZF), and the margin profile was structurally higher than the old IT services business. The consolidation was a classic J-curve — temporary earnings depression masking value creation.

Persistent Systems (Legacy product-development IT → Cloud/digital engineering): +2,000%. The legacy OPD business for ISVs was under pressure from hyperscaler disintermediation. FY20 revenue grew only 5.9%, EBITDA fell 15.1%, PAT declined 3.2% — negative operating leverage on a barely-growing base. Through FY18–FY19 revenue was essentially flat (₹3,113 crore → ₹3,366 crore) with declining EBITDA — classic stagnation of a legacy model.

The pivot was into cloud-native engineering, data/AI, and hyperscaler-partner-led modernization (AWS, Azure, Salesforce, Google Cloud) — a structurally different revenue engine from time-and-materials OPD. Executed through acquisitions: MediaAgility (Google Cloud Premier partner), Data Glove (Microsoft Azure, dedicated Microsoft Business Unit), and Sureline Systems (cloud migration). CEO Sandeep Kalra explicitly targeted cloud, data, security, and Europe expansion as the new axes. This is a genuinely different business — productized hyperscaler-partner delivery rather than legacy OPD. ROCE expanded from ~19% to ~33%. The stock went from ₹210 (2020 low) to ₹6,789 (2024 peak) — roughly a 32x move. The cleanest of the four.

Persistent — Why It Worked

The legacy OPD/IBM book was genuinely stagnating; the cloud-and-hyperscaler rebuild was genuinely different and higher-return. The acquisitions were capability buys in the $10–50M revenue range, each bringing a pre-existing hyperscaler partnership. The stock consolidated in 2022 (down 21% that year) then rewarded as the thesis proved out. The incremental ROCE was underwritable because the new revenue was partnership-led with committed hyperscaler pipelines, not merchant capacity hoping for buyers.

Gravita India (Lead recycling → Lithium-ion battery recycling + copper): Mixed. The core lead recycling business was structurally margin-constrained. Revenue grew from ₹1,029 crore (FY18) to ₹1,410 crore (FY21), but EBITDA margin was volatile and thin: 4.7% (FY19), 7.2% (FY20), 8.6% (FY21). CARE Ratings explicitly flagged the “low value addition nature of its business and weak entry barriers” and “limited pricing flexibility” due to LME-linked pricing. This is a commodity pass-through with structurally constrained returns.

The new businesses are genuinely different. Lithium-ion battery recycling — a 6,000 MTPA plant commissioned at Mundra in FY26 — uses different feedstock (Li-ion batteries vs. lead-acid), different technology (hydrometallurgical vs. pyrometallurgical smelting), and serves a different end-market (EV battery materials). Copper and copper alloys via the ₹562 crore acquisition of Rashtriya Metal Industries (RMIL), completed March 2026, adding ~50,000 MTPA copper capacity and ~₹900 crore revenue. Total planned capex to FY29 is ₹1,200–1,300 crore.

But the outcome is mixed. ROCE has declined from 31–32% (FY22–23) to 17% (FY26) as capex on new verticals is still ramping. Working capital days stretched from 40 (FY22) to 115 (FY26). RMIL currently earns ROIC of only ~14% at 50% utilisation, with management targeting 25% over time. Lithium-ion recycling is pre-revenue scale, with technology still being “streamlined.” The stock delivered a 3-year CAGR of ~35% but is down ~15% in 2025–26 as margins compressed. The jury is still out on whether the genuinely new businesses will earn their target returns.

Gravita — The Underwriting Question

The core lead business ROIC of ~25% is proven. The new businesses (Li-ion, copper) are not yet earning their cost of capital at current utilisation. The underwriting depends on: (a) Li-ion technology stabilisation and scale, (b) RMIL utilisation ramp from 50% to 80%+, (c) working capital normalisation from 115 days back toward 60–70. If these materialise, the incremental ROCE will be attractive. If they don’t, Gravita will have diluted a 25% ROCE core with sub-15% new capital — value destruction.

Ashok Leyland (Diesel commercial vehicles → Electric CVs via Switch Mobility): Partial. The diesel CV business was cyclical, not terminally declining, but Ashok Leyland was ceding market share and margins were thin (EBITDA ~4.1% in FY22). The new business — battery-electric buses and LCVs via Switch Mobility — is a genuinely different technology stack: separate electric powertrain, separate IP, separate manufacturing. Ashok Leyland had already invested over $130 million into Optare’s EV business over the prior decade; in 2021 it committed a further $150–200 million, scaling to a $500 million-plus 3–5 year EV plan. Switch India committed ₹1,000 crore to a dedicated EV plant (30,000 e-LCV + 10,000 e-bus capacity). A further ₹1,200 crore was infused in 2024.

The incremental ROCE was negative for four years (FY21–FY25). Switch UK failed (plant closure, losses written off). Switch India finally turned PAT-positive in FY26 with over ₹100 crore PAT, delivering 1,530 e-buses (up 238% y-o-y) and ~1,600 e-LCVs. But the eMaaS arm (Ohm) is still loss-making. The stock went from ₹17 (2020 low) to ₹215 (Feb 2026 peak), but the rerating was driven as much by the diesel cycle recovering (margins from 4.1% to 12.7%) as by EV success. No external investor has come in at the hoped-for $1.4–1.8 billion valuation. A real but expensive optionality play.

Ashok Leyland — The Duration Problem

The EV business is genuinely different from diesel trucks. The technology works. Switch India is now a credible #1/#2 in e-buses. But four years of negative incremental ROCE is a very long J-curve. The parent stock’s rerating was driven as much by the diesel cycle recovering as by EV success — the market rewarded the old business healing, not the new one earning. The lesson: a capex transition can be “directionally right” and still take so long that the incremental ROCE is irrelevant to the stock performance during the transition.

· · ·

The Five-Question Framework

The framework rests on five questions. The first three screen. The fourth decides. The fifth determines durability.

1. Is the core genuinely stagnant?

Not “slowing from 20% to 10%” — genuinely not growing, margins under structural pressure. Persistent’s revenue was flat at ₹3,100 cr with EBITDA falling 15%. KPIT’s automotive revenue declined in FY21. Gravita’s EBITDA was 5–9% on a commodity pass-through. Ashok Leyland’s CV margins were 4.1%.

2. Is the new business genuinely different?

Not more capacity of the same thing. KPIT’s automotive software is different from IT outsourcing. Persistent’s cloud engineering is different from OPD. Gravita’s Li-ion recycling is different from lead. Ashok Leyland’s EVs are different from diesel trucks.

3. Is the demand real, contracted, or imaginable?

A binding term sheet is visible. An MoU is hoping. A TAM forecast is imagining. The difference between a ₹10,000 cr factory at 80% utilisation and the same factory at 35% changes everything.

4. Can the incremental ROCE be calculated?

Build it bottom-up: capacity × utilisation × price = revenue. Revenue × margin = EBITDA. Subtract depreciation for EBIT. EBIT ÷ capital = ROCE. This is the decisive question.

5. What prevents competitors from destroying the return?

If a factory earns 30% ROCE, why won’t competitors build the same one? High ROCE persists only with a moat: policy protection, captive demand, technology barriers, or scale.

The Demand Quality Hierarchy

The J-Curve: Temporary Dip or Permanent Destruction?

The J-curve is not automatically bullish. A J-curve followed by a high incremental ROCE is a temporary earnings depression — KPIT and Persistent both went through this. A J-curve followed by a low incremental ROCE is permanent capital destruction. Gravita is in the middle of its J-curve right now: ROCE has fallen from 32% to 17% as new capital deploys, and the recovery depends on whether Li-ion and copper can earn their target returns. Ashok Leyland’s J-curve on EV lasted four full years before Switch India turned profitable — a very long wait for the incremental ROCE to turn positive.

The Three Hidden Variables

Working Capital: A ₹5,000 crore plant may require another ₹1,000 crore of inventory, receivables, and ramp-up losses. Gravita is the live example: working capital days stretched from 40 to 115 as new verticals ramped, absorbing capital that doesn’t show up in the plant capex headline number but absolutely affects the incremental ROCE.

Depreciation: Capex of ₹10,000 crore with EBITDA of ₹2,000 crore looks like 20%. But if depreciation is ₹1,200 crore, EBIT is only ₹800 crore — an 8% return that may be below cost of capital. The final question is steady-state EBIT ROCE, not EBITDA return.

Non-Uniform Cost of Capital: A 15% ROCE from a regulated asset may be excellent. A 15% ROCE from a new battery chemistry may not be enough. The hurdle rate must reflect the risk.

The ₹100 Test

The framework compresses into one question: If I gave management ₹100 today, how much sustainable operating profit will that ₹100 produce once the project reaches maturity? If ₹25–30, extraordinary. If ₹15, examine risk and duration. If ₹8–10, need a very good reason. If “we will know once the plant ramps,” the market should not be paying as though the economics are proven.

KPIT’s ₹100 test was clear: software engineering talent at ~₹15–20 lakh/employee generating ₹35–40 lakh revenue/employee. Persistent’s was clear: hyperscaler partnership contracts with committed pipelines replacing merchant OPD billing. Gravita’s is still unclear: the Li-ion plant is commissioned but not yet at scale. Ashok Leyland’s took four years to answer — and the answer was modest, not extraordinary.

· · ·

The Thesis in Action: Waaree Energies

Waaree is the most important current test of this framework. The framing matters enormously: this is not “a company with a growing core investing in more capacity.” This is a company whose module business is the stagnant core under pressure, and whose cell backward integration is the genuinely new business.

India closed 2025 with approximately 140 GW of module manufacturing capacity against domestic demand of ~45 GW — a 3x oversupply. Module utilization collapsed from ~72% (FY23) to ~40% (FY26). DAM Capital warns of “materially weakened return ratios for pure-play module assembly businesses.” This is the stagnant core: commoditised, oversupplied, with structurally declining returns.

The cell backward integration is the new business. Domestic ALMM-listed cells sell at a 130–160% premium to imported cells. DCR modules using domestic cells command 2–3x the margin of non-DCR modules. Waaree’s current cell-to-module integration ratio is only 18.3%, but management targets 93% by FY28E. The cell capacity is being scaled from 5.4 GW to 15.4 GW (10 GW incremental, equipment already at site), plus a 10 GW ingot-wafer facility at Nagpur.

The underwriting holds for the next ~18–24 months because ALMM-II creates a domestic cell mandate and cell supply (18 GW) is well short of demand (50 GW). But ICRA projects cell capacity scaling to ~100 GW by December 2027. The cell scarcity premium is itself finite. Waaree’s blended ROCE is already declining in forward estimates — from ~30.6% (FY26) to ~24% (FY27E) to ~21% (FY28E) — as the capex base triples.

· · ·

The Verdict: Should You Invest in Waaree?

Yes — but with a defined time horizon, not as a permanent compounder.

The incremental ROCE on cell capex is ~25–32% — underwritable and clearly above cost of capital. The three reinforcing pillars hold: ALMM-II policy mandate (21 GW cell shortfall), captive offtake (cell feeds own 26 GW modules), and equipment already at site (10 GW incremental cell, commissioning H2 FY27). The underwriting is real.

But the cell scarcity premium is finite. Cell capacity is scaling from 18 GW to ~100 GW by December 2027. The premium that drives today’s 25–32% incremental ROCE will normalise as supply arrives. Waaree’s next leg — ingot-wafer at Nagpur (FY28) — extends the runway, but at lower incremental returns because wafer manufacturing is more capital-intensive and less policy-protected than cells.

Play the cell backward integration cycle. Do not extrapolate 25%+ ROCE indefinitely. The underwriting window is roughly 18–24 months before cell margins start normalising. The stock may consolidate during this period as the market waits for the cell capex to commission and ramp — that consolidation is the accumulation window, because the incremental ROCE is underwritable and the captive offtake de-risks utilisation. Once the cell premium normalises (FY28–29), the thesis shifts to the wafer leg, which is a different underwriting with lower returns.

Waaree: Framework Verdict

PLAY — with a time horizon. The incremental ROCE on cell capex (~25–32%) is underwritable and above cost of capital. Captive offtake (cell feeds own 26 GW modules) de-risks utilisation. ALMM-II creates a domestic cell demand floor with a 21 GW shortfall. Equipment is at site.

BUT: Cell capacity scaling to 100 GW by December 2027. The cell scarcity premium is finite. Underwriting window: ~18–24 months before normalisation begins. Next leg (ingot-wafer, FY28) extends runway at lower returns. Do not extrapolate 25%+ ROCE indefinitely.

· · ·

The Trap: Kitex Garments

Kitex is the textbook “can’t underwrite” case. The standalone infantwear business is structurally high-margin (~24% EBITDA, ~20% ROCE) but faces acute tariff pressure. The ₹3,550 crore Telangana expansion into adultwear is a genuinely different business — new product, new customers, new geographies where Kitex has no track record. The stock rallied from ₹37 to ₹324 on pure capex optimism — no committed offtake, no project IRR guidance. Then the 50% US tariff took effect. Q4 FY26 margin collapsed to 1%. Consolidated ROCE collapsed to ~2%. The stock crashed 53% from peak. The market paid for a return profile that was never quantifiable.

When the Underwriting Says No: Deepak Nitrite

Deepak Nitrite shows what happens when you can underwrite, and the answer is “no.” Core phenolics face structural margin pressure. ROCE collapsed from 35% to 10.7%. The new bet is polycarbonate via ₹3,720 crore capex. But PL India estimates PC plant ROCE at only 9–10% — below cost of capital. The underwriting exists, but the answer is “no.”

· · ·

Cash Reinvestment: The Compounding Wonder

If you find a company that can reinvest at high incremental ROCE with a multi-year runway, you have found a compounding machine. The equation: Earnings Growth = Incremental ROCE × Reinvestment Rate × Duration. Duration is critical. KPIT sustained ~30%+ ROCE for 4+ years and created enormous value. Persistent’s cloud rebuild expanded ROCE from 19% to 33% over a similar period. Gravita’s new verticals are still in the “prove it” phase. Ashok Leyland’s EV business took four years to reach profitability. The runway length matters as much as the return height.

Operating Leverage: The First 30% Matters Most

The first 30% of utilisation takes the plant from loss-making to break-even. The next 30% generates the bulk of EBITDA. This is why captive offtake is the single biggest de-risking mechanism — it guarantees the first 30%.

The Capex Cycle: Scarcity to Normalisation

Every investment cycle goes through: scarcity → high returns → capital inflow → capacity expansion → competition → price pressure → normalised returns. The investor’s job is to identify which companies retain attractive returns after the capital has arrived. India’s module expansion already shows this: 140 GW capacity vs 45 GW demand. Cell capacity is scaling from 18 GW to 100 GW. The cell scarcity premium driving Waaree’s incremental ROCE today is itself finite.

The Value Accretion Chain

The Project Returns Maths

· · ·

The Capex Transition Checklist

#CategoryQuestion1The old businessIs the core genuinely stagnant or structurally declining?2The old businessWhat is current ROCE? How much free cash flow does it generate?3The new businessWhat is being built? Is it genuinely different from the core?4The new businessHow much capital is required (including working capital and ramp-up losses)?5The new businessWhat is the expected capacity, utilisation, and selling price?6The new businessWhat EBITDA margin? What is steady-state EBIT after depreciation?7DemandWho is the buyer? How much of capacity is contracted or captive?8DemandWhat happens if utilisation is 20 percentage points lower?9CompetitionWhy can’t competitors replicate the economics?10Capital allocationWhat is the incremental ROCE after working capital, depreciation, and ramp-up time?

· · ·

The Bottom Line

The most important mistake investors make is confusing growth of the asset base with growth of intrinsic value. A company can double capacity and destroy value. It can triple revenue and destroy value. It can enter a trillion-dollar market and destroy value — if the capital earns less than the cost of capital.

The opposite is also true. KPIT spent years rebuilding as a pure-play automotive software company while the stock languished below ₹150. Persistent rebuilt its entire revenue engine around cloud and hyperscaler partnerships while the market ignored it at ₹337. In both cases, the temporary earnings weakness was not the thesis. The return on the capital being deployed was the thesis.

Gravita is the live test of whether a 25% ROCE core can absorb sub-15% new capital without destroying value. Ashok Leyland is the test of whether “directionally right” is enough when the J-curve lasts four years. The framework does not promise certainty. It gives you the discipline to know which question you are answering — and which one you are avoiding.

The framework comes down to one question: when the factory is finally running at full speed, will the capital inside it be worth more than the capital that was put into it?

For Waaree, the answer is yes — for the cell backward integration cycle, with a defined time horizon. For KPIT and Persistent, the answer was yes, and the market eventually agreed. For Kitex, the answer was always no. For Gravita and Ashok Leyland, the answer is still being written.

The Trap Test

Before you invest in any capex transition story: Can I build a bottom-up incremental ROCE calculation? Do I know the selling price, utilisation, cost structure, working capital, and depreciation? Does the resulting EBIT return clear the appropriate hurdle rate?

If you cannot answer with specific numbers, you are not investing — you are betting on a narrative.

In 2020, India’s data center capacity stood at a modest 375 megawatts — enough to keep the lights on in a few industrial parks, but barely a rounding error on the global stage. Fast forward to 2026, and that number has ballooned to roughly 1.65 gigawatts. That’s a fifteen-fold expansion in half a decade. And here’s the part that should make you pause: another 8.33 gigawatts is already in the pipeline.

To put that in perspective, 8.3 gigawatts is more than the total installed capacity of several European countries. If even half of that pipeline materializes, India will have built more data center infrastructure in the next few years than it accumulated in its entire digital history up to this point.

First, the big picture

• A data centre is priced in megawatts (MW) of IT load — the power actually available to run servers.

• Two very different products exist under one name: a standard facility (normal cloud, storage, enterprise workloads) and an AI facility (GPU racks, liquid cooling, much higher power density).

• The gap between them isn’t small. It roughly triples the cost.

Read that chart carefully: the numbers on the right (India and global "facility" cost) don't include GPUs. Once you add GPUs, an AI data centre costs $30–45 million per MW globally — the chips are more expensive than the building.

How much does India charge, and is it cheap?

• India builds a standard MW for $5–8 million, against a global average of $11.3 million.

• That’s roughly 45–55% cheaper than the world average.

• But India is not the cheapest place on earth. China, Vietnam and Taiwan all build for less.

So what’s India’s actual edge, if it’s not the lowest price tag?

• A large, English-speaking technical workforce

• Genuine domestic demand — India generates ~20% of the world’s data but holds only ~3–4% of global data centre capacity, so most of this is being built to serve India itself, not just export capacity

• 100% FDI allowed automatically, no prior approval needed

• A 20-year tax holiday (through 2047) for foreign cloud firms operating out of India-based data centres

Taiwan and Vietnam are cheaper. They aren’t offering all of that.

Where does the money actually go?

Split any data centre project into six buckets. Electricity-related equipment — not the building — eats the biggest share.

What changes when you go from standard to AI:

• Cooling nearly doubles its share (20% → 31%) — air conditioning can’t handle GPU racks running 120+ kW; you need liquid cooling

• Civil construction — the actual building — shrinks (18% → 9%)

• Electrical infrastructure grows even further (40% → 50%)

The one-line summary: an AI data centre is not “a bigger warehouse.” It’s a much denser electrical and cooling system stuffed into a smaller shell.

The India price tag, in real numbers

A quick reader question, answered directly: “I read AI data centres cost ₹200 crore/MW — you’re saying ₹70–95 crore. Which is right?”

Both, depending on what you mean.

• ₹70–95 crore/MW is India’s cost for a generic “AI-ready” facility — 20–50 kW per rack, liquid cooling bolted on.

• ₹200 crore/MW matches the global benchmark for AI-optimised facilities ($15–25M/MW), converted to rupees.

• India usually undercuts the world because land and labour are cheap here. But the equipment that makes an AI data centre AI — switchgear, coolant distribution units, high-density UPS — is imported and priced in dollars everywhere. India’s discount barely applies to that part.

• L&T’s own Dholera project (Gujarat) is priced at ₹100 crore/MW — sitting right between the two numbers.

• Bottom line: if someone’s building a real frontier-density AI campus in India today, ₹150–200 crore/MW is a believable number. ₹70–95 crore/MW describes a more modest “AI-ready” build. Always ask what’s included before comparing two numbers.

What a 1 MW facility looks like, cost by cost

For a standard Tier III facility at ₹58 crore/MW:

Why that last row matters: a small facility and a large one on the same grid can pay a similar substation connection fee. Small projects get hit disproportionately hard by this one line.

Running the thing: power is almost the whole story

Once built, a data centre spends 8–15% of its revenue on operating costs. Of that, power alone is 50–65% — and can go past 70% for an AI facility running GPUs near-continuously.

Everything else — staff, security, insurance, maintenance — is small change next to the electricity bill.

The single biggest lever an operator has isn’t engineering. It’s how they buy power.

• DISCOM (state utility) commercial rate: ₹8–11 per unit

• Open access (buying renewable power directly, common above 1 MW): ₹3.50–5.00 per unit

That’s a 2–3x swing on the biggest cost line in the business.

A 100 MW facility can pay anywhere from ₹591 crore to ₹1,248 crore a year in power costs, just depending on how it sources electricity. That’s a ₹657 crore/year gap — bigger than the entire land cost of building the facility in the first place.

Water, for comparison: a 100 MW facility uses about 2.55 billion litres of water a year, costing roughly ₹12.75 crore. That’s about 1% of the power bill. Water gets a lot of attention because most Indian data centre hubs — Mumbai, Chennai, Hyderabad, Bengaluru — are also water-stressed cities. The cost is trivial. The reputational and regulatory risk isn’t.

Who actually builds all this

Land and construction are Indian. The electrical “brain” — much of it isn’t.

LayerWho’s in itConstruction (EPC)L&T Construction, Sterling & Wilson (Shapoorji Pallonji), Tata ProjectsUPS, switchgearVertiv, Schneider Electric, ABB, Siemens, EatonDiesel generatorsCummins, Caterpillar (global); Kirloskar Oil Engines, Mahindra Powerol (domestic)TransformersHitachi Energy, Siemens, ABB (global); CG Power (domestic)Cooling / chillersVertiv, Stulz, Johnson Controls (global); Kirloskar Chillers, Blue Star (domestic)Racks / AI serversNetweb Technologies — India’s own listed HPC/server maker, now partnered with Vertiv on liquid-cooled AI racksGPUs100% imported — Nvidia, deployed via Yotta, Tata Communications, and othersOperators (who owns the capacity)NTT (~20%), STT GDC (~19%), Sify (~19%), Nxtra/Airtel (~15%), CtrlS (~15%), plus fast-scaling newer names — AdaniConneX, Yotta, Digital Connexion (Jio/Brookfield/Digital Realty)

The bottleneck nobody talks about: transformers and high-voltage switchgear are under 10% of total project cost — but they take 12+ months to arrive, and that’s rising, not falling. Global switchgear lead times now stretch past 52–80 weeks, with manufacturers sold out through 2028. A capacity plan built purely on construction-cost math will underestimate how long delivery actually takes. The real constraint on India hitting its targets isn’t land or labour. It’s transformers.

What actually slows a project down

None of these are about money. All of them are about time.

• Approvals: a new facility needs 40+ clearances — land use, fire safety, power connection, environmental consent, water permits — spread across state, municipal and utility processes. India has no single national data centre policy; every state runs its own version.

• Power connection: usually the single longest step in the entire approval chain.

• Land: hyperscale facilities need large, contiguous plots near both a power substation and a fibre route. In Mumbai, Chennai, Bengaluru and Hyderabad, that land is genuinely scarce.

• Water: every one of India’s five biggest data centre hubs is also a water-stressed city. There’s currently no binding national rule on how much water a data centre is allowed to use.

• Skilled labour: India is projected to be short 100,000+ specialised workers (power systems, cooling, compliance) by 2033.

• Equipment supply chains: transformers, switchgear, and copper are all facing global shortages and lengthening lead times — this is a worldwide problem, not an India-only one, but India is competing in the same queue as everyone else.

A 60 MW facility running late costs roughly $14.2 million a month in lost revenue. These delays are not academic.

Where it’s all landing — and why

Mumbai has 51% of India’s live capacity. Here’s the actual reason, not the assumed one:

Over 97% of the world’s international internet traffic travels through undersea cables. Mumbai hosts or is building seven separate cable systems landing there. Add India’s financial centre sitting on top of that — banking and trading data can’t tolerate latency — and you get a location advantage that no other city can copy with a subsidy.

But the pattern is starting to crack, and this is the interesting part:

• CtrlS is building a 612 MW campus in Hyderabad, not Mumbai

• AdaniConneX’s flagship AI campus is in Chennai, not Mumbai

Why? AI training workloads don’t care about latency the way banking transactions do. So AI-cluster siting is now chasing power availability and large land parcels — both more available in Hyderabad and Chennai than in land-locked Mumbai. This is the first real shift in a decade-old pattern, and it’s happening for a genuine economic reason, not a policy push.

Who’s actually demanding all this capacity:

• Hyperscalers (AWS, Microsoft, Google, Meta) account for 60–70% of new demand

• Microsoft has committed $17.5 billion cumulatively to India; Google $15 billion for a single Andhra Pradesh campus

• Data localisation rules (RBI payment-data mandate, the DPDP Act, penalties up to ₹250 crore for violations) force a chunk of this demand to exist in India regardless of global cloud economics — this is a durable demand base, not a cyclical one

If India builds all 8.33 GW — what does that cost?

Three scenarios, depending on how much of the pipeline ends up AI vs standard:

These numbers are facility cost only — no GPUs included. Add GPUs at global pricing for the AI-heavy scenario and the number jumps past ₹20 lakh crore ($240 billion+).

That actually explains something that looks confusing on the surface: India’s government has floated a $200 billion investment target, and cumulative announced investment already sits around $126 billion. Those numbers don’t match a “facility-only” calculation — they match a mix of facility cost, land, power infrastructure, and a growing share of GPU spending. Different people quoting different scopes of the same pipeline.

None of these are directly comparable to each other. Some include power generation assets. Some are multi-year investment envelopes, not pure construction spend. Whenever you see a big headline number attached to a data centre deal, the first question worth asking is: what’s actually inside that number?

The five things worth remembering

1. India builds a standard data centre for 45–55% less than the world average — but it’s not the cheapest country on earth. Its edge is workforce, policy, and real domestic demand, not raw price.

2. Power, not the building, decides whether a data centre makes money. A facility’s power-sourcing strategy can swing annual costs by 2–3x.

3. For AI facilities specifically, the building is the smaller bill. GPUs cost more than the shell around them.

4. The real bottleneck to India hitting its 8 GW target isn’t money or land — it’s transformers and switchgear, which take over a year to arrive worldwide.

5. Mumbai’s dominance is genuine, not lazy — it sits on India’s undersea cable landing points. But AI-specific capacity is already choosing Hyderabad and Chennai instead, for reasons that make economic sense.

All figures in INR unless stated; USD converted at ~₹88/$. Full source list available on request — data drawn from CareEdge Ratings, IMARC Engineering, Blackridge Research, CEEW, Takshashila Institution, SBI, Cushman & Wakefield, JLL, Turner & Townsend, Knight Frank India, MeitY, and company disclosures, 2025–2026.

India ships about 9 billion online orders a year. Two-thirds of them never touch a hired logistics company. The fight is over the rest — and over what flows into it.

Section 5 rebuilds the denominators and shows which claims survive. Most sell-side work on this sector is a definitional muddle; the arbitrage is in doing the parcel arithmetic yourself.

Anatomy of a Parcel — First Mile, Mid Mile, Last Mile

A parcel passes through three different businesses on the way to a door. Mixing them up is how people end up comparing Shadowfax with Blue Dart.

The three legs are three different businesses. Mid-mile rewards capital and density: a sort centre running at 80% utilization crushes one running at 40%, and this is where Delhivery's decade of capex lives. Last-mile rewards people management: there are no assets to sweat, only a few hundred thousand gig riders to recruit, keep and route. First-mile barely shows up in the cost line, but it decides stickiness, because that is where the 3PL's software plugs into the seller's warehouse.

This is why like-for-like matters. Blue Dart is a mid-mile business — own aircraft, premium B2B lanes, 22% ROE — that happens to deliver parcels. Shadowfax is a last-mile business — 2.6 lakh gig riders, about four turns of capital — that rents its mid-mile. Delhivery is the only Indian player running all three legs at scale. That is its margin opportunity (express service EBITDA is already 16–18%), and it is also why its capital intensity is high and its ROE has lagged.

KEY TAKEAWAY — The last mile is 40–50% of parcel cost and roughly 0% of parcel capex. It is labour, organised by software. That one fact explains the sector: it is why clients can take the work in-house (nothing to build, just riders to hire), why Shadowfax tripled revenue in three years on ₹431 Cr of lifetime capex, and why nobody here will earn Blue Dart margins without buying planes.

OPINION (Margin note) — Blue Dart bought aircraft. Shadowfax rents thumbs. Both get filed under "logistics," and that filing error is where most bad comparisons in this sector begin — one of these businesses takes a decade and a balance sheet to copy; the other takes a client with orchestration software and 7,000 franchisees. Keep that asymmetry in your head through every section that follows.

The Six Arenas — Segments, Sizes, and Who Plays Where

Delivery demand comes from six arenas. Each grows at its own speed, pays its own rates, and carries its own risk of the client taking the work in-house.

3.1 Who plays where — the presence matrix

Read the matrix and two things jump out. First, Shadowfax and Delhivery make money in different places. Delhivery earns from mid-mile density — express and freight sharing the same trucks and hubs. Shadowfax earns from managing riders in cities. They clash head-on only in express parcels, which is also where both face the captives. Second, the captives are moving down into 3PL territory. Ekart and Amazon Shipping now sell spare capacity to outsiders. Yesterday's customer is becoming tomorrow's competitor — Delhivery's own management flagged this on the Q4FY26 call.

KEY TAKEAWAY — Where a company plays is its risk profile. Shadowfax sits in the two arenas where clients most want to take the work in-house — express and quick commerce. That is where the growth is, but it is a fair-weather book. Delhivery's spread across freight, warehousing and returns is ballast: the market forgets it exists when it panics about express, and forgets the risk when express is running hot.

Operational Intricacies — Why Each Segment Is a Different Sport

A 10-minute promise and a 2-day promise are not the same business at different speeds. The clock decides everything: batching, rider pay, and whether scale helps you or hurts you.

This is the deepest fact in the sector, and it cuts both ways for Shadowfax. The upward-bending q-comm curve is the reason platforms tolerate a 3PL at all. When your own cost per order rises with scale, a company that can pool demand from Blinkit, Zepto, Instamart and food delivery in the same neighbourhood has a utilization edge you cannot copy on your own. A Shadowfax rider who stacks a food order, a q-comm drop and a return pickup across one day earns three to four times what a single-app rider earns. He stays longer. Density improves. Cost falls. That pooling across apps is Shadowfax's real moat. The ">50% share" line is just its symptom.

4.1 What each arena actually demands

KEY TAKEAWAY — Specialization is real and it compounds. Express rewards address intelligence and sortation accuracy accumulated over billions of parcels; q-comm rewards rider-pool liquidity across demand sources; reverse rewards trained doorstep judgment. This is why the sector does not collapse into one winner — and why the correct comparison set for each player is arena-by-arena, never a single "logistics" league table.

The Insourcing War — Captives, Valmo, and the Outsourcing Floor

Every big client eventually asks the same question: why pay a middleman ₹7.6 a parcel? The ones with scale build their own network. What matters is how much volume can never be taken in-house.

Read the crossover carefully. Valmo did not beat the 3PLs with better technology. It beat them by deleting the spread. Its 6,000–7,500 franchise partners carry the assets and the labour risk. Meesho keeps only the software. The ₹7.6 a 3PL would have kept as profit goes back into Meesho's P&L as savings. It is the same asset-light trick Shadowfax played on Delhivery, played one level higher — the client cutting out the middleman who had cut out the middleman. Amazon (~85% in-house) and Flipkart (~90% through Ekart) reached the same place through sheer density. Quick commerce is the extreme case: when the product is a 10-minute promise, delivery is the product, so ~85% stays in-house by design.

5.1 The outsourcing floor — why 3PL demand cannot go to zero

OPINION (the insourcing pendulum has a return swing) — Insourcing is not a one-way ratchet; it is a capital-allocation decision that looks different at different points in the funding cycle. Meesho pre-IPO optimized for cost (build Valmo). Listed Meesho will optimize for reported margins and asset-lightness — and there were already "outsourcing recovery signals" in broker channel checks by early 2026. Zepto's IPO (₹8,010 Cr fresh issue, filed Jun-26, expected listing H2-CY26) forces the same discipline on q-comm: once public, burning capital on captive fleet expansion competes with showing a path to profit. Every logistics-heavy platform that lists is a marginal argument for outsourcing. The float is the 3PLs' friend.

KEY TAKEAWAY — The war is real but the map is settled. Clients above roughly 1.5bn parcels a year build their own network and chase Valmo's sub-₹40. Everyone else rents. The 3PL pool is the long tail plus the surge lanes plus the new launches, and it compounds with India's platform economy, not with any one client. Shadowfax's job, put plainly, is to move its revenue from the clients who insource (Meesho, Flipkart) to the ones who never will, before the move is forced on it. Shadowfax 360, the Amazon onboarding and the 100-store plan are that move. The pace of it is the stock.

Bottom-Up TAM — Rebuilding the Market Parcel by Parcel

Forget the $20bn TAM slides. The market a 3PL can bill is simple: parcels shipped, times the share the captives don't keep, times the fee per parcel. We count each piece.

6.1 The anchor: how many e-commerce parcels does India actually ship?

The cleanest anchor in the public domain: Meesho shipped ~2.7bn orders in FY26 (The Ken, May-26) and JM Financial pegs Meesho at ~30% of India's total e-commerce shipments ex-hyperlocal (Jan-26). That triangulates total e-commerce parcel volume at ~9.0bn parcels in FY26 — roughly 24–25mn parcels a day, consistent with RedSeer's "1bn parcels a month by 2030" trajectory (currently ~750mn/month). Composition estimate: Meesho ~2.7bn, Flipkart group ~2.2–2.4bn, Amazon ~1.6–1.8bn, everyone else (Myntra within FK, Ajio, Nykaa, D2C, ONDC, long tail) ~2.2–2.4bn.

6.2 The D2C and vertical pool — the customers who can never leave

The waterfall above hides its best part inside "everyone else." Roughly 2.2–2.4bn parcels a year come from vertical platforms and direct-to-consumer brands: Nykaa, Ajio, Tata Cliq, pharmacy and grocery verticals, and the long tail of D2C labels — around 250,000 of them ship through Shiprocket alone (~$4bn GMV a year). Myntra is the exception that proves the rule: it sits inside the Flipkart group, so Ekart carries most of it. The rest cannot build an Ekart. They are too small, forever. This is the one demand pool with no insourcing risk at all.

This is also where Shadowfax's client-concentration problem gets fixed, if it gets fixed. Shadowfax 360 (self-serve shipping for small brands), the CriticaLog purchase (jewellery, luxury, high-value electronics) and Prime Large (heavy parcels, 10,000 pin codes, up 170% in a year) are all aimed at the same thing: replacing Meesho-and-Flipkart revenue with hundreds of small clients who pay more and cannot leave for a captive. Brokers already expect the top-client concentration to ease to ~81% of e-com volumes by FY28. Watch that number more closely than the growth number.

6.3 Quick commerce: the TAM everyone is mis-sizing

Q-comm is quoted in GMV ($35–54bn by 2030-32), which is irrelevant to a 3PL. What a delivery company can invoice is the delivery leg. Build it up: FY26 order volumes are now disclosed — Blinkit 916.6mn, Zepto 640.2mn, Instamart 412.2mn (company filings via Outlook Business, May-26), JioMart ~1.6mn/day exiting Q3FY26 (RIL investor deck), plus Flipkart Minutes, BBNow and the tail — ~2.7–2.8bn orders in FY26, running at ~7.8mn/day in Jan-26 (RedSeer).

Two things follow from the store table. First, the store build-out is already inside our order projections — 10bn orders in FY31 is the same claim as 12–14,000 dark stores at ~2,000 orders a day. If the store capex cycle stalls (it is the biggest fixed-cost line in q-comm; Blinkit's estate alone runs an estimated ₹5+ Cr a month in rent), the order number stalls with it, and so does the outsourced delivery pool. Watch store openings the way you would watch cement dispatches. Second, the stores themselves are a market. Somebody has to staff, pick and run them. Shadowfax's 15-to-100 store plan is a bid for that work — store operations as a service, sold to brands and smaller platforms that want q-comm shelf presence without building anything. We have deliberately left this out of the ₹2k Cr pool; if it works, it is upside on top.

OPINION (the pool is small, the flow is huge) — A ₹2k Cr pool looks like a rounding error next to Shadowfax's ₹14.6k Cr market cap. But the pool is not the point. The growth of the pool is. It is order volume (compounding 40%+) times the outsourcing ratio (floored at 10–15%, and it rises whenever platforms chase new cities or need to show profits) times the per-drop fee (sticky). Hold RedSeer's trajectory and the outsourced pool alone triples past ₹6k Cr by FY29 with zero share gain. There is exactly one scaled independent bidder for it. Value the hyperlocal business as a royalty on q-comm order growth, because that is what it is.

6.4 Cannibalization — when a parcel becomes a drop, the 3PL loses two-thirds of it

Quick commerce does not only add orders. It also steals them. A bottle of shampoo that used to arrive as a next-day e-commerce parcel now arrives in ten minutes from a dark store. The categories are already moving: grocery first, then beauty and personal care, now fashion and small electronics — Blinkit sold iPhones last Diwali. For the shopper this is an upgrade. For a 3PL it is a pay cut, and here is the exact size of it.

So: yes, cannibalization is in the numbers. It is the reason our FY31 express base is 13bn and not 16bn, and it is the honest answer to why the whole 3PL pool grows at 9–10% while orders everywhere seem to be exploding. It also sets up the sector's cleanest swing factor. Today a migrated order pays a 3PL ₹7 instead of ₹22. If the q-comm outsourcing ratio rises from 15% to 30% — one funding winter, one post-IPO margin push — the same migrated order pays ₹14, and the cannibalization drag halves without a single new customer. The outsourcing ratio is the most leveraged number in this industry. Nobody publishes it. Zepto's IPO papers will be the first time we see it printed.

6.5 The definitional muddle — whose "market share" is real?

6.6 The pool in FY31 — what five more years does to the arithmetic

Project the same three terms forward — parcels, captive share, fee — and stop pretending precision. The honest anchors: RedSeer's "1bn parcels a month by 2030" puts e-com volume near ~12–13bn by FY31 (a conservative ~8% CAGR against the 15–20% history; if the Meesho-style value-commerce wave keeps running, 15%+ holds and you get 18bn — I carry that skew in the bull case, not the base). Q-comm order volume compounds at 40%+ for another two years before fading toward the mid-20s — call it ~10bn orders by FY31. Captive share creeps up in the base case because Valmo isn't finished and q-comm stays control-obsessed. Fees inflate barely at all; this is a deflationary industry that hands its productivity gains to clients.

Two things in that chart matter more than the totals. One: in the base case the pool compounds at 9–10% while parcels compound at 8% and q-comm orders at ~30% — the pool grows slower than the traffic, because captive creep taxes every incremental parcel. Anyone underwriting "logistics grows with e-commerce" is underwriting the wrong line. Two: the mix migrates. Q-comm goes from ~9% of the investable pool to ~21% by FY31 — the market drifts toward Shadowfax's arena even while express remains Delhivery's fortress. And run the endgame: extrapolate both listed names' revenue paths to FY31 (Delhivery ~₹19–20k Cr on its consensus trajectory, Shadowfax ~₹9–10k Cr on a fading 20%+ path) and they absorb ~₹28–30k Cr of a ~₹34k Cr pool. By FY31 the two stocks essentially are the market. You are not buying share gainers anymore; you are buying the pool's inflows — which is a cleaner, duller, and frankly better thesis than the one the sector is marketed on.

KEY TAKEAWAY — The entire investable for-hire pool — express plus q-comm outsourcing — is ~₹21–22k Cr in FY26, roughly $2.5bn. Delhivery and Shadowfax together already invoice ~₹14.7k Cr of it. This is a consolidated duopoly-plus-one (XpressBees) fighting over a pond, with the real upside coming from two flows into the pond: 15–20% underlying parcel growth, and any reversal of captive insourcing. Price the flows, not the headline TAM — and note where they lead: ~₹34k Cr by FY31 in the base case, with the q-comm slice more than doubling its weight (Exhibit 15).

The ₹53 Autopsy — Per-Transaction Economics

A platform pays a 3PL about ₹53 to move one parcel. Who keeps what inside that ₹53 explains the margins, the insourcing, and who survives.

Three things fall out of this. First, the margin ceiling is real and permanent. About 70% of the fee is labour and trucking bought at market rates. The 3PL is a middle layer earning a spread, and spreads do not turn into software margins. Company-level EBITDA tops out in the high single digits. Anyone modelling 15% is modelling a different business. Second, the maths shows why clients insource. A captive running its own last mile does not save the whole ₹53. It saves the 3PL's ₹7.6 spread, minus its own inefficiency. Valmo clears that bar because its franchise model runs the whole stack under ₹40 (Section 5). Ekart and Amazon clear it on density. Smaller platforms never clear it — which is exactly why they rent. Third, the q-comm transaction is a different animal. The 3PL bills ₹42–48 a drop and the rider takes ₹25–35, so the spread is thinner than in express. But there is no truck and no sort centre to pay for. Almost no capital, a decent return on a thin spread. Thin margin, endless turns.

KEY TAKEAWAY — The ₹7.6 spread is the bull case and the bear case at once. It is small enough that any client above ~1.5bn parcels a year saves money by building its own network, and small enough that everyone below that line saves money by renting. So the 3PL market is not everyone's logistics. It is, permanently, the logistics of the small, the seasonal and the new. India produces those clients faster than anyone insources them.

Operational Benchmarking — Pins, Fleet, Throughput

Networks are measured in pin codes, riders, trucks and turns of capital, not adjectives. Here is the scorecard — captives included, because they set the bar.

Read the table slowly. Delhivery's ~19,000 pins against Shadowfax's 15,600 looks like a small gap. It is not. The last 3,000 pins are the expensive ones — thin rural routes where only mid-mile density, or India Post, makes the sums work. That reach is what the Ecom Express deal bought Delhivery, and it is a real moat with national sellers. On the other side, Shadowfax doubled its orders in a year while its network grew only 10–30%. That is a density flywheel turning: more orders per touchpoint, more pay per rider, less churn, better service, more orders. A flywheel like this breaks in only one way — a mega-client walking out. That is what concentration risk means on the ground, not just in the accounts.

KEY TAKEAWAY — On operations, the gap between the two listed names is specialization, not quality. Delhivery owns reach and the mid-mile factory. Shadowfax owns urban density and rider liquidity, built for a tenth of the capex. The one number neither can answer is Valmo's sub-₹40. Until a listed 3PL beats it, that cost line — not any competitor — is the sword hanging over the sector.

OPINION (Margin note) — The most important cell in the scorecard isn't a Shadowfax or Delhivery number at all — it's Valmo's sub-₹40. When a management narrates network pride on a call (pins, hubs, riders, touchpoints), there's only one question worth asking back: what does it cost you to move one parcel, and is it under forty. Nobody listed clears that bar yet. Until someone does, every pin-code slide is decoration.

Year-Wise Projections — Testing the 40% Claim

Shadowfax has guided 38–40% growth for FY27 and management talks about holding rates like this for five to seven years. Delhivery's street numbers say mid-teens. Run both forward, year by year, and check them against the pool.

Start with what the 40% claim means in plain arithmetic. ₹4,202 Cr growing 40% a year for five years is ₹22,600 Cr by FY31. For seven years it is ₹44,300 Cr by FY33. Our base-case investable pool in FY31 is ₹34,000 Cr — express and quick commerce combined, everything a 3PL can bill in this country. So the seven-year version of the claim needs Shadowfax to become the entire market plus some, and the five-year version needs it to take two-thirds of the pool while Delhivery, XpressBees and the captives politely stand aside. Neither happens. The honest version: 40% is real for FY27 (guided, visible in Q1), and then the pool starts doing the talking.

The collision row is the point of this section. Even on our faded path — not management's — the two listed names are eating the whole base-case pool by FY31. Only three ways to square it: the pool is bigger than our base (the bull case, ₹47k Cr, needs the insourcing pendulum to swing back), or the two keep taking share from XpressBees, DTDC and the long tail until there is nothing left to take (mostly true, mostly done by FY29), or growth fades harder than anyone is guiding (the quiet base case for FY30 onwards). Practically: believe the 40% for FY27, price maybe two more years of 25–30%, and treat any five-year 40% story as a claim on a pool that does not exist. When a CEO says seven years of 40%, he is not lying — he is assuming the bull-case pool and telling you nothing about the middle years.

KEY TAKEAWAY — The projections and the TAM cannot both be right past FY29. That is not a reason to avoid the stocks — it is the timetable for when growth stories hand over to margin stories. Delhivery's handover already happened (that is what 238x on 17% growth means). Shadowfax's comes around FY29. Own it for the growth years; be positioned for the handover before the market reprices it.

Financials & Valuation — The Inversion the Street Missed

At IPO the market priced Shadowfax at 407x and broke the issue. Today it trades at 67x — a third of Delhivery — while growing three times faster. The risk moved into the client list, not the price.

How to read the multiple gap. The market is not being stupid giving Delhivery 238x. It is paying today for a margin ramp it can already see: 16–18% express service EBITDA delivered, ~16% ROIC in transport, overheads still to be spread. And it is charging Shadowfax for two real problems — 75% of revenue from two clients, and doubt about how long hyperlocal lasts. Our argument is with the width of the gap, not its direction. Per unit of near-term earnings growth, Shadowfax costs about a third of Delhivery, and its Q1FY27 print — margins up 204bps, guidance raised — attacks that doubt head-on. Gaps this wide usually close from both ends: Delhivery drifting toward the mid-100s as growth settles, and Shadowfax re-rating if, and only if, client concentration prints below ~60%.

KEY TAKEAWAY — Identical EV/Sales, wildly different P/E: the market has decided both revenue lines are worth the same and only Delhivery's will ever carry real margins. That is the single most attackable assumption in the sector's pricing — Shadowfax's service EBITDA per order (₹7.6, 14.4%) is already inside Delhivery's normative band. If corporate overhead scales the way FY26→Q1FY27 suggests, the P/E gap is a coiled spring.

Forward View & House Call

Own the pool, or own the bet that the pool grows. Delhivery is the first. Shadowfax is the second. Hold them accordingly.

11.1 The house call

Delhivery — the diversified core. The only Indian name with all three legs of the parcel value chain plus PTL, warehousing and software/NBFC optionality, now consolidating a thinning field (Ecom Express absorbed; XpressBees the next likely domino via IPO-or-merger). You are paying ~238x trailing for a ~2% ROE business — which is really paying for the express margin band (16–18%) to reach the company level. Hold it as sector ballast; add on multiple compression, not on momentum.

Shadowfax — the leveraged claim on outsourced velocity. Cheaper than the market believes (67x trailing, ~61x broker FY28E), faster than the sector, and structurally long the three durable demand pools — q-comm growth, the D2C long tail, and surge lanes. The position is a bet that client concentration falls faster than insourcing rises. Q1FY27 said the bet is working; the July VC-exit blocks say you will get supply-driven entries. Size it as the high-beta satellite, never the core, until two-client concentration prints below ~60%.

The unlisted watchlist. Shiprocket is the structurally safest business model in the sector (asset-free take-rate on the long tail — immune to the insourcing war) and the most interesting future listing. XpressBees is the consolidation piece. Porter is adjacent. Ekart's rumoured external-3PL opening is the wildcard that could compress everyone's express pricing.

11.2 What we are watching (the falsifiers)

WATCH — Zepto IPO prospectus & post-listing conduct (H2-CY26): first hard disclosure of a q-comm player's delivery cost stack and outsourcing mix; lets us replace the ~15% assumption with a printed number.

WATCH — Shadowfax client-concentration disclosure each quarter: the single number that re-rates or breaks the stock; watch for below 60%.

WATCH — Meesho AGM / earnings commentary on Valmo share: plateau vs push-to-80% decides Shadowfax's express base case.

WATCH — Ekart / Amazon Shipping external-3PL pricing: captives selling spare capacity is a price war on the ₹48–51 benchmark.

WATCH — Labour-code implementation costs: ₹/order inflation hits the ₹21–24 rider layer; tests who has pricing power to pass it through.

WATCH — Shadowfax dark-store count vs 100-store FY27 guide: proxy for the q-comm enablement pivot beyond pure delivery.

FINAL TAKEAWAY — India crosses a billion parcels a month around 2030. The captives have already taken the easy two-thirds. What is left for hire is the hard third — festive peaks, long tails, new launches and ten-minute promises. That is a smaller market than the TAM slides claim and a better business than the bears admit: two listed players, almost no capital in the last mile, and a floor under demand that physics and franchise economics both hold up. Own the factory (Delhivery) for the decade. Rent the rocket (Shadowfax) while the q-comm pool compounds. And read every client-mix disclosure like the risk report it is.

^{Sources & confidence. Live prices/multiples: NSE/BSE market data, 3 Aug 2026. Company data: Shadowfax Q3FY26/Q4FY26/Q1FY27 earnings presentations and filings; Delhivery FY26 filings and Q4FY26 earnings call; Eternal FY26 shareholder letter; Zepto UDRHP coverage. Industry: RedSeer, JM Financial via YourStory (Valmo economics), DAM Capital via Business Today/Outlook Business (q-comm store economics), The Ken (Meesho/Valmo), Inc42, QCommerce Guide, Xylem Investment, The Silk Road Nexus, Base.com, Mann Ashar Shadowfax model. All bottom-up pool constructions, reconciliations, floor estimates and scenario bands are the author's own and flagged as such in each exhibit. Captive-share figures (Amazon ~85%, Flipkart ~90%, q-comm ~85%) are broker estimates, not audited disclosures.}

^{Disclaimer (SEBI). This document is an independent analytical research note prepared for informational and educational purposes only, published at equitybypiyush.substack.com. It is not investment advice, an offer, or a recommendation to buy or sell any security, and does not consider any reader's financial situation, objectives, or risk tolerance. Equity investments are subject to market risks; estimates and scenarios herein are analytical constructs, not forecasts of certainty. The author is not a SEBI-registered investment adviser or research analyst; please consult a SEBI-registered adviser before acting. The author holds no position in the securities mentioned as of the publication date.}

What I believe, upfront, in six lines:

• The world runs 420 GW of nuclear, is building 78 GW more, and the new buyers are hyperscalers with better credit than most governments. This revival is financed, not just announced.

• Uranium already repriced. American SMR stocks over-repriced. The Indian supply chain has barely started, and its order flow is the most visible of the three.

• India’s 100 GW by 2047 splits into two claims. 22.5 GW by FY32 is sanctioned and funded; I underwrite that. The rest depends on the SHANTI Act’s liability rules working; I track that.

• The in-hand programme alone generates ₹25,000-45,000 crore of orders a year through FY33 on my math. Suppliers book this on construction, years before any electron flows.

• The best pool sits in nuclear-island equipment, where qualification takes years and maybe six Indian companies hold membership cards.

• What flips the story bigger: the first privately funded reactor sanction, and the first Indian component win in a Western tender. No dates exist for either. That’s precisely why they’re worth watching.

SECTION I

How nuclear works, in plain words

Ninety seconds of physics buys you the whole sector. A reactor is a kettle with a very exotic flame.

The chain of events, start to finish:

• Fission. A uranium-235 atom absorbs a neutron and splits. The split releases heat, plus two or three fresh neutrons.

• Chain reaction. Those neutrons split more atoms. A moderator (ordinary water, or heavy water in India’s reactors) slows the neutrons so the splitting sustains itself. Control rods of boron or cadmium absorb neutrons; push them in, the reaction slows, pull them out, it speeds up. That’s the throttle.

• Heat to steam. A coolant carries the heat away from the core and boils water into steam, usually through a separate loop so the radioactive side never touches the turbine side.

• Steam to power. Steam spins a turbine, the turbine spins a generator. From this point it’s the same machine as a coal plant. That’s why retiring coal sites make natural reactor sites: the back half of the plant, the grid bay, the water and the workforce already exist.

• Fuel maths. One kilogram of uranium holds about as much energy as three thousand tonnes of coal. Which is why fuel is a small share of nuclear’s cost and the concrete-and-steel is nearly all of it.

The main machine types, one line each:

Three videos I’d actually share with readers (each explains this better than text can):

• Kurzgesagt, Nuclear Energy Explained: How Does It Work? — the five-minute animated version: youtube.com/watch?v=rcOFV4y5z8c

• TED-Ed, How do nuclear power plants work? (Ramana & Saini) — the balanced classroom version: youtube.com/watch?v=R7WPEYGr1Vs

• Animagraffs, How Nuclear Power Works — a full 3D teardown of a modern plant, the best visual of the lot: youtube.com/watch?v=PRWwXeRIvoI

KEY TAKEAWAY: Two design facts carry the whole India story. PHWRs run on natural uranium, so India never queued for enrichment. And heavy-water reactors scale down gracefully, which is why India’s SMR is a shrunk version of a machine it has built sixteen times, not a paper design.

SECTION II

The value chain, link by link

Nine links from mine to megawatt. Everything later in this note, every profit pool, every stock, lives on one of these links.

Exhibit 1: The chain on one page. Green means India is present and credible, amber partial, red absent. (Structure and judgments mine; link facts from WNA, IEA, company disclosures)

What each link means for an investor, fast:

• Mining (link 1): a commodity market run by Kazatomprom and Cameco. Cyclical, already repriced, investable only through global names. India’s play here is contracts, not mines.

• Conversion and enrichment (2): the true global choke. Rosatom holds ~40% of enrichment; the US makes under 1%. India sidesteps the whole link because PHWRs burn natural uranium. Most countries can’t say that.

• Fuel fabrication and heavy water (3): NFC Hyderabad and the Heavy Water Board make India fully self-reliant for its own fleet. A quiet annuity, government-owned, not directly investable, but strategically priceless.

• Reactor design and IP (4): perhaps eight entities on earth own buildable large-reactor designs. BARC/NPCIL is one of them. Design ownership is why India’s costs are its own and its export option exists at all.

• Nuclear island equipment (5): the margin link. Reactor vessels, steam generators, calandrias, end shields, precision internals. Qualification takes years, the vendor club is tiny, and this is where I’d concentrate research hours. The Indian club: L&T, BHEL, Walchandnagar, MTAR, Godrej.

• Turbine island (6): conventional engineering. BHEL supplies 56% of India’s installed nuclear turbine capacity. Solid, established, less scarce.

• Civil and EPC (7): the biggest single cheque, the most ordinary margin. HCC has built civil works behind ~60% of the installed fleet; L&T HCI won the Kudankulam 5&6 package.

• Operation (8): a regulated annuity running 40-60 years per plant. NPCIL’s monopoly, cracked open only via NTPC’s JV today and privates tomorrow.

• Back-end (9): waste, reprocessing, decommissioning. India is one of very few countries reprocessing at scale, and PFBR is the bridge to burning thorium. Deferred economics, strategic weight.

KEY TAKEAWAY: Notice the pattern the colours make: India is red or amber exactly where it doesn’t need to be present (enrichment, mining) and green where the money and the moats are (design, equipment, fuel self-reliance, construction). Very few countries’ chains colour in that conveniently. That’s not an accident; it’s sixty years of the three-stage programme doing its job.

SECTION III

Why nuclear came back

Nuclear didn’t stall for forty years because of physics. It stalled because nobody would finance it. What changed is the buyer.

• The stock: 420 GW operating across 30+ countries; 78 GW under construction in fifteen. 2025 looked flat (3 GW added, 3 retired) but ~15 completions are expected in 2026, the best year in decades.

• The money: investment running ~$65 bn a year, heading to ~$120 bn by 2030 if pledges hold. The US executive order targets 400 GW by 2050; Washington and Ottawa each announced ten-reactor programmes in June 2026.

• The three drivers, compounding: energy security (Europe’s 2022 lesson), decarbonisation (the only firm zero-carbon source proven at scale), and AI load (Section IV). Each failed alone for years. Together they’ve moved actual capital.

• The pattern I’d trade on: money and policy turn two to four years before capacity charts do. Solar 2010 and BESS 2020 both traded exactly this way. Waiting for the capacity curve means paying after the supply-chain stocks have moved.

Exhibit 2: The stock, the pipeline, and who holds it. (Source: IEA Global Energy Review 2026)

SECTION IV

Who builds, whose designs — and the AI bid

One statistic reorganised my view of this sector, and one buyer class changed its financing forever.

Exhibit 3: Construction starts by design origin, past decade. 2025’s ten starts: nine China, one Russia. (Source: IEA)

• China: ~60 GW operating, ten units approved in one $27 bn sweep, largest nuclear market by 2030, and the world’s first commercial land-based SMR (Linglong One) this year.

• The flip side: everyone who isn’t China or Russia must rebuild atrophied capability — forgings, qualified welders, people who’ve built one before. India kept its muscle through the entire global winter. Outside China and Russia, that makes India’s PHWR ecosystem close to unique, with an export option written into it.

The AI bid, in numbers:

• Thirteen hyperscaler deals, ~9.8 GW committed, twenty-year tenors. The queue-jumping logic: 2,600+ GW stuck in US interconnection queues with five-year waits.

• Restarts deliver first: Microsoft’s Three Mile Island restart (Crane, 835 MW) hits full power in 2027, a year early. Holtec’s Palisades runs in parallel.

• Fleets deliver now: Amazon’s 1.92 GW Susquehanna deal; Meta’s blocks with Vistra and Constellation.

• SMR cheques deliver in the 2030s, if at all: Google-Kairos (500 MW), Amazon-led $700 mn into X-energy, Meta’s 5.2 GW across Oklo and TerraPower.

• India read-through: the Bharat Small Reactor captive model is the domestic version of this structure. Indian data-centre and industrial load will want the same product. Someone signs the first Indian corporate-nuclear deal within a few years; keep a list of who’s shortest of firm power.

Exhibit 4: Committed GW by buyer, mid-2026. (Source: smrintel tracker; company statements)

KEY TAKEAWAY: Near-term cash sits with restart owners and operating fleets. SMR equity is venture risk in utility clothing. And the scarcity signal in the 94% chart is India’s export brief, not China’s victory lap.

SECTION V

SMRs, properly explained

SMR means Small Modular Reactor. Each of the three words is doing work, and the middle one carries the whole investment thesis.

• Small: under ~300 MW of electrical output, against 700-1,600 MW for conventional plants. Small enough for a factory campus, a steel plant, a data-centre park, or a retiring coal site’s footprint and grid bay.

• Modular: the thesis word. Instead of pouring and welding everything on site over a decade (where nuclear’s cost blowouts live), you manufacture standard modules on a production line and ship them out. Series production is supposed to cut costs 30-50% by the nth unit. Nobody has proven that yet. That single unproven claim is what every SMR valuation rests on.

• Reactor: same fission physics as Section I, packaged smaller, with passive safety designs (many can cool themselves without pumps or operators in a failure).

• Why buyers care: smaller cheques (hundreds of crores, not tens of thousands), faster permits in theory, siting next to demand, and a size that matches a single industrial customer, which is exactly what hyperscalers and Indian captive users want.

Where the race actually stands:

KEY TAKEAWAY: If I could hold only one SMR exposure it wouldn’t be a stock. It would be the Indian component makers supplying BSMR-200 if it proceeds: paid on orders, zero design risk.

SECTION VI

Fuel

Fuel repriced first, as it always does. I treat it as ballast in this theme, not the upside.

Exhibit 5: Uranium’s round trip. Spot points approximate; the term price is the tell. (Source: Cameco series via ANS; INN; Trading Economics)

• Demand: WNA sees requirements up 28% by 2030, possibly doubling by 2040 as the construction pipeline converts to fuel contracts.

• Supply: Kazatomprom’s 2026 guidance has swung between a cut and a raise across announcements; deliberate opacity, in my reading. Acid shortages and depleting Tier-1 mines keep the deficit alive into the 2030s. Cameco talks $140-150 contract ceilings.

• The real choke is enrichment: US capacity under 1% of the world, Russian imports banned from 2028, HALEU (the SMR fuel) barely commercial. DOE’s $2.7 bn rebuild is years from mattering.

• India’s quiet competence: a ~$2.8 bn ten-year Cameco agreement inside the 2026 Canada package, renewed Australian supply, Kazakh and Russian contracts, UCIL at home, and no enrichment dependence at all for the PHWR fleet. Fuel for the 100 GW plan is being contracted a decade early. Nobody writes about this, and it’s the most professionally run piece of the entire programme.

SECTION VII

Nuclear relative to everything else

The question every reader will ask: why pay ₹21 crore a MW when solar costs four? Because a MW is not a MWh, and after sunset the comparison inverts.

Exhibit 6: Where nuclear sits in the world’s grids. (Approximate 2024-25 generation shares; author-compiled from WNA/IEA data)

Exhibit 7: The honest comparison: capex, how much each technology actually runs, and what the power costs. Ranges and figures are author-compiled approximations for India, mid-2026; nuclear tariff band runs from Kudankulam’s existing ~₹4.1 to my ~₹5-6 estimate for new PHWRs. (Author-compiled)

How I’d frame the relative maths for readers:

• Per MW, nuclear looks absurd: ~₹21 crore against solar’s ~₹4. That’s the chart critics use.

• Per MWh delivered, the gap collapses: nuclear runs ~83% of the hours in a year for 40-60 years; solar runs ~22% for 25. One nuclear MW produces roughly four times the annual energy of a solar MW, at the hours the grid actually struggles with.

• The right competitor isn’t bare solar, it’s solar-plus-storage: and against firm round-the-clock renewable structures (which price well above ₹4.5-5/kWh once you firm them properly for all hours and seasons), new nuclear at ₹5-6 is competitive, not crazy. Existing nuclear at ₹4.1 is outright cheap firm power.

• Scale of the bet, relatively: the 100 GW build at TERI’s ₹23-25 lakh crore is about twelve semiconductor missions, or nine times the entire solar+BESS equipment market to 2030. India has never proposed anything this size in energy.

• And the mix context: India is at ~3% nuclear generation share. The 100 GW plan takes it to roughly 8-10% of a much larger 2047 grid. France sits at 65%. Even full success leaves India a renewables-first grid with a nuclear spine, which, for the record, is the correct design.

Exhibit 8: Programme scale against everything else this series has covered. (Sources: TERI; prior deep-dive figures)

KEY TAKEAWAY: Sell readers the right sentence: nuclear is not cheap capacity, it is cheap firm energy bought in forty-year instalments. Comparing it to daytime solar on capex is how the sector gets mispriced, and mispricing is the whole reason to write about it.

SECTION VIII

India: the bankable leg and the hopeful leg

I underwrite 22.5 GW. I track 100 GW. Mixing those two jobs up is the standard analytical error in Indian nuclear commentary.

Exhibit 9: The two legs, drawn the way I size them. (Source: DAE replies in Parliament, Jul 2026; my framing)

The bankable leg — projects in hand:

• 24 reactors operating, 8.8 GW. Workhorse: the indigenous 220/540/700 MW PHWR line; Kudankulam’s VVERs are the large-unit exception.

• ~6.6 GW under construction: remaining Kudankulam units, the 700 MW fleet-mode series, GHAVP Haryana.

• ~8 GW sanctioned, including the ten-reactor fleet approval and ASHVINI’s 4×700 MW Mahi Banswara (foundation stone laid September 2025, first unit targeted FY33).

• Sum: ~22.5 GW by FY32, on a design Kakrapar and Rajasthan have already proven in serial production, at roughly $2-2.5 mn per MW, a fraction of Western cost. Fleet mode is the whole trick, and India has been running it for years.

The hopeful leg — the roadmap beyond:

• +32 GW of NPCIL PHWRs and LWRs after FY32, per DAE’s parliamentary reply.

• +~46 GW from PSUs, states, JVs and private entrants under the new law, to reach 100 by 2047.

• Coal-site conversion is the smartest siting idea in it: land, water, grid and workforce pre-exist.

• The honest constraints: the private wedge needs liability rules that vendors will actually sign under, and 4-5 GW of commissioning a year through the 2030s is a run-rate only China has sustained. I want it. Wanting isn’t underwriting.

• For a 5-7 year horizon the debate barely matters: the bankable leg alone keeps supplier order books growing through FY33. The 100 GW is free optionality on a trade that works without it.

SECTION IX

The reform stack

India fixed the actual constraint. Nuclear’s binding problem was never physics or fuel; it was law and finance, and SHANTI goes straight at both.

On liability, both sides deserve a fair line. The 2010 law made India effectively uninsurable for global vendors, so softening it is the admission price for foreign technology and private capital. The counter, that victims’ recourse shouldn’t be traded for investment, isn’t frivolous. The narrow question I care about: do the final rules give a supplier’s board enough certainty to sign? Jaitapur and EDF are the first live test.

KEY TAKEAWAY: The first privately funded reactor sanction is this sector’s Micron moment: the event that converts law into pipeline. No date exists. Markets don’t schedule their catalysts.

SECTION X

Profit pools, mapped onto the chain

Now that the chain exists (Section II), the money question becomes concrete: a twin-unit plant is a ₹22,000 crore shopping list, and each line of it belongs to a specific link and specific companies.

Exhibit 10: Capex anatomy of a twin 700 MW plant, mapped to chain links 5-7. Recent sanctions run ₹20,000-25,000 cr a pair; splits are industry-typical, my estimates. (NPCIL sanction disclosures; standard cost structures)

• Civil (~25%, link 7): the largest single contract, contested by big EPC houses. Volume, ordinary margins.

• Nuclear island (~31%, link 5): the margin pool. Vessels, steam generators, calandrias, end shields, precision internals. Years of qualification to enter; a handful of Indian members. Pricing power lives here.

• Turbine island (~15%, link 6): established suppliers, BHEL’s stronghold. Moderate margins.

• BOP, electricals, I&C (~18%): spread across dozens of mid-size vendors; where smaller listed names take share.

• The annuities (links 3 and 8): O&M on a fleet going 8.8 → 22.5 GW compounds to roughly ₹1,500-2,000 cr a year by the early 2030s on my math; fuel and heavy water similarly. Small now, recurring for forty years, and valued by nobody yet.

Exhibit 11: The pools over FY27-FY33, with deliberate ranges. All author estimates, parameters in text. (My model)

• Sizing: 13.7 GW of in-hand build at ₹17.5-25 cr/MW is a ₹2.4-3.4 lakh crore programme; equipment and systems take ₹1.4-2 lakh crore of it after civil.

• Next wave: early packages from the 32 GW start landing before FY33 if the mission holds momentum.

• Exports, the wildcard: every Western restart and new build needs non-Chinese components. Even ₹10-25k crore of Indian wins would transform two or three companies’ earnings mixes.

• My pool ranking: nuclear island first (scarcity, pricing power), exports second (small base, huge option), civil third (volume, no moat), annuities the sleeper (mispriced because they’re small today).

SECTION XI

The 5-7 year order flow and catalysts

Suppliers get paid on orders, and orders front-run commissioning by three to six years. FY27-FY33 is the awarding window for everything headlined to 2035.

Exhibit 12: My order-flow model for the in-hand programme. The level is indicative; the shape is the point. (My model)

• Civil dominates early; nuclear-island and long-lead equipment overtake from FY29 as fleet-mode units enter fit-out. Those are the high-margin awards.

• A private or BSR sanction anywhere in FY29-31 adds a step nobody is modelling.

• The O&M base grinds up with every unit that enters service, regardless of everything else.

SECTION XII

The equity lens: companies, numbers, stances

Names with numbers, not names with adjectives. Every figure below is from disclosures or reported coverage as of mid-2026; my stances sit beside them, clearly separated.

Exhibit 13: How I map the ways to own the theme: repricing done vs demand visibility. Positions are my judgment, mid-2026. (My assessment)

Two flags I’d print in bold for readers. The thorium-theme basket has already run up to 225% in a year in places, so parts of this story are priced; entry discipline matters more than discovery now. And every eventual operator still sells to DISCOMs or captive industry, so tariff absorption and payment behaviour cap realised returns exactly as they do in solar. The supply chain is partly insulated from that; developers aren’t. One more reason picks-and-shovels rank above operators in my ordering. None of this is a recommendation; the names illustrate chain links, and the stances are analytical positions I expect to update.

Closing view

What I’m carrying out of this note, as pointers:

• Demand for firm clean power went contractual the day hyperscalers started signing twenty-year deals. That bid has better credit than most sovereigns and isn’t leaving.

• Outside China and Russia, people who can actually build reactors are scarce. India kept its capability alive when nearly everyone else let theirs rot. Scarcity meeting returning demand is the oldest setup in the book.

• India’s programme finally has the three missing pieces: funded projects, proven designs, a law that lets capital in. The 22.5 GW is bankable; the 100 GW is tracked optionality.

• My research hours go to the nuclear-island suppliers and the export watchlist. My calendar is Section XI. My map is Exhibit 13.

• When the numbers move, I’ll update the note, not defend it.

Data as of late July 2026. Estimates and models are labelled and mine. Conflicting official signals, Kazatomprom’s guidance and PFBR’s status among them, are flagged rather than smoothed over.

Sources

IEA Global Energy Review 2026 · DAE replies in Parliament, Jul 2026 · Deccan Herald, Down To Earth (SHANTI, mission) · TERI (₹23-25 lakh cr) · WNA, Nuclear Business Platform · smrintel trackers, Partnership for Global Security (hyperscaler deals, SMR status) · ANS, INN, Sprott (uranium) · Company-level: Goodreturns/expert coverage (BHEL turbine share, MTAR ₹650 cr nuclear book), Equitymaster (L&T NPCIL MoU, EDF MoC; NTPC/ASHVINI, MBRAPP Sep-2025), Torus Digital (HCC 60%/₹3,800 cr, MTAR book detail), News9 (BHEL header assemblies, Power Mech ₹18,300 cr, NTPC subsidiaries and site studies), Walchandnagar company disclosures (Class-I prequalification, component list), Business Today (thorium basket returns), Business Standard (L&T KKNPP 5&6, steam generator history) · Videos: Kurzgesagt, TED-Ed, Animagraffs as linked in Section I · Capex splits, order-flow model, pool sizes, tariff band for new PHWRs and the priced-vs-visibility map are my estimates and judgments. Nothing here is investment advice.

Solar’s constraint is no longer cost or capacity — it is the clock. Storage is how the clock gets solved, and India has just committed to solving it at scale.

The world added 664 GW of solar in 2025, and for the first time the industry is guiding to a flat year in 2026 — not because demand for clean power fell, but because grids are struggling to absorb what has already been built. Integration, not generation, is now the gating factor. That is the single most important context for everything in this note.

India had its best solar year ever. FY26 additions reached 44.6 GW — nearly double the previous record — taking cumulative capacity past 150 GW in March 2026 and 162 GW by June. India overtook the United States as the world’s second-largest solar market by annual installations. Rooftop finally moved, with PM Surya Ghar covering over 47 lakh households. Manufacturing scaled even faster than deployment: module capacity of 172–210 GW now runs at roughly 40% utilisation, while cells remain the thin part of the chain — the same pattern of assembly-first, upstream-later that runs through every Indian manufacturing story.

Storage is where the next five years of value creation sit. Globally, 307 GWh of batteries were added in 2025 (up ~50%), with China turnkey systems touching $63/kWh. In India, more than 130 GWh of storage tenders were floated in 2025 alone; discovered two-hour BESS tariffs have fallen ~86% since 2022 to ₹1.48 lakh/MW/month; the Cabinet approved a second, eight-times-larger VGF round (₹5,400 crore for 30 GWh) in May 2026; and the CEA’s requirement of ~236 GWh by 2031-32 implies a 25-fold build-out from what will be commissioned in 2026. The caution that belongs beside the excitement: tariffs have fallen twice as fast as battery costs, so a meaningful share of awarded capacity is priced for perfection — execution and balance sheets will separate winners from casualties.

For investors, the pairing matters more than either half. Solar developers are becoming storage developers; module makers are adding cell and battery lines; the tender pipeline is migrating from plain solar to solar-plus-storage and firm-and-dispatchable (FDRE) structures with real penalties. The capital-goods layer — cells, containers, power electronics, EPC — is where the volume growth is most visible and least priced.

Introduction

Every gigawatt of cheap midday solar makes the evening more valuable. That transfer of value — from generation to flexibility — is the trade.

Solar is the cheapest electricity most of the world has ever produced, and it all arrives in the same six hours. As penetration rises, midday power prices sink toward zero while evening prices firm — the value of a solar plant erodes precisely because solar succeeded. Storage reverses that erosion: it buys the cheapest hours and sells the most expensive ones, converts variable energy into firm capacity, and lets a grid retire the diesel and delay the wires.

This note reads as a pair: Sections I–III cover solar — the global plateau, India’s record year, and the manufacturing ledger. Sections IV–VI cover storage — the global surge, India’s tender machinery, and the economics with their risks. A companion frame worth keeping in mind: grid capex and DISCOM health sit underneath both stories; where relevant we flag them, and they deserve their own note.

STRATEGIC OBSERVATION: Solar and storage are converging into one asset class. The market still prices them as two. That gap closes over the next two to three years — through FDRE tenders, hybrid PPAs, and manufacturers integrating across both — and repricing follows convergence.

SECTION I

The Solar Century Hits Its First Speed Bump

664 GW in one year — and the industry’s own association says the next constraint is integration, not installation.

Global solar additions reached 664 GW in 2025, another record, with Europe adding 81.6 GW and India overtaking the US as the second-largest market. The remarkable part is the guidance: 2026 is expected to be roughly flat — the first pause in a decade of 25%+ compounding. The reason is not demand for energy; it is the system around the panels. Midday saturation, negative prices, curtailment and slow grid connections are now the binding constraints in China, Europe and parts of India alike. SolarPower Europe’s own framing captures the shift: scaling solar is no longer about deploying capacity but about integrating it — grids, batteries and flexibility are the urgent investments.

Exhibit 1: Global annual solar additions, 2021–2026F. 2025 set the record; 2026 guides flat as integration becomes the constraint. (Source: SolarPower Europe GMO 2026; earlier years industry data, approx.)

KEY TAKEAWAY: The solar plateau is not a demand problem — it is the market’s signal that value has migrated downstream, to storage, grids and flexibility. Read it as the starting gun for the storage decade, not the end of the solar one.

SECTION II

India’s Record Year

44.6 GW in twelve months — the target was 34. India is now the world’s #2 solar market and still accelerating.

FY26 was the year India’s solar machine hit full speed. Additions of 44.6 GW grew 87% over FY25’s 23.8 GW and beat the government’s own target by a third. March 2026 alone saw 6.65 GW — the largest single month ever. Cumulative capacity crossed 150 GW at March 2026 and stood at 162.15 GW by June 30 (a 57-fold rise from 2.82 GW in 2013-14). Total renewable capacity reached 275 GW, and non-fossil capacity 283 GW — third in the world.

What drove it: the completion of projects from MNRE’s 50 GW-a-year bidding trajectory begun in 2023; a rush to beat the ISTS-waiver deadline of June 2025; a strong C&I open-access segment; and — the newest leg — distributed solar. Rooftop added 8.7 GW (up 69%), powered by PM Surya Ghar, which has covered 47.26 lakh households with ₹14,771 crore of central assistance disbursed. Rajasthan (12.1 GW) and Gujarat (9.0 GW) took 61% of FY26’s ground-mounted additions between them.

Exhibit 2: India’s solar trajectory — annual additions and cumulative capacity, FY22–FY26. (Source: MNRE / JMK Research; cumulative at fiscal year-end)

Exhibit 3: Where FY26’s capacity landed — state and segment split. (Source: JMK Research, MNRE)

The concentration deserves a flag rather than a cheer: seven states hold ~85% of installed capacity, and the same two that dominate solar also lead the grid-congestion and curtailment lists. India’s solar geography is running ahead of India’s wire geography — which is exactly why the storage sections of this note exist.

The decade ahead: demand outlook to 2035

The forward numbers matter more than the record year. Industry and company assessments (MNRE, SEIA, BNEF-derived, via the Waaree Q1FY27 deck among others) put global annual additions at ~860 GW by 2030 and ~993 GW by 2035 (~4% a year from here), while India compounds at ~10% — roughly 72 GW a year by 2030 and 100+ GW by 2035. Two structural demand layers are new since last cycle: data centres (physical-infrastructure spend grew 28% YoY to $12 bn in Q1’26 alone) and grid investment itself (~$100 bn planned in Indian transmission, storage and pumped hydro), both of which buy firm clean power, not bare solar. H1 2026 evidence supports the India trajectory: 26 GW added versus 18 GW in H1 2025.

Exhibit 3A: Annual solar additions outlook, 2025-2035. Projections, not fact — shown as directional. (Source: MNRE, SEIA, PIB, industry/company assessments via Waaree Q1FY27 presentation)

KEY TAKEAWAY: India’s solar volume problem is solved; the machine now reliably delivers 40+ GW a year. The open questions have moved — to where the power flows after 6 pm, and to who makes the equipment. Those are the next two sections.

SECTION III

The Integration Problem: Why Solar Needs a Partner After 6 PM

India’s peak demand now arrives after sunset. Every policy lever of the last two years — storage mandates, grid-forming inverters, transmission plans — is a response to that one fact.

The shape of the problem is simple. Solar generation peaks at noon; India’s demand peaks in the evening. As solar’s share grows, the grid’s net load (demand minus solar) sinks deep at midday and ramps steeply at dusk — the “duck curve” that California made famous now shows up in Indian state load despatch centres, sharpest in the two states with the most solar, Rajasthan and Gujarat, which already live with evacuation constraints and curtailment. Curtailment is the quiet killer of project returns: every megawatt-hour a plant is told not to generate is revenue the tariff assumed. This is why capacity additions alone stopped being a sufficient measure of progress around 2024-25.

The policy response has been unusually coherent, and it is worth listing as a stack because each piece attacks one layer of the problem:

KEY TAKEAWAY: Integration is no longer a footnote — it is the procurement design. The 2-hour ESS mandate alone converts every future GW of utility solar into ~2 GWh of storage demand, which is how the BESS runway in Sections VI-VII becomes contractual rather than aspirational.

SECTION IV

Making It at Home: The Manufacturing Ledger

India now has more module capacity than annual demand — the value migration upstream, to cells and ingots, is the part that is still being built.

Module nameplate capacity stands between 172 GW and 210 GW depending on the count (JMK vs industry tallies — we show the band rather than pick a side), against FY26 demand of 44.6 GW. Utilisation is around 40%, partly a technology transition (MonoPERC lines giving way to TOPCon) and partly plain overbuild ahead of ALMM enforcement. Cells are the thin link — roughly 25–27 GW class ~est. — which is precisely why ALMM List-II (domestic cells, mandatory for government schemes from June 2026) matters so much: it converts policy into cell-line economics.

Exhibit 4: The manufacturing ledger — module capacity vs demand vs cell capacity. (Source: JMK/industry reports; cell figure estimated, flagged)

The upstream ledger: cells, wafers, polysilicon, glass

The module story is finished; the value migration now runs through four upstream gates, each with its own policy protection and its own build-out underway:

Add the demand-side shields — 20% basic customs duty on module imports, PLI for high-efficiency modules, PM Surya Ghar (~US$9 bn) and PM-KUSUM pulling certified domestic product — and the design becomes visible: India is fencing each upstream step in sequence, exactly as it fenced modules first. For investors, each fence creates a 3-5 year protected-margin window for whoever builds inside it first; the glass ADD plus the 5,400 TPD gap is the clearest current example.

KEY TAKEAWAY: Modules are done; cells, wafers and glass are the investable frontier. Watch utilisation and export lanes, not nameplate announcements — nameplate is now the least scarce thing in Indian solar.

SECTION V

The Equipment Stack Around the Panels

A solar+BESS plant is one-third panels by value. The other two-thirds — inverters, transformers, EPC, and increasingly hydrogen — are markets that double to twenty-fold by 2035, with less crowding and better margins.

Exhibit 5A: Adjacent markets, 2025 vs 2035, global. (Source: Research & Markets, Precedence, Mordor, BNEF via Waaree Q1FY27 deck — company-compiled, directional)

KEY TAKEAWAY: The panels are the crowded trade; the stack around them is not. Transformers and inverters have demand visibility from three directions at once (solar parks, BESS, RDSS) — and unlike module-making, nobody accuses these of overcapacity.

SECTION VI

The Global Storage Surge

307 GWh added in 2025, $63/kWh turnkey systems in China, LFP at ~95% share — storage is repeating solar’s cost curve, faster.

Global storage additions reached 112 GW / 307 GWh in 2025 — up ~48% — and BNEF forecasts 158 GW / 459 GWh for 2026, even as solar and wind slow. China (61 GW / 173 GWh) and the US (18 GW / 55 GWh) were ~70% of the world. BESS shipments grew 75.5% to 421 GWh, with 600 GWh projected for 2026; storage-cell production is running ahead of demand, so no supply constraint is in sight. A new demand engine has appeared alongside grids: AI data centres, which are turning on-site storage from a UPS afterthought into core infrastructure.

Exhibit 5: Global storage additions and the 2025 geographic split. (Source: BloombergNEF Energy Storage Outlook H1-2026; InfoLink; 2024 estimated from growth rates)

Price is the engine. Lithium-ion pack prices fell 84% in a decade to $115/kWh in 2024 — the biggest single-year drop since 2017 — and Chinese turnkey systems touched $63/kWh in 2025 before firmer lithium began nudging cell prices up. LFP chemistry has taken roughly 95% of new utility-scale awards on cost, cycle life and thermal stability; sodium-ion is progressing at CATL and peers but remains next-cycle. The practical read for a project developer hasn’t changed: reliability and predictable degradation matter as much as chemistry, and flexible production lines that can run Li-ion today and Na-ion tomorrow are the future-proof configuration.

Exhibit 6: The price fall that created the industry. (Source: BNEF via Ember Global Electricity Review 2025; Benchmark Mineral Intelligence for 2025 system pricing)

KEY TAKEAWAY: Storage in 2026 is where solar was in 2015 — costs down 80%+, volumes compounding at 50%, and the biggest markets just getting started. The difference: storage earns from price spreads that solar itself keeps widening. The two technologies feed each other.

SECTION VII

India’s Storage Build-out: Tenders, Tariffs, VGF

130 GWh of tenders in one year, tariffs down 86% in three, and a Cabinet-approved second VGF round eight times the first — the machinery is built and running.

The policy stack

The tender machine, by the numbers

• Volume: 65 BESS tenders totalling 58.7 GWh plus ~74 GWh of pumped-storage tenders were floated in 2025 — over 130 GWh in one year. VGF-backed projects expanded to 22 projects / 28.5 GWh, of which ~22 GWh already awarded. About 9.2 GWh across 12 projects commissions in 2026.

• Price discovery: two-hour standalone tariffs fell from ₹2.26 lakh/MW/month (2024, VGF T-I) to a record ₹1.48 lakh (APTRANSCO, late 2025); four-hour systems discovered ₹2.85 lakh. Cumulative fall since the first 2022 tender: ~86%.

• The structures are maturing: from plain standalone BESS to FDRE tenders with teeth — SECI’s FDRE-IX (June 2026) packages 1,200 MW with a mandatory 4.8 GWh of four-hour storage and penalties at 1.5x tariff for peak-hour shortfalls. SJVN’s 265 MW/530 MWh Haryana tender (July 24, 2026) extends the VGF template further.

• Who’s winning: Pace Digitek, Oriana Power, Bondada, Stockwell, Ecoren among private names; NTPC Green, Tata Power and Power Grid on the PSU side. Reliance is building toward 120 GWh of annual cell+BESS capacity (first 40 GWh phase commissioning 2026); Pace Digitek is scaling 2.5 → 10 GWh of integration capacity.

Exhibit 7: Discovered 2-hour BESS tariffs — an 86% fall in three years. 2022 point back-derived from the reported cumulative decline, flagged as an estimate. (Source: IESA, IEEFA, MVA Pulse tracker)

Exhibit 8: Pipeline vs requirement — the shape of a 25x build-out. (Source: IESA; MNRE/Cabinet; CEA National Electricity Plan requirement)

The longer arc: industry projections put global annual BESS additions at ~1 TWh by 2035 (from 247 GWh in 2025) — and India’s installed base at 236 GWh by FY32 from just 1.1 GWh in 2025, a 214x expansion, with ~80 GWh a year of combined BESS+EV cell demand between FY27-35. Pumped storage runs alongside: ~74 GWh of PSP tenders were floated in 2025 alone, and PSP remains the only proven 6-hour-plus technology at Indian scale — read BESS and PSP as complements by duration, not competitors.

Exhibit 6A: The storage runways. India’s FY32 requirement is a CEA planning figure; the 2035 global number is an industry projection — both directional. (Source: CEA; Mercom/CES via Waaree deck; IESA)

KEY TAKEAWAY: India has done for storage what it did for solar a decade ago: built an auction machine that manufactures demand visibility. 9 GWh commissioning in 2026 against a 236 GWh requirement by 2032 is the clearest volume-growth runway anywhere in Indian capital goods today.

SECTION VIII

Economics, Risks and the Investment Map

The opportunity is real and so is the sharpest number in this note: tariffs fell 71% while battery packs fell 36%. Somebody is underwriting perfection.

The economics, honestly

At current costs, two-hour BESS paired with solar delivers evening power at rates DISCOMs will sign — that is what the tender volumes prove. The addressable market through 2030 (central plus state procurement) is estimated at 60–80 GWh; at ₹3.5–4 crore per MWh of system cost, that is ₹2.1–3.2 lakh crore of equipment procurement before a single state scheme is added. Revenue models are also maturing: fixed capacity charges (₹ lakh/MW/month, paid on availability) dominate, with FDRE structures layering energy delivery obligations on top.

The caution, from IEEFA’s 2025 tariff study: between 2022 and 2025 tariffs fell over 71% while battery pack prices fell ~36% — auctions have priced in cost declines that haven’t fully happened yet. Roughly three-quarters of the two-hour capacity awarded sits below IEEFA’s viability benchmark; the four-hour segment looks healthier, with over two-thirds of capacity aligned to benchmarks. This is solar’s 2017-18 story replaying: aggressive bids, a shake-out among thin-balance-sheet winners, consolidation toward players who can procure cells well and finance patiently. It is a reason for selectivity, not for sitting out.

What to own, what to watch (snippets, not recommendations)

Case study: what full integration looks like — Waaree’s Q1FY27 numbers

One company’s disclosures, fresh from its July 30, 2026 earnings deck, show what the twin build-out looks like from inside a balance sheet — useful here as a template, not a recommendation. The headline machine: 26 GW of module capacity (the largest non-Chinese maker globally), 5.4 GW of cells (India’s largest), a ₹61,500 crore order book (25.2 GW of modules), Q1FY27 revenue of ₹7,932 crore (+79% YoY) with ₹1,440 crore EBITDA and reaffirmed FY27 EBITDA guidance of ₹7,000-7,700 crore — on net cash (-0.08x net debt/equity) and ~28% ROCE.

Exhibit 8A: One balance sheet across the whole stack — capex announced vs deployed by vertical, as of June 30, 2026. Note the ₹10,000 cr BESS leg — the second-largest bet after solar itself. (Source: Waaree Energies Q1FY27 investor presentation)

Read the programme as a map of where an insider thinks the value sits: ₹12,047 cr into upstream solar (cells 10 GW, ingots/wafers 10 GW, polysilicon stake), ₹10,000 cr into BESS (20 GWh cell+pack+container by FY28 — 5.15 GWh of container capacity already commissioned against 3.5 GWh planned, plus a 1,520 MWh BESS EPC order in hand), ₹3,900 cr into glass behind the anti-dumping wall, ₹4,475 cr into power infrastructure and T&D (including the ₹1,225 cr APSPL transmission-structures acquisition), and comparatively tiny cheques into transformers (₹192 cr for 20,000 MVA), inverters (₹180 cr for 4 GW) and electrolysers (₹676 cr for 1 GW) — capital-light entries into the Section-V stack, riding the retail network (27 states, 600+ franchisees, ~1 in 6 Indian module installations). The strategic read: the largest module maker is spending its module profits to stop being (only) a module maker — which is the whole thesis of this note executed by one management team.

The wider player map, with numbers

Risk register

KEY TAKEAWAY: Buy the build-out, respect the bids. The volume story is as close to certain as infrastructure gets; the margin story depends on tariff vintage, cell procurement and balance sheets. The equipment and enabler layers carry the volume with less of the auction risk.

Conclusion

Solar built the problem; storage is being paid to solve it. India has both machines running at once — that is rare, and it is the opportunity.

India enters FY27 with a solar engine proven at 44+ GW a year and a storage engine that has just been handed its scale-up mandate — an eight-times-larger VGF round, tender structures with real delivery obligations, and a national requirement that implies a 25-fold build-out by 2032. The global backdrop helps rather than hinders: battery costs near all-time lows, no supply constraint, and the world’s biggest markets proving every use case India needs.

The discipline this moment asks of an investor is the same one solar taught a decade ago: separate the certainty of volumes from the uncertainty of margins. Volumes accrue to equipment, integration and enablers almost regardless of who wins each auction. Margins accrue to whoever bid sensibly, procures cells well and is financed to outlast a shake-out. Hold both ideas, and the twin build-out is one of the most investable multi-year themes in Indian capital goods.

Figures marked ~est. are derived or estimated and flagged where they occur; conflicting source counts (module capacity, cell capacity) are shown as ranges rather than resolved silently. All data as of July 27, 2026.

Sources

MNRE / PIB and minister statements (FY26 additions, 162.15 GW Jun-2026, PM Surya Ghar 47.26 lakh households) · JMK Research via pv-magazine India (FY26 segment/state splits, 275 GW RE) · SolarPower Europe Global Market Outlook 2026 via pv-magazine (664 GW, 2026 plateau) · BloombergNEF Energy Storage Outlook H1-2026 via Energy-Storage.news (112 GW/307 GWh, 2026F) · InfoLink via Balkan Green Energy News (geographic splits, 421 GWh shipments) · Benchmark Mineral Intelligence via ESS News ($63/kWh, 315 GWh, lithium firming) · Ember Global Electricity Review 2025 ($115/kWh pack) · IESA via Saur Energy (130+ GWh tenders, 9.2 GWh 2026, VGF 28.5 GWh) · IEEFA May-2026 tariff viability study (₹1.48L/₹2.85L, 71% vs 36%) · MVA Pulse BESS tracker (winners, 86% fall, VGF caps) · Cabinet May-14-2026 via Anantam (₹5,400 cr / 30 GWh / PSDF) · SECI FDRE-IX and SJVN tender documents via trade press · CERC Petition 31/AT/2026 (NVVN ₹6.64/kWh comparison) · Construction World / Taiyang News (manufacturing capacity range, trade data) · Waaree Energies Q1FY27 investor presentation, Jul-30-2026 (company figures; demand-outlook and adjacent-market projections compiled therein from MNRE/SEIA/Research & Markets/Precedence/Mordor/Mercom — treated as directional and labelled).

• What happened: India’s third attempt at chips, launched Dec 2021 with ₹76,000 cr, approved 12 plants worth ₹1.64 lakh cr across 6 states. Three are producing (Micron, Kaynes, CG Semi — all packaging, all 2026). The one fab (Tata-PSMC, ₹91,000 cr, 28-110nm) ships wafers in 2028. Display fabs are the one piece that has not taken off.

• The pattern: success tracked technology-transfer difficulty, not administration. Packaging (easy to transfer) beat plan; fabs (hard) are late; displays, locked up by a two-country duopoly, found no partner.

• The money mechanics: the Centre pays ~50% of construction capex pari-passu, states add 15-30%, so companies own plants with ~20-30% of their cost. Micron: $2.75 bn plant, Micron’s share $825 mn.

• Committed ≠ disbursed: ~₹3.1 lakh cr announced across the policy stack; ~₹25k cr cash paid. ISM itself: ~₹10-11k cr through FY26, almost all to packaging; the fab line disbursed near zero; the design incentive (DLI) spent 0%, 15%, 34%, 52% of its small budgets over four years.

• What 2.0 changes (Cabinet, Jul 15, 2026): ₹1,27,500 cr, six pillars. 1.0 bought factories (fab + packaging); 2.0 buys the ecosystem — equipment, materials, chemicals, gases, ingots, design IP, R&D, technicians. The right sequence, and the harder half.

• The value chain in one line (§2): India today operates only in the chain’s thinnest-margin links — OSAT is an ~$8 bn global profit pool; the links 2.0 targets hold ~$150 bn between them.

• Adjacent money: ECMS ₹40,000 cr for components (75 projects, ₹61,671 cr investment: PCBs, camera and display modules) and Mobile Scheme 2.0 ₹62,500 cr for devices — three layers, three schemes, routinely confused.

• Where India stands globally (§11): largest exporter of design labour (~20% of world design workforce), zero sovereign wafers until 2028, packaging scaling fast. Spending ≈ 0.62% of GDP — 3x the US relative effort, behind only Korea.

• Why it matters (§23): electronics is India’s #2 import (₹8.3 lakh cr FY26, half an oil bill); chips are its irreducible core, sourced through Taiwan and China. The jobs case is modest (~₹3.2 cr of support per direct job) — the real case is capability and security, and at 0.16% of the budget a year it is inexpensive insurance.

• The 2030 honest case (§20): one fab at yield, top-5 packaging destination, forming materials base, import share bending down — and India still importing ~70-85% of its chips. A genuine achievement, with plenty of road still ahead.

• What to watch next (§22): 2.0 guidelines (due within weeks of Jul 15), TSAT Assam production, ECMS first payouts FY27, Dholera tool-in 2027.

• The clearest opportunity: design is India’s strongest suit, and 1.0 gave it the smallest cheque with rules that pushed successful startups out. 2.0 makes IP ownership a full pillar — whether the new guidelines drop the foreign-capital bar and settle IP rights is the best early read on 2.0’s quality.

Contents

1 | Background: what ISM is and when it started

Start date: December 15, 2021. The Union Cabinet approved the Semicon India Programme with an outlay of ₹76,000 crore. The India Semiconductor Mission (ISM) was set up inside Digital India Corporation as the nodal agency to run it. The trigger was the 2020-21 global chip shortage, which stopped Indian car production lines and exposed the fact that India imported essentially 100% of its chips while consuming (then) roughly $25-30 billion of semiconductors a year, headed toward $100 billion+ by 2030.

The stated goal: build a domestic semiconductor and display manufacturing ecosystem — fabrication, packaging, design, and displays — so that India’s electronics industry (now the country’s fastest-growing export category) stops being an assembly business built entirely on imported silicon.

Context that matters: India had tried twice before. SCL Mohali (1983) was a genuine fab that fell decades behind after a 1989 fire and underinvestment. Fab proposals in 2007 and 2013-14 (two consortia were approved letters of intent) died without construction. ISM 1.0 is the third attempt, and the first with money that actually moved.

2 | The value chain: sizes, profits, and where India stands

A chip passes through nine industries before it reaches a device. No country controls all nine. Every national programme is a choice of which links to buy, build, or rent. The poster below is the whole framework on one page — read it top to bottom, then the numbers table.

2.1 The numbers behind the poster (2024, global)

Sources: SEMI (materials $67.5 bn, 2024), industry equipment data ($109 bn, 2024), WSTS-class total (~$600-627 bn, 2024). Margins are industry-typical; profit pools = revenue × margin, rounded. India cells marked ~est. where no official series exists.

2.2 Profit pools in absolute dollars

The point in one sentence: India’s entire current position (OSAT) competes for an ~$8 bn global profit pool, while the links 2.0 targets — equipment, materials, EDA/IP, design — hold roughly $150 bn between them, and foundry another ~$75 bn. The concrete version: of a packaged chip leaving Sanand, India captures roughly the packaging conversion fee — a single-digit share of the chip’s value — because the die, substrate, leadframe, mold compound and every tool are imported. Localising substrates and leadframes alone would roughly double capture per chip; localising consumable chemicals and gases converts one-time capex subsidies into revenue a fab pays every month, forever.

2.3 What 1.0 funded vs what 2.0 funds

Reading the strip: 1.0 went in at the middle of the chain — fab, packaging, design services — and packaging is where it delivered. 2.0 turns nearly every square dark: outward into equipment, materials, wafers and IP; upward into advanced packaging; downward into technician training. The order makes sense. Suppliers don’t set up shop before there’s a customer, and 1.0 built the customers. The next stretch is slower by nature — hundreds of firms qualifying products to fab-grade specs, the way Japan and Taiwan did over a couple of decades. The money is in place; what it buys now depends on patience and steady procurement.

3 | ISM 1.0: design and incentive terms

How the money actually works: ISM 1.0 pays for construction, not production. The government reimburses about half of verified capital spending as it happens — plant, machinery, cleanrooms — with the state adding its share on the same base. On a ₹100 project, the company puts in about ₹30 and owns the whole plant. Two things follow: money moves at the speed of concrete and tools, and the deal is attractive enough to have brought in Micron and Tata. It did its job.

4 | Envisioned vs actual: the 1.0 scorecard

Why this pattern and not another: packaging plants cost ₹3-25k crore, take 12-24 months, and use mature, transferable technology — so they got built. Fabs cost ₹90k crore, take 4-5 years, and need an ecosystem that does not exist yet — so the fab is still in the ground. Displays need technology from a Chinese-Korean duopoly that has no reason to share it — so displays died. The scorecard is not a judgment on execution quality; it is a map of technology transfer difficulty.

5 | The 12 projects

The Aug 2025 tranche (SiCSem + 3D Glass + CDIL + ASIP) totalled ₹4,594 cr officially, with 2,034 direct skilled jobs committed. Figures marked ~ are announced or derived, not audited.

6 | The money: who paid what, on what conditions

5.1 The template, using Micron as the worked example

5.2 All projects: the funding split

What the state race means: Gujarat set the template at 20% and others raised it — UP to about 25%, Odisha and Andhra to 30%. So the newest projects in smaller states get the best terms, close to 80% total support. That is competitive federalism doing what it should: spreading the industry beyond one coast. Worth tracking quietly alongside it: whether state budgets deliver those top-ups on schedule.

7 | Committed vs disbursed: the budget reality

Cross-check (MeitY data, Jun 2026): the five companies investing by FY26 spent ₹15,799 cr in total on their projects, of which ₹5,692 cr was their own money in FY25-26; the balance came pari-passu from the Centre and states. Another ₹15,500 cr of company money is expected in FY27 (₹31,299 cr cumulative). This co-flow figure runs slightly ahead of union budget actuals — the difference is state money and tranche timing. The defensible statement is a band, not a point: ₹10-11 thousand crore of central cash through FY26.

Why the design incentive under-delivered — and why it’s fixable: three design choices held it back. Startups that raised meaningful foreign capital lost eligibility, which filtered out the successful ones. The government’s rights over company IP were left vague, which made founders cautious. And the nodal agency was itself a design player. Total exposure stayed under ₹1,100 crore — small money pointed at India’s strongest capability. All three issues are choices, not laws of nature, and 2.0’s IP pillar exists to change them. The new guidelines are the test.

8 | State incentives compared

9 | Semicon 2.0: the full design

10 | Adjacent schemes: chips vs components vs devices

Three schemes, three layers of the same stack — and the most common confusion in Indian semicon commentary. Chips are ISM. Components (PCBs, camera modules, display modules) are ECMS. Devices (phones) are PLI, now Mobile 2.0.

10.1 ECMS: the component money

10.2 Devices: PLI → Mobile 2.0

Two payment logics, both working: PLI pays when goods are produced — it has paid out about a third of its outlay and phones became India’s top export. ISM pays while plants are built — about 14% paid, with the results now coming out of the ground, three plants at a time. ECMS borrows PLI’s logic for components. The 2028 wafers are the moment the construction-side bet starts producing PLI-style proof.

11 | Global comparison: sovereign capabilities

11.1 The money, with mechanisms

11.2 Capability scorecard

● strong ◐ partial ✗ absent

Where India stands: India comes into 2.0 with the world’s largest pool of chip-design talent — about a fifth of the global design workforce, over a lakh engineers, some two thousand chips a year designed here — and its first sovereign wafers arriving in 2028. The nearest comparison is the UK turned inside out: the UK holds design crown jewels with no fab plans; India has the scale, the fabs under way and the state appetite, and is now going after the IP. Against Korea, Taiwan and Japan the gap is accumulated experience, not money — and pillars 2, 5 and 6 of 2.0 are a serious first attempt at closing it. Nobody has compressed that journey below about fifteen years. India has started the clock.

12 | Talent, employment, market size

13 | Ancillaries: built and buildable

13.1 What exists today because of 1.0

• Gas and chemical contracts around the anchor sites — industrial-gas majors building on-site plants at Dholera and Sanand; Archean Chemical appears in ISM’s own ecosystem announcements.

• Equipment multinationals building Indian bases — Applied Materials’ ~$400 mn Bengaluru engineering centre; Lam Research’s ~₹10,000 cr Karnataka plan plus a pledge to train 60,000 engineers; ASM International in Chennai; Tokyo Electron training tie-ups with Tata.

• Design centre density — Intel, Qualcomm, Nvidia, AMD, MediaTek, Broadcom, TI all run large India design centres; design-GCC count projected to grow ~30% by 2030.

• ECMS-funded component base — PCBs, camera and display modules, passives, laminates, magnets: the board-level layer beneath chips.

• Still ~100% imported: substrates, leadframes, mold compounds, photoresists, wafers, all fab tools.

13.2 The 2.0 whitespace, ranked by how realistic it is

14 | Interpretation: ten judgments

• 1. The mission delivered where technology travels easily, and is still working where it doesn’t. Packaging beat its plan. The fab takes longer because fabs everywhere take longer. Displays found no willing technology partner. Read the scorecard as a map of how hard each technology is to move, not as a report card on the administrators.

• 2. Headline outlays and actual spending are different things, on purpose. Around ₹3 lakh crore stands announced across the policy stack; roughly ₹25 thousand crore has been paid out. Money is released as plants get built and goods get made, so the yearly fiscal cost stays small and the programme survives budget cycles. The heavy spending years are FY27-FY30, and that is when commitment gets tested.

• 3. Companies own the plants while putting up about a quarter of the cost. That is a generous deal, and it worked — it brought Micron and Tata in. It also means announcements come cheap. So count production, not press releases.

• 4. Gujarat earned its lead. It moved first with a clean 20% policy, ready land and utilities, and it shows in the results: 6 of 12 projects. The flip side is concentration — a lot now rides on one state’s power, water and politics, and the mission gets stronger every time a plant succeeds outside it. Assam in 2026 is the one to watch.

• 5. States are competing to host this industry, and that is new. Gujarat offered 20%, UP 25%, Odisha and Andhra 30%. Later projects in smaller states get the best terms — a deliberate way of spreading the industry beyond one coast. The thing to keep an eye on is whether those state budgets can carry the promises they’ve made.

• 6. The clearest fix available is in design. Design is India’s strongest suit — a fifth of the world’s chip designers — and 1.0’s design incentive was its smallest cheque, with rules that pushed out startups the moment they raised serious money. 2.0 makes IP ownership a full pillar. If the new guidelines drop the foreign-capital bar and settle the IP-rights question, this becomes the highest-return rupee in the whole programme.

• 7. 2028 is the year everything prices off. When Dholera ships its first commercial wafers, every future fab application, every equipment maker’s India decision, and 2.0’s standing all re-rate at once. Until then, patience — that is how fab construction works everywhere.

• 8. Read sanctions, not headlines. The one public rupee sanction (HCL-Foxconn, ~₹1,500 cr on a ₹3,706 cr project) came in near 40%, not 50% — eligible capex is narrower than project cost. Apply that discount mentally to every derived figure, including ours.

• 9. The materials pillar is where value capture multiplies. Nine packaging plants importing every substrate and leadframe capture a small slice per chip today. Each material that localises raises that slice, and a fab buys materials every month, forever. This is the quiet compounding story inside 2.0.

• 10. India’s realistic 2030: a genuine achievement, honestly described. A working mature-node fab, a top-5 packaging destination, a forming materials base, the world’s largest design talent pool starting to own IP — and continued imports for equipment, EDA and advanced chips. Every country that made it started exactly here. The Substack can say both things with a straight face.

15 | Data gaps and sources

What is genuinely not public

Sources

Primary: ism.gov.in (Semicon 2.0 approval, six pillars, timeline, SCL ₹4,500 cr, project announcements) · PIB (Micron Cabinet notes Jun/Dec 2023; Aug 12, 2025 four-project tranche ₹4,594 cr; ECMS tranches Oct 27, 2025 / Jan 2, 2026 / Feb 2026 cumulative / Mar 30, 2026) · Union Budget documents FY23-FY27 via the Takshashila/Kotasthane semiconductor tracker (BE/RE/actuals by sub-scheme; DLI utilisation 0/15/34/52%; fab-line zero disbursal; FY27 ~₹8,000 cr) · Invest UP (Semiconductor Policy 2024 terms) · ecms.meity.gov.in.
Secondary, dated in text: Business Standard (EFC ₹1.25 lakh cr, Jun 30, 2026; five-company investment ₹15,799 cr, Jun 11, 2026; Cabinet approval, Jul 15, 2026) · BusinessToday and Electrical Mirror (six-pillar detail, Jul 15, 2026; ECMS ₹40,000 cr and beneficiaries, Feb 1, 2026) · Deccan Herald (HCL-Foxconn ₹1,500 cr; Karnataka 20% offer) · Outlook Business/PIB (Micron $825 mn / 50% / 20%) · Expertlancing and IMARC (AP 60%, TN 50%, Odisha 30% state stacks) · IBEF (DLI 24 projects, 105 companies, 23 tapeouts; PLI FY25 disbursements ₹10,114 cr total, ₹5,732 cr electronics) · Lexology / India Briefing (ECMS mechanics) · CRS R47558, CSIS, globaldeal.io (US/EU/JP/KR/TW/CN programme mechanics) · PwC “Semiconductor and Beyond” (workforce 2030) · Company disclosures (CG Power, Kaynes, Micron) as referenced.

Prepared July 27, 2026. Figures marked ~est. are policy-arithmetic derivations, not sanctioned amounts. Conflicting official figures are flagged where they occur rather than averaged.

Part II — The expansion

Demand economics · the full failure history · country lessons · listed-equity map · 2030 scenarios · risk register · 12-month watch calendar

16 | The demand side: what India buys, and from whom

The demand-supply match is a real advantage. India’s chip demand is concentrated in mature-node parts — power chips, microcontrollers, display drivers — which is exactly what a 28-110nm fab and nine packaging plants make. The country is building capacity for demand it already has and currently imports. One thing to explain to readers early: demand is growing so fast that the import bill keeps rising even as domestic supply comes online. The right scorecard is import share, which falls — not the bill, which won’t for a while.

17 | The road to the one that worked: three earlier attempts

Attempt 1 (1983-): SCL Mohali

India built a genuine state-owned fab in 1983 — at the time only a few years behind the global frontier. A 1989 fire destroyed much of the facility; rebuilding took years, investment never kept pace, and SCL settled into a small 180nm-class R&D and space-grade role. The lesson institutionalised for decades: a fab is not a one-time purchase but a permanent spending commitment, and India stopped making it. The ₹4,500 cr modernisation announced in Nov 2025 (with a no-privatisation assurance) is the state finally re-engaging with its own asset — though budget lines for SCL have gone essentially unspent for years.

Attempt 2 (2007-2014): the letter-of-intent era

The 2007 SemIndia proposal and the 2013-14 approvals to two consortia (one anchored by Jaypee with IBM/Tower technology, one by HSMC with STMicroelectronics) all died without construction. Common causes: no anchor customer, financing that never closed, technology partners with licences but no equity, and incentive frameworks that expired before decisions were made.

Attempt 3 (2022-2023): the first ISM fab round — three applicants, zero fabs

Why the second round worked: every 2022 fab applicant was a financier hunting for a technology partner. The approval that stuck — Tata with PSMC — flipped that: an industrial house with a real balance sheet, and a foundry putting its own process technology and engineers into the project. The government’s insistence on a production-grade technology licence looked harsh when Vedanta-Foxconn unwound; it turned out to be the mission’s best call, because it meant only buildable projects got approved. Twelve approvals and three plants producing is the payoff of that discipline.

18 | Country lessons

19 | The listed-equity map

A tracking note: every listed name here carries some of the semicon story in its price already. The edge is in the boring places: grant receipts and receivables in annual-report notes, capacity utilisation disclosures, and the gap between announced and commissioned capacity. Commissioned capacity is the number that settles arguments.

20 | 2030 scenarios

All three scenarios share one property: India remains a net chip importer in 2030 by a wide margin. The mission’s honest 2030 success metric is not self-sufficiency; it is (a) one fab at commercial yield, (b) packaging at global top-5 scale, (c) a forming materials base, and (d) import share bending downward. Anything claiming more is marketing; anything dismissing that as failure ignores what every other late entrant’s first decade looked like.

21 | Risk register

22 | The 12-month watch calendar (Aug 2026 – Jul 2027)

23 | How big is this bet, really? Relative comparisons and criticality

23.1 Against India’s own programmes

23.2 Against the import bill: the strategic exposure

Electronics is India’s second-largest import — half an oil bill — and chips are its irreducible core: the phone, the PCB and the module can all be localised (and are being, via PLI and ECMS), but every one of them is built around imported silicon. Oil dependence at least has many suppliers; chip dependence runs through one strait — Taiwan fabricates over 60% of the world’s chips and ~90% of the advanced ones, and China supplies roughly a third of India’s IC imports. A Taiwan contingency is simultaneously an energy-crisis-scale economic event for India’s electronics industry and a defence-supply event, with no strategic petroleum-reserve equivalent possible: chips cannot be stockpiled meaningfully because designs churn.

23.3 Against other nations’ effort, relative to size

In absolute dollars India’s ~$24 bn is mid-pack (§11). Scaled to GDP it reads differently: roughly three times the US CHIPS Act’s relative weight, in the same class as Japan and China, behind only Korea. Two honest caveats cut opposite ways: India’s figure is a decade-length commitment against one year’s GDP, flattering the ratio — but every country’s figure is computed the same way, and India’s number also excludes state top-ups (add ~₹40-50k cr ~est. of committed state money and the ratio rises further). Relative to income, India is spending as much as anyone outside Korea. The true constraint, as §2 and §8 argue, is absorption — ecosystems, suppliers, technicians — not money.

23.4 Criticality scorecard (author assessment)

Putting it together: by money, this is India’s largest sector bet and still a small slice of the budget. By strategic exposure, it goes straight at the most concentrated vulnerability in the import basket. By jobs, it is modest — the case rests on capability and security, and it is a strong one. A fair one-line description for the Substack: affordable insurance against an expensive dependency, with a real industrial upside attached.

24 | The source library: what to read, and what each source is for

Reading order for someone starting cold: Chip War → §2 of this document → the Takshashila tracker → then live sources. For quarterly maintenance of this document: PIB + sansad.in annexures + company filings cover 90% of what changes.

Two Averages Deck Final

195KB ∙ PDF file

Download

Download

Let me start with a number you already know. India’s per capita income is $2,697 a year. You’ve seen it in every budget speech, every “fifth largest economy” celebration, every panel discussion where someone in a blazer says the word “trajectory.”

What I want to convince you of over the next twenty minutes is that this number, while technically correct, tells you almost nothing about the country you actually live in. And once you see why, a lot of things that feel confusing right now, including why lakhs of students are sitting at Jantar Mantar over a leaked exam paper, start making uncomfortable sense.

That’s all I’ll say about the protest, by the way. The leak matters, but it’s the spark, not the fuel. The fuel is what’s below.

Part 1: The two Indias, measured

So here’s a small exercise. The top 10% of Indian earners take home 58% of all the income in the country. That’s from the World Inequality Report, and we’ll get to whether you should trust it later, because I know some of you are already composing the WhatsApp reply. For now just run the arithmetic with me: take those roughly ten crore people out of the picture, and divide what’s left among the remaining 90%.

You get about $1,270 per head.

For context, the average for Sub-Saharan Africa is around $1,700. So the India that 125 crore people wake up in every morning is, on income, poorer than the average African economy. At the same time, the average member of our top 1% earns around $64,000 a year, which is three times the average Pole. Poland is in the European Union.

Neither of these groups lives in the $2,697 country. Nobody does. It’s just the midpoint between two populations that don’t shop in the same places, don’t study in the same schools, and, as we’ll see, don’t even queue for the same things.

The rupee version, from the World Inequality Lab’s 2022-23 estimates, since it makes things concrete:

I did the maths on that last row a few times because I didn’t believe it. The gap between the average bottom-half Indian and the average member of that top sliver works out to about 6,800 times. If you earn ₹71,000 a year, you’d need to work for longer than recorded human civilisation to make what they make in one year. At some point a gap stops being a gap and becomes two different universes that happen to share a currency.

And income is actually the gentler measure. Wealth is where it gets properly lopsided. The top 1% owns about 40% of everything in this country. The bottom half, which is 70 crore people, owns 6.4%. Hurun counted 284 dollar-billionaires in India last year, together worth about ₹98 lakh crore, which is a third of our GDP and bigger than Saudi Arabia’s entire economy. Widen the lens a little and about 1,683 Indians are worth over ₹1,000 crore each; their combined pile is ₹167 lakh crore, which is getting close to half of GDP, held by a group of people who would fit comfortably into one large wedding hall.

Now, you could say fine, the rich got richer, that’s capitalism, what about everyone else? Here’s what happened to everyone else over the same six years. That ₹1,000-crore club grew its wealth from ₹31 lakh crore to ₹167 lakh crore, up 440%. In the same period, ordinary household debt doubled, from ₹70 lakh crore to ₹137 lakh crore, and it now sits at 47.8% of GDP, the highest the RBI has ever recorded. Household savings meanwhile fell to about 5% of GDP, the lowest in decades. So the top multiplied its wealth five times over while the rest of the country borrowed more and saved less to stay in place. When I put those three lines next to each other I stopped thinking of it as two lanes moving at different speeds. It’s more like one lane is financing the other.

The bit about the British that I can’t get over

There’s one comparison in all this research that genuinely stopped me, and I want to spend a minute on it.

Because the British taxed us, India has income tax records going back to 1922, which means economists can trace exactly how much of the country’s income the top 1% has taken, every year, for a century. In 1940, at the height of colonial extraction, that number was 20.7%. After Independence it fell and fell, all the way down to 6.1% by 1982. And then it turned around. By 2022-23 it stood at 22.6%.

Read that again. On the same measure, using the same kind of tax data the colonial administration used, the top 1% of Indians today takes a larger share of India’s income than it did under the British. The economists who assembled the series didn’t bury this in a footnote; they put it in the title of their paper. They called the era the Billionaire Raj.

Now, to be fair, and this matters: 1991 worked. India is roughly six times richer per person in real terms than it was in 1990, and I’m not remotely nostalgic for the licence raj. But it’s worth being honest about what else happened. Everything Independence achieved on distribution between 1947 and 1982 has been unwound completely, and then some.

The awkward China comparison

Whenever this comes up, someone says every fast-growing poor country goes through this phase, it’s the price of growth, look at China. Which is convenient, because China is actually the perfect test of that claim, and it fails.

In 1990, China was poorer than us. $318 per head against our $368. I’ll just leave the table here:

They’re now four and a half times richer than us per person, and we’re roughly where they were in 2007. That part everyone knows. The part that kills the “price of growth” argument is the distribution. Both countries opened up, and yes, inequality rose in both. But China’s top-10% share climbed from 27% to about 43% and then it stopped, around 2010, and has drifted down since. Ours went from 32% to 58% and has never stopped climbing.

Why did theirs stop and ours didn’t? Not because their billionaires are more generous. It’s the design of the growth. China grew through factories, and factories pull crores of farm workers into wage jobs whether anyone intends it or not; they’d also put nearly every kid through school before the boom started, and after 2005 they actively pushed the gap down with rural tax cuts, social insurance and rising minimum wages. We grew through services, and services are wonderful if you’re in the English-speaking top decile (fluent English carries about a 34% wage premium in India, per published research) and mostly irrelevant if you’re one of the 90% of workers doing informal work with no contract, no pension, no anything.

So inequality wasn’t the price we paid for our growth model. It IS our growth model. Which brings us to why the exams matter so much.

The queue that doesn’t move

India’s median age is 29. China’s is 39, America’s is 40, Japan’s is 48. Roughly 90 lakh to a crore young people join our job market every single year, and everyone calls this the demographic dividend, and on paper it should be.

Except look at what they’re joining. Youth unemployment is around 16%. Among graduates it’s about 29%, which is its own kind of dark comedy: in India, the more educated you are, the more likely you are to be unemployed. Fewer than one in ten workers has any social insurance. And the doors out of all this? NEET admits 4%. IIT admits 2%. The army’s Agnipath scheme got around 50 lakh applications for 46,000 posts, which is under 1%. State police recruitments routinely pull 40 lakh to a crore applications.

But the finding that really settled it for me came out in 2024, in one of the top economics journals. Three researchers, Asher, Novosad and Rafkin, measured what happens to a child born into the bottom half of India’s distribution, tracking birth cohorts from the 1950s onwards. Their conclusion, and I’m quoting directly: mobility in India has been “constant and low since before liberalization.”

Constant. Thirty years of 7% growth, the IT boom, the startup wave, all of it, and a bottom-half kid’s odds of climbing did not improve at all. The prizes got much bigger. The odds never moved. (Two details worth knowing from the same paper: mobility for Scheduled Castes actually rose, and the authors credit reservations, so policy evidently can move this number when it tries; and Muslims’ mobility fell by about the same amount, while girls do worse than boys everywhere.)

So when those students sit at Jantar Mantar defending the sanctity of one exam, understand what they’re defending. It’s a queue that measurably hasn’t moved in seventy years, and it’s the only queue they’ve got.

Part 2: What the young are actually doing about it

Alright. If you’re 22, reasonably smart, and you can see all of the above even if you’ve never read a single paper about it, what do you do?

First you’d check whether plain old work pays. And here I don’t need to cite any activist; the Finance Ministry’s own Economic Survey did the job.

Corporate profits for the Nifty-500 hit 4.8% of GDP in FY24, the highest in fifteen years, up from 2.1% two decades ago. In the same window, rural real wages, that is, what a rural worker’s pay actually buys, were lower in 2023-24 than in 2017-18. Lower. Profits at those big listed firms rose 22% in one year; employment at the same firms rose 1.5%. Even the Chief Economic Adviser said out loud that companies hoarding profits instead of paying wages are eating their own future demand.

So wages don’t pay. What does a rational 22-year-old do when the wage escalator is broken? He buys a ticket. And once you see it, you see it everywhere, because young India is buying five different kinds of tickets, all with roughly the same odds.

1. Exam Ticket

2. Stock Market

3. Phone Camera

4. Gaming apps

5. Leave India

The first is the exam itself, and there’s an entire industry selling that ticket: coaching is now a ₹58,088 crore business heading towards ₹1.34 lakh crore by 2028. Twenty-seven percent of all Indian students take private coaching, and it eats 10 to 11 percent of the family budget, for a 2 to 4 percent chance of a seat. Kota takes in two lakh students a year and about 3% make it. One more twist: urban families spend four times what rural families do on this ticket, so even the lottery is rigged towards those who need it least.

The second ticket is the stock market, and this one we know with unusual precision because SEBI went and counted every single trader. Ninety-one to ninety-three percent of individual F&O traders lose money, every year they’ve measured it. They lost ₹1.8 lakh crore between FY22 and FY24 and another ₹1.05 lakh crore in FY25 alone. And who are these people? Three quarters earn under ₹5 lakh a year. Nearly half are under 30. Seventy-two percent are from outside the big cities. On the other side of their trades sit algorithms, which made 97% of the foreign institutional profits in this segment. Over four years, roughly ₹2.9 lakh crore has flowed from small-town young men to algo desks and the taxman, which for scale is about three full years of MGNREGA. And here’s the detail from SEBI’s study that I find most telling, and least funny: three quarters of the people losing money keep trading even after years of losses. That’s not stupidity. That’s people who have looked honestly at their job prospects and concluded that a 7% lottery beats them.

The third ticket is the phone camera, and this is the one people romanticise. India has something like eight crore people trying to make it as creators. Of those, maybe 8 to 10 percent earn anything at all from it, against about 40% in mature markets. YouTube’s own commissioned study found its entire Indian ecosystem supports around 9.3 lakh full-time-equivalent livelihoods, which against eight crore aspirants is a conversion rate of about 1%. The same as Kota, if you’re keeping score. The median creator makes ₹60,000 to ₹80,000 a year, which, and let this sink in, is less than the average income of the bottom half they’re trying to escape. There’s even an exchange-rate insult built in: an Indian view pays roughly a tenth of an American view, because ads price against Indian wallets while the platform’s cut is global.

The fourth ticket was the gaming app. Real-money gaming, fantasy sports, rummy, opinion trading, grew into a ₹30,000-plus crore industry with some 45 crore users, until Parliament banned the whole thing in August 2025, citing addiction and destroyed families. I keep coming back to the irony here: the only lottery India has actually outlawed is the one that admitted it was a lottery. The exam, the options screen and the feed, with the same or worse odds and bigger total losses, remain legal, celebrated, and taxed.

And the fifth ticket is simply leaving. This one is being bought at both ends of the pyramid at once. At the top, about 4,300 dollar-millionaires left India in 2024 alone. At the bottom, Gulf migration hit records and remittances reached $129 billion, the highest of any country on earth and roughly six times our FDI inflows. In between are the families selling land to pay ₹40-60 lakh for the donkey route to America at worse-than-coin-flip odds. Whatever we tell ourselves about the India story, our most reliable export is people who examined the domestic lottery and walked away from the counter.

The house that always and always wins

Now notice something about all five games. The players lose in every one of them. But there’s a layer in the middle that wins in every one of them, and honestly this layer is where the inequality machine actually lives.

The coaching companies, obviously, growing 18% a year on the fuel of pure desperation. The brokers and exchanges and, let’s be honest, the government, which collects STT and GST on every one of those losing trades. The platforms, which monetise the free content of eight crore hopefuls and pay out to one percent of them. And then there’s the one I’d read twice if I were you: the lenders. RBI’s latest stability report shows that people earning under ₹10 lakh now account for about 75% of new personal loans and 78% of the fresh bad loans. Microfinance has shrunk for seven straight quarters, which tells you the bottom of the pyramid has run out of unsecured credit entirely. So what’s the fastest growing loan product in India right now? Gold loans. Up more than 100% in a year, now the single biggest retail lending category outside housing. Families are pledging the wedding gold, the last thing any Indian household touches, partly to get by and partly, I suspect, to keep buying tickets.

One uncomfortable footnote for readers of this newsletter specifically: the house is listed. Coaching, brokers, exchanges, gold-loan NBFCs. As a basket it has comfortably beaten the Nifty. The cleanest publicly traded bet on Indian inequality was never a luxury stock. It’s the ticket sellers.

Part 3: The gap has stopped being a statistic

Everything so far is money, and money can in principle be redistributed. What convinced me this is now a different kind of problem is how much of the gap has stopped being financial and become physical. It’s in bodies, on the map, inside the tax code and the prisons. Let me just walk through it.

Start with children, because this is the least abstract version. 35.5% of Indian kids under five are stunted, and 57% of women are anaemic; that’s the government’s own NFHS survey. Stunting is inequality measured in centimetres, decided before a child’s fifth birthday, and it mostly doesn’t reverse. Around 40% of Indians, about 59 crore people, cannot afford a healthy diet by FAO’s measure, and yes, that’s actually a big improvement from a few years ago, but it’s still six Frances’ worth of people priced out of decent food in a country sitting on record grain stocks. And if you want the state’s own confession of scale, it’s this: free foodgrain for 81.35 crore people, extended till December 2028, at a cost of ₹11.8 lakh crore. Think about what that ration list actually is. The government of the world’s fifth-largest economy has formally certified that 58% of its citizens cannot be relied on to feed themselves. Meanwhile, in the very same quarter, the fastest-growing words in FMCG earnings calls are “premium” and “gourmet.” Same country, same week.

Then look at the map. Telangana and Karnataka are at roughly ₹3.1-3.5 lakh per person; Bihar is at about ₹60-66 thousand. That’s a five-times-plus gap between states of the same union, and it has roughly doubled since 1991. Fertility has split along the same line, around 1.4-1.6 children per woman in the south, which is European territory, versus 2.6-3 in the Hindi belt. So we now have two demographic countries as well: one getting old before it gets rich, one young without jobs. And this collides head-on with delimitation after 2026, when parliamentary seats get redrawn towards the poorer, younger north while the richer, older south keeps writing the tax cheques. The two-Indias problem is about to acquire a border and a seat count.

Then the tax system, which quietly runs the same transfer as the labour market. Personal income tax used to be 38% of direct taxes; it’s now 53%, and corporate tax has gone the other way. In 2025, for the first time in our history, salaried individuals paid more direct tax than all of corporate India combined. Sit with that alongside everything in Part 2: record corporate profits, taxed relatively less; falling real wages, taxed relatively more, and then taxed again at the shop through GST, which by its nature takes a bigger bite of a poor family’s budget than a rich one’s.

The jails, too. 73% of India’s prisoners have never been convicted of anything; they’re undertrials in jails running at 113% capacity, and overwhelmingly they’re just poor, people who couldn’t post bail or pay lawyers, some of whom serve longer waiting for trial than their alleged crime even carries. Dalits, Adivasis and Muslims are all overrepresented, every single year. In practice, bail in this country is a product, and the same money that buys coaching seats buys freedom.

And my favourite quiet statistic in all of this: India employs somewhere between 90 lakh and a crore private security guards, several times our 21 lakh actual police. It’s one of the largest private armies in the world, made up of men earning ₹12-15,000 a month standing twelve-hour shifts outside property worth crores. If you want one image for this whole essay, that’s it: one India paying the other India minimum wage to guard it from itself. And it doesn’t stop at guards. When the water fails, there are tankers and RO. When the air fails, purifier sales spike every November while delivery riders breathe whatever the AQI says. Private schools when government schools fail, private hospitals when public ones do, and out-of-pocket health costs are still over 40% of all health spending, which remains one of the surest ways an Indian family falls back into poverty. The pattern, once you notice it: the top 10% hasn’t escaped India’s failures. It has bought private exits from every one of them. And every private exit removes exactly the people with the clout to demand the public thing gets fixed.

A pause, because I can hear the objections

There is a genuine case on the other side and it deserves better than a strawman, so let me make it properly and answer it.

The strongest point first: extreme poverty really has collapsed. 171 million Indians moved out of it between 2011 and 2023 by the World Bank’s count, and the extreme poverty rate is now down to somewhere between 2 and 5 percent depending on which line you use. This is real, it’s enormous, and nothing in this essay contradicts it, because poverty and inequality are different questions. One is about the floor, the other about the ladder. India raised the floor, genuinely. The ladder got longer faster than anyone could climb it.

Then there’s the headline you probably saw in July 2025: the World Bank supposedly ranking India the fourth most equal country on earth, Gini of 25.5, ahead of everyone except Slovakia, Slovenia and Belarus. I’d encourage anyone waving this around to actually open the database. That 25.5 is a consumption Gini, and consumption always looks equal at the top, because rich people save most of their income and barely respond to household surveys anyway. The same World Bank database carries an income Gini for India, and that one rose from about 52 to about 62 between 2004 and 2023. The 25.5 is also a Bank-adjusted figure, 29.5 before adjustments, computed on a redesigned survey that the Bank itself warns isn’t comparable to the old one, and then ranked against other countries’ income-based numbers. A country cannot be the fourth most equal society on one page of a database and more unequal than the British Raj on another page, unless the two pages are measuring different things. They are. This essay is about the income and wealth page.

Third objection: welfare closes the gap. Partly true, and the poverty numbers above are partly welfare’s achievement. But notice what this defence concedes. The argument for India’s income distribution has become “yes, but 58% of people get free food.” That’s a floor, not a ladder, and whether the exchequer can hold that floor up forever is exactly the question Part 4 turns on.

And finally: yes, the Billionaire Raj estimates are contested. Top-income numbers built off tax data involve assumptions, and serious people have challenged them. Fine, throw the whole series out. You are still left with SEBI’s trading data, RBI’s loan-stress data, the Economic Survey’s own profit-versus-wage numbers, PLFS wages, the NFHS stunting figures, five crore pending court cases, and a peer-reviewed mobility study, all pointing in one direction. And above all you’re left with behaviour. When crores of people move their families’ savings into games with 3-7% odds, they are telling you something about their real prospects that no survey will ever capture.

Part 4: How long we have, what breaks, and the way out

First, the referee

Every fix I’m going to suggest runs through one institution, so we should look at it honestly, because it’s in worse shape than almost anyone admits.

There are 5.34 crore cases pending in Indian courts. About 1.2 lakh of them have been pending for more than thirty years. We have roughly 21 judges per million people; the Law Commission recommended 50, and it recommended that in 1987. We spend less than 0.1% of GDP on the judiciary, and the single biggest litigant clogging it is the government itself, party to about half of everything. Enforcing a contract in India takes 1,445 days on the World Bank’s last count, which ranked us 163rd in the world. Various studies put the cost of all this delay at 1.5 to 2 percent of GDP, every year. Even our flagship bankruptcy reform tells the story: the IBC says resolve a company in 330 days; the actual average is past 700, and lenders recover 33 paise on the rupee.

Now here’s the connection to everything above, which almost nobody draws. When enforcing a contract takes four years, you only do business with people you already trust, which in India means family, caste, community. So firms stay small. Small firms can’t hire at scale, and mass hiring is the only proven cure for Part 1. And when banks can’t recover money through courts, they stop lending to enterprises and lend only against things they can seize without a judge, which is to say your salary and your gold. That gold-loan boom we talked about? It’s partly a court-delay problem wearing a jewellery box. Meanwhile the people who can afford to have exited the courts altogether; Indian companies are the largest foreign users of Singapore’s arbitration centre, year after year. Even justice has a private exit now.

On paper we compete with China. On enforcing a contract, we compete with Congo.

The honest answer to “how long do we have”

Nobody can time a rupture, and I won’t pretend to; Tunisia’s own economists didn’t see theirs coming. What you can do is date the clocks, and there are three of them running.

The demographic one gives us about twenty years. Our fertility rate is already at 2.0, basically replacement, and the working-age share of the population peaks in the late 2030s or 2040s. After that, India starts ageing before it gets rich, which is the trap China walked into at $12,000 per head and we’d be walking into at $4,000-5,000. The window in which a young workforce can still be turned into broad prosperity closes somewhere around the mid-2040s. That’s one investment cycle. Not two.

The political clock is shorter. Delimitation after 2026 moves Lok Sabha seats towards the poorer, younger north at exactly the moment the richer south is paying most of the bills, so the state-level gap from Part 3 turns into a constitutional fight over seats and money within this decade.

And the trust clock has no dial at all, only events. Jantar Mantar is one. The Maratha, Jat and Patidar agitations are another, and think about what those actually are: historically dominant communities rioting to be officially declared backward, which is the single clearest signal Indian politics can produce that people no longer believe the open economy pays. Trust doesn’t decline smoothly. It holds and holds and holds, and then reprices in a day. The nearest thing to a number comes from IMF research: high inequality roughly halves how long a growth boom lasts.

What actually breaks

If nothing changes, my honest ranking of what gives way, in order of likelihood:

The consumption engine goes first, and it’s already going. This is the boring ending and the most probable one: not an explosion but a slow puncture. Mass demand flattens (it already has: FMCG volumes, small cars, affordable housing). Households can’t borrow more, because the debt is at a record and the savings are gone. Companies respond by selling pricier things to the top 10% instead of building capacity for the rest, growth drifts down towards 5%, and the whole catching-up-with-the-world project quietly dies. Brazil and Mexico reached our income level decades ago and have basically stayed there since. Nobody rioted. They just stopped converging. That’s the modal outcome, and every leading indicator in this essay says it has already started.

Second, the household balance sheet, because the stress is visible in RBI’s own numbers: sub-₹10-lakh earners are 78% of the new bad loans, and the marginal borrower is now secured by wedding gold at record gold prices. If gold ever corrects properly, the margin calls land on the one asset the bottom half has left.

Third, the federal bargain, through delimitation and the tax-sharing fight; not secession, nothing dramatic, more like a fiscal cold war of hoarded revenues and competitive freebies while big reforms freeze.

Fourth, the street, in episodes rather than one big bang: every leak, every recruitment scam, every viral billionaire wedding shortens the interval. Worth remembering that Egypt’s economy was growing at 5-7% right up until its revolution. Growth didn’t save anyone, because it wasn’t reaching the average 22-year-old, and the trigger, when it came, looked trivial.

And fifth, the quiet ones: talent leaving from both ends of the pyramid, and concentrated wealth meeting election finance until policy simply stops being written for the other 99%. A democracy doesn’t have to fail loudly. It can just gradually stop working for most of its people.

The way out

You could read all of this and despair, and I want to argue against that, because the genuinely striking thing about the fix list is that none of it is new. Every piece has been done, somewhere, at scale, within living memory. Five things.

Jobs first, because a tight labour market is the only force in history that has compressed inequality without confiscation. That means labour-heavy sectors: factories, construction, tourism, care work, not just capital-intensive production incentives. The target has to be something like a crore of formal jobs a year. For calibration, Germany, with a seventeenth of our population, runs about 1.3 million apprenticeships; we run about five lakh. Vietnam, meanwhile, is eating the factories we keep missing. China ran this entire playbook between 2000 and 2015; there’s no mystery in it.

Second, the courts, and I’d honestly put this above everything else. Going from 21 to 50 judges per million means roughly 50,000 more judges, which fully loaded costs something like 0.1% of GDP, against the 1.5-2% of GDP we burn every year on delay. That’s a fifteen-to-twenty-times return, better than any highway this country has ever built, and it’s the fix that makes every other fix work, because nothing scales in an economy where a contract takes four years to enforce. Alongside: the government imposing settlement discipline on itself, cheque-bounce cases moved to a fast civil track, more NCLT benches, and finished digital land titles.

Third, children, and specifically the pipeline before the exam rather than the bottleneck at it. Stunting is decided by age five and reading by age ten, and ASER says half our Class 5 kids can’t read a Class 2 text. Every rupee spent at the coaching end of the pipeline is fighting over a fixed number of seats; every rupee at the anganwadi end makes the pipe itself bigger. And keep widening the doors: MBBS seats have roughly doubled since 2014, so we know expansion works; add real vocational tracks so that one exam stops carrying an entire generation’s expected value on its back.

Fourth, wages. The Economic Survey has already made the intellectual case; what’s missing is treating it as a monitored problem. Social security that actually follows the worker, minimum wages that are actually enforced, and the profit-share versus wage-share gap tracked the way we track inflation, because a corporate sector sitting on 22% margins while paying real wages below 2017-18 levels is manufacturing its own demand crisis for 2035.

And fifth, the tax-and-competition rebalance. Widen the direct tax net instead of squeezing the same salaried class ever deeper; look honestly at GST’s burden at the bottom; and take profit concentration seriously, because by industry estimates a couple of dozen firms now take most of India’s corporate profit pool, which is Gilded Age geometry. On wealth and inheritance taxes I’ll give you both sides straight rather than a slogan: the American Progressive Era shows this can work, and the 4,300 millionaires already leaving every year shows design matters more than intent. What isn’t negotiable is the direction. You cannot keep taxing labour more for earning less while taxing capital less for earning more, and expect the arithmetic to hold.

The deadline, once more, without drama. The demographic window shuts around 2045. The political one is this decade. The trust one is unknowable, which means you treat it as now. Shift about five percentage points of income share from the top 10% to the bottom 90% over ten years, which is roughly the dose China administered to itself after 2005 and America after 1913, and the bottom-90% income path in 2036 lands about 25% higher, wide enough to restart the consumption-and-investment loop. Miss all three windows and India doesn’t blow up. It just becomes the largest country in history to grow old in the waiting room.

Last thing

If you take one paragraph from all of this, take this one. India’s average income is a number nobody lives in. Our top 1% takes more of the pie than it did under the British, on the British tax system’s own data. Wages have stopped working, by the Finance Ministry’s own admission, so an entire generation has rationally switched from selling its labour to buying lottery tickets, in exams, in options, on the feed, in visas, and the companies selling those tickets are among the best-performing stocks on the exchange. The gap is set in children’s bones before they turn five, drawn across the map ahead of a seat-redistribution fight, coded into a tax system that charges workers more as it pays them less, sorted into jails by who can afford bail, and staffed, ninety lakh guards deep, at the gates between the two Indias.

In a normal economy, a leaked paper is cheating. In a lottery economy, it’s rigging the only casino the poor still believe in. The protest will fade, protests do. The clocks won’t. And the number I’ll be watching isn’t the Gini or the billionaire count. It’s the year young Indians stop buying tickets, because working started paying again. Nothing in the data suggests that year has been scheduled.

Sources: World Bank WDI & Poverty and Equity Briefs; IMF WEO; World Inequality Report 2026 / WID; Bharti, Chancel, Piketty & Somanchi (World Inequality Lab, 2024); Asher, Novosad & Rafkin (AEJ: Applied, 2024); Economic Survey 2024-25; PLFS/NSO; SEBI F&O studies FY22-FY25; RBI Financial Stability Reports; NFHS-5; FAO SOFI; PIB (PMGKAY); Hurun 2025; CFA Wealth Tracker 2026; Infinium Global Research; CMS/NSS Education Survey 2025; BCG/TRAI & Kofluence; YouTube-Oxford Economics 2024; Online Gaming Act 2025; Henley & Partners; NJDG; Law Commission; World Bank Doing Business (final edition); DAKSH; IBBI; NCRB Prison Statistics 2024; CAPSI estimates; UN WPP 2024; ASER; Azam-Chin-Prakash; Berg & Ostry (IMF); Marcellus. “~” means approximate; derived figures are computed from these primaries.

The Cabinet cleared a ₹37,500 crore scheme to push coal gasification. This is the third try.

The first was 2007. JSPL started building India’s first commercial coal-to-syngas plant at Angul in Odisha. It came online in 2014. It still runs well below its design capacity.

The second was January 2024. The same Cabinet cleared an ₹8,500 crore scheme. Eight projects worth ₹6,233 crore in grants are now in different stages of build. The first commercial plant — BCGCL at Lakhanpur — placed its first big equipment orders only four months ago.

Today is the third try. It aims to pull in ₹3 lakh crore of private money and gasify 75 million tonnes of coal a year by 2030.

The scheme exists — that part isn’t in doubt. The Cabinet brief, the PIB release, and the Minister’s press conference all confirm it. The real question: does anything actually change this time?

For 18 years, coal gasification policy has been “in the works” — the Cabinet’s own words from 2024 — and has produced almost nothing at commercial scale. The same five reasons explain every failed cycle:

One, the Coal and Petroleum ministries kept fighting over who got the revenue. They sorted that out in 2022. Two, Indian coal has 30–45% ash. Most global gasifiers were built for coal with half that ash. We had to develop our own technology, which took ten years. Three, nobody wanted to buy syngas products. Urea is subsidised, methanol blending stayed under 1%, and SNG had to compete with cheap LNG. Four, the economics didn’t add up — ₹40,000–50,000 per tonne of installed capacity, five to six years to build, NPV negative under most coal-oil-rupee scenarios. Five, no carbon price. So the cleaner option was also the more expensive one.

The 2024 scheme dealt with reasons three and four by throwing money at the problem. The 2026 scheme makes those grants 4.4× bigger and reshapes them to attract big private capital. Neither fixes the ash problem. And the carbon math is coming back as a headwind — India’s carbon trading scheme starts setting benchmarks this fiscal year.

What Will Coal Gasification Actually Replace?

The pitch is import substitution. India bought ₹2.77 lakh crore worth of stuff last year that coal gasification can — in theory — replace. Look closely at what makes up that ₹2.77 lakh crore.

What sits inside the ₹2.77 lakh crore import bill that coal gasification is supposed to replace

Coking coal alone is ₹85,000 crore — but coal gasification only replaces this one indirectly, through the syngas-to-DRI route which itself competes with electric arc steel. LNG is ₹1.2 lakh crore but only the industrial slice swaps in; transport-grade LNG doesn’t. Methanol and ammonia — the cleanest cases — add up to roughly ₹16,000 crore. Real, but not earth-shaking.

Now run the optimistic math. Say India hits 100 MT of coal gasification by 2035. That’s five years past the official target, which is the honest timeline given how long these plants take to build. Annual import substitution at that point: ₹60,000–90,000 crore.

That’s 22–32% of last year’s substitutable import bill. The import bill itself grows in the meantime.

One more thing on the briefing: the Minister described ammonia as “100% imported today.” That isn’t accurate. Technical-grade premium ammonia is mostly imported. Bulk ammonia for captive urea production is largely made in India. The substitution case is real on the premium slice. The headline statement was sloppy.

How The 2026 Scheme Differs From The 2024 One

The 2026 scheme isn’t a bigger version of the 2024 one. It’s a redesign.

Look at the caps. A 20% subsidy on plant-and-machinery cost — which is roughly 60% of total project capex — means the ₹5,000 crore per-project cap kicks in around ₹42,000 crore of project capex. The ₹12,000 crore per-group cap kicks in at about ₹1 lakh crore of group capex.

Read this as a deliberate choice. The scheme is built to pull in two to four really big projects — think the New Era Cleantech model in Vidarbha, ₹20,000+ crore per cluster — rather than 25 smaller builds. A single corporate group can bid for up to roughly ₹60,000 crore of capex before the subsidy starts getting diluted.

The other big change: no PSU set-aside. The 2024 scheme reserved ₹4,050 crore for PSU-led projects. The 2026 scheme is fully open. That’s a clear shift toward private capital.

Where does the trade sit

It sits in the equipment and services that take 30–40% of project capex but get no headlines.

Three pieces dominate roughly half the plant capex: the gasifier vessel (20%), the air separation unit (17%), and downstream synthesis (13%). The first and third go to primary EPCs — BHEL and L&T. The second is the under-priced ancillary play I'll come back to. The remaining 50% — pipework, refractories, water treatment, instrumentation, captive power, civils — spreads across mid-cap and small-cap suppliers that nobody is putting in a gasification basket today.

What 25 Years Of Chinese Coal-To-Chemicals Tell Us

The spine of this whole piece is the China comparison. The question isn't whether coal gasification can work at scale — it can; China proved it. The question is what we should realistically expect from India's late entry. So let me spend some time on this comparison rather than gesture at it.

Some facts to anchor the rest. China consumes ~380 million tonnes of coal a year in coal-to-chemicals. Roughly 70% of Chinese methanol output is coal-route. Roughly 80% of Chinese ammonia is coal-route. The Shenhua Ningxia coal-to-liquids plant alone gasifies 44,000 tonnes of coal per day — almost double India’s entire commercial coal gasification capacity today.

How China actually built this

China’s coal-to-chemicals push started with the 11th Five-Year Plan (2006–2010). Three drivers underwrote it. Energy security after the 2003–2008 oil price spike. Abundant cheap coal in the western provinces — Inner Mongolia, Shaanxi, Ningxia, Xinjiang. And state-owned coal majors (Shenhua, China Coal) that could underwrite mega-projects at command-economy capital cost.

The actual buildout had a clear arc.

The marquee plants tell the same story.

The Chinese plants made money — for a while

Between 2010 and 2014, when Brent was $100+ and Chinese coal was $40–60 a tonne, most of these plants ran with healthy cash margins. Shenhua Ningxia’s published unit economics were $50–60 per barrel of liquid product — competitive against $80+ Brent.

The 2014–2016 oil crash flipped the math. Several smaller methanol units shut down. The big plants kept running because the state absorbed the losses. By 2017, Beijing stopped approving new coal-to-liquids projects on emission grounds. Coal-to-olefins kept building but at a much slower pace. The aggregate capex on Chinese coal-to-chemicals has dropped roughly 60% from its 2014 peak.

The detailed parameter-by-parameter comparison

Three things India can usefully copy from China

One: a commercial coal-to-chemicals industry is possible at scale. China demonstrated the technology stack, the unit economics, and the industrial cluster model. India isn’t trying something untested. It’s trying something that worked 14 years ago somewhere else. That’s a meaningful de-risking of the technology question — but only of the technology question.

Two: build in the coal belt, not near the consuming market. China built mostly in Inner Mongolia, Shaanxi, Ningxia, Xinjiang — where coal is abundant and cheap. The chemicals get railed or piped out. India’s 2024 scheme follows this — Lakhanpur (Odisha), Sonepur Bazari (West Bengal), Vidarbha (Maharashtra), Niljai (Maharashtra) are all coal-belt locations. This is one place the policy got it right.

Three: state support has to be sustained, not one-shot. China’s incentives ran for ~15 years and were renewed through multiple Five-Year Plans. India’s ₹37,500 cr scheme is large but one-shot. The Chinese experience suggests we’ll need a second incentive tranche by FY30 — particularly if oil cycles weakly between FY28 and FY32 when most Indian plants try to ramp.

Three reasons India can’t replicate China’s speed

One: Indian coal has 2–3× the ash content. China’s Inner Mongolia coal averages 8–15% ash. Indian Gondwana coal averages 30–45%. This isn’t a problem you solve once and forget — it’s an ongoing operational drag on yield, ash disposal, and equipment life. Indian-coal gasifiers will always run 10–15% more expensive per tonne of product than Chinese-coal ones, structurally. The indigenous tech we’re celebrating (BHEL PFBG) makes high-ash usable. It doesn’t make it cheap.

Two: water is a binding constraint here in a way it wasn’t there. Coal gasification consumes 8–12 cubic metres of water per tonne of coal. China rationed coal-belt water through state allocation — projects got the water they needed because the state said so. India’s coal-belt states face genuine summer water stress, and allocations get challenged in court and at the NGT. Expect 6–12 month average delays on commissioning purely from water clearance disputes.

Three: India can’t take environmental shortcuts. China built its first wave of coal-to-chemicals plants with effective EIA bypass. India’s plants face Environmental Clearance as the binding constraint at 12–18 months — before NGT challenges, which can add another 12–24 months. The 5–6 year build cycle isn’t poor execution. It’s the system working as designed in a democracy with environmental rule of law.

The Contrarian Tail

1. The catalyst supplier is the next import dependence

Every coal-to-methanol plant burns through $15–25 of catalyst per tonne of product over a catalyst cycle. Similar spend on iron-based catalysts in ammonia. Across 25 plants over 20-year lives, that’s $8–12 billion of cumulative catalyst purchases.

This market belongs to Johnson Matthey (UK), BASF (Germany), Haldor Topsoe (Denmark), and Clariant (Switzerland). No Indian player at scale. The scheme creates a brand new import dependence — on consumables instead of feedstock — that nobody is addressing. Aatmanirbhar Bharat at the gasifier shell. Still imported at the catalyst layer.

2. Petronet LNG is the un-discussed short

One of the 2024 scheme projects — CGIL’s plant at Sonepur Bazari — will pump 1.83 MMSCMD of synthetic natural gas into GAIL’s grid. If the 2026 scheme produces six similar SNG projects (the unstated but logical split), India adds 10–14 MMSCMD of coal-route gas into the system by FY33.

That displaces 10–15% of LNG terminal regassification throughput. Petronet’s Dahej contracts hold; the marginal economics don’t. By FY32, this is a real offtake risk worth pricing in. The same scheme that lifts GAIL hurts Petronet LNG. Only the first is being talked about today.

3. The carbon cost is invisible today. It won’t stay invisible.

Coal gasification without carbon capture puts out about 3.5 kg of CO₂ per kg of methanol — versus 0.5 kg via the gas route. Similar math for ammonia and urea. India’s Carbon Credit Trading Scheme Phase 1 starts setting benchmarks this fiscal year. By the time the 2026 scheme’s plants commission in FY32–FY34, coal-route plants will face explicit carbon liability — plausibly ₹500–2,000 per tonne of CO₂, depending on how strict the benchmarks are. EU CBAM, in force from January 2026, taxes coal-route ammonia at the border — which limits the export option.

The scheme is best understood as import substitution at a moderate carbon premium, not as clean energy. ESG funds will reject it. Sovereign and energy-security funds will buy it. The split between those two pools is itself something to watch.

The Seven Names That Pass The Test

Seven names pass the four-rule test. They sit at different points on the plant. Some are obvious only when you look closely. Some are mid-caps the market currently puts in different buckets. Together they make up the differentiated trade.

1. Engineers India Ltd

EIL runs project management for every plant in the pipeline. Each project pays them ₹50–100 cr in fees. Across 25 plants that’s ₹1,500–2,500 cr cumulative — and these are pure-margin earnings with no execution risk. The market still treats EIL like a sleepy PSU consultancy. The order book visibility through FY28–FY34 is what eventually rerates the stock.

TAM share: 15–25% of PMC

2. VA Tech Wabag

Wabag built the wastewater treatment and zero-liquid-discharge system for JSPL’s gasification plant at Angul. That single reference makes them the only Indian water company with an operational coal-gasification track record — decisive when EPCs build their technical pre-qualification list on every new tender. Coal gasification needs 8–12 cubic metres of water per tonne of coal; ZLD is mandatory. The market files Wabag under “municipal water” and misses the industrial side. The Angul reference is buried in a 2024 academic chapter, not on their investor deck.

TAM share: ~30–40% of ZLD

3. Ratnamani Metals & Tubes

Stainless seamless pipes — austenitic, duplex, super-duplex. Coal gasification piping runs at 40–250 bar in wet H₂S, ammonia, and methanol vapour service. That’s exactly the metallurgy Ratnamani is approved for across refineries, fertilizers, nuclear, and petrochem. Process piping is 7% of plant capex; cumulative addressable share is ₹3,000–4,000 cr over the cycle. That maps to an 8–10% revenue tailwind on top of the base business for years.

TAM share: ~15% of piping

4. Anup Engineering

This one surprised me. Anup already supplied HURL’s ammonia-urea fertilizer complex — exactly the equipment class needed for coal-to-ammonia plants. They built hot-box reactors for HPCL Mumbai’s hydrogen generation unit, directly applicable to syngas reactors. Gas purification technology licensors have approved them for PSA and TSA equipment — very few Indian fabricators carry these approvals. They’re a global licensee of Lummus Technology’s Helixchanger heat exchanger IP. A small-cap pure-play almost nobody is positioned in.

TAM share: ~20% of process equipment

5. Linde India

Air Separation Units are the bottleneck nobody talks about. 75 MT of coal gasification needs about 165,000 tonnes per day of pure oxygen — that’s 30–35 large ASUs, roughly 50× India’s current industrial oxygen base. Linde India is the only listed company with proprietary cryogenic ASU tech. They run a build-own-operate model with take-or-pay offtake — the highest-quality earnings stream in the whole stack. Catch: premium multiples already, and a possible Inox Air Products IPO would create competition. Buy for steady compounding, not for a rerating.

TAM share: ~30% of ASU build

6. Orient Refractories

Smaller than IFGL or Vesuvius, but Orient has a decisive parent: RHI Magnesita is the world’s #1 refractory company, with operating coal gasification references in China and South Africa. Gasifier linings need chrome-magnesite, chrome-alumina, and silicon carbide at 1,200–1,500°C — RHI’s specialty. Linings need replacement every 18–30 months, so this turns into recurring revenue rather than one-shot capex.

TAM share: ~25–30% of linings

7. ISGEC Heavy Engineering

The market reads ISGEC as a sugar + presses + EPC conglomerate. The chemical-process equipment side is mispriced. They’ve built an ammonia converter using Haldor Topsoe’s process. They make CFBC boilers in JV with Foster Wheeler — directly adjacent to BHEL’s PFBG. They make pressure vessels for fertilizer plants. Cumulative addressable share over the cycle is ₹10,000–12,500 cr.

TAM share: ~20% of process equipment

The honest read: India has been failing at coal gasification for 18 years. The 2026 scheme finally fixes the financial and policy bottlenecks. It doesn't solve the high-ash technology problem at commercial scale, and it doesn't deal with the carbon math that gets worse as CCTS and CBAM tighten.

— Piyush

This is opinion, not advice. May hold positions in some of the names mentioned.

equitybypiyush.substack.com

I’m glad we went. Not just for the views, though I’ll remember those for a long time. But because something about being in a place that doesn’t care about you — that will test you regardless of your mood or your intentions — cuts through a lot of noise. I came back thinking more clearly than I have in months.

This is what I brought back.

---

Before you even leave for the hills, you have to be ready. Not halfway ready. Not “I’ll figure it out” ready. Actually ready. Because the mountain doesn’t pause while you get your bearings. The ice doesn’t wait for you to check your tyres. You either came prepared or you didn’t, and you’ll find out which very quickly.

I watched people that day who hadn’t taken it seriously. Wrong shoes. No chains. Windows fogging up and no idea why. They weren’t reckless people. They just hadn’t respected the mountain enough to prepare for it. And they were the ones stuck, frightened, waiting for someone else to solve it.

I’ve made exactly those mistakes in markets. Entered positions without knowing what would make me exit. Sized as if the thesis was certain when it wasn’t. Told myself I’d manage the risk later, as if later was a real plan. It never is. The market, like the mountain, doesn’t warn you before it tests you. When the test comes, your preparation either was real or it was comfort theatre. There’s no middle ground and no negotiating with it.

But the people who were prepared — and you could tell immediately who they were — they were the ones actually enjoying the day. Stepping out into the snow, laughing, taking photographs. Not because nothing could go wrong. Because they’d already thought through what could, and they were ready. That kind of readiness doesn’t just protect you. It frees you. It’s the difference between watching the portfolio with your jaw clenched and actually being present in the journey you signed up for.

---

On the road itself, the pressure to go faster is constant and almost entirely an illusion. The car that overtook us on the narrow stretch — impatient, confident — we passed him twenty minutes later, tilted into the shoulder, someone already stopping to help. Slow on ice isn’t timid. Slow is what keeps you on the road. Slow is what gives you time to read what’s coming before it’s already on top of you.

I’ve spent years confusing speed with progress in investing. Needing to have a view on everything. Needing to act on it. The anxiety of holding still while others seemed to be moving. What I’ve slowly learned — and it is slow, this particular lesson — is that the returns I’m most proud of came from doing almost nothing. Buying something real and then sitting in it through the boredom, through the months when nothing happened, through the period when it looked wrong to everyone including sometimes me. That’s where compounding lives. Not in the activity. In the patience.

And then small things. On that mountain, what kept us smooth wasn’t one big decision. It was a dozen quiet ones. Tyres right. Knowing when to brake and when to coast. Warm flask. Phone charged. None of it dramatic. All of it essential. The day went well because of details, not brilliance. I think about my process the same way now. The edge isn’t the great call. It’s the checklist you actually use. The position size rule you don’t override. The exit criteria written before you’re emotionally attached to being right.

We rolled down the window and asked. A man who’d driven that road a thousand times gave us five minutes that were worth more than everything I’d read about the route. He wasn’t polished. He didn’t have a deck. He just knew — which patch ices first, which bend looks safe but isn’t, where to stop if the weather turns. That knowledge doesn’t live in apps or travel blogs. It lives in people who’ve actually done the thing in all conditions. We listened more than we talked. We took the route he suggested. It was right.

I keep coming back to this with research. Primary sources — people who actually know the terrain — beat secondary ones every time, without exception. The investor who’s spent months talking to distributors in a supply chain sees something that isn’t in any broker note. The person who’s sat across from management in a bad quarter understands something the transcript won’t tell you. When you find someone who actually knows, stop performing intelligence and just listen.

---

At some point the weather shifted. Not suddenly — these things rarely are. Just the light changed, the road surface looked different, the temperature dropped in a way you felt before you measured it. The instinct was to push through. We’d come this far. Turning back felt like conceding something.

But there’s a kind of knowing that lives below the ego — the part of you that recognises the difference between conviction and stubbornness. We found shelter. Waited. An hour later the conditions had changed and we moved. We got to where we were going. Just not on our original timetable, and that was fine.

I’ve watched people — myself included — blow up positions not because the thesis was wrong but because they couldn’t take shelter. Couldn’t sit in cash. Couldn’t admit that right now, in this weather, the right move is to wait. Adding to losing trades because stopping would mean admitting the market is smarter than you today. It always is, on some days. The capital that survives the storm is the capital that gets to act when things clear. That’s not a small thing. That’s the whole game across a long career.

There was a car in the snow. Buried, really — past the axles. The people inside had the look of people who’d been sitting there long enough to stop expecting help. We stopped. Helped dig them out. What else do you do.

I think about how many people are navigating their financial lives without a map, without anyone who’s been on this road before. Not because they’re careless — because nobody gave them the right information at the right time. If you understand this terrain and you don’t share what you know, that’s a choice. A poor one. Help people who are stuck when you have the means to help them. That’s not wisdom. That’s just basic.

---

Here’s what I keep coming back to. It snowed in Sissu in April. Nobody forecast it. Nobody expected it. And as we were leaving — coming down the mountain, job done, experience had — we passed a stream of cars heading up. Into the snow. Into the conditions we’d already been through and were now leaving behind.

I don’t say this to be smug. I say it because it captures something about how things actually work. By the time everyone is moving toward something, the people who were already there are on their way home. The best of it is already done. The crowd arrives after the experience, not during it.

Every investment that’s genuinely worked for me was uncomfortable to hold when I bought it. Quiet. Unloved. Sometimes I was the only person I knew who thought it made sense. That loneliness isn’t a bad sign. It’s often the sign. By the time it’s obvious, by the time everyone is driving up the mountain, you should already be thinking about the road back.

But you have to stay. That’s the other half. Once you’re in something real, the work becomes endurance. Staying in it while the noise argues against you. Not selling the run short because waiting got uncomfortable. The themes that actually pay — the multi-year cycles, the quiet compounders, the structural shifts that take longer than anyone expects — they reward the investor who can just stay in the snow long enough to see what it becomes.

We stepped out of the car at one point and just stood in it. Snow coming down, the valley below us, nothing moving. Nobody said anything for a while. There was nothing to say. We were already there. That was the whole point.

---

I’m still thinking about that trip. I think I’ll think about it for a long time. Not because it was perfect — parts of it were genuinely uncomfortable and I made decisions I’d make differently. But because the mountain is honest in a way that most things aren’t. It shows you exactly who you are and how prepared you actually were. No flattery. No allowances. Just the road and the conditions and what you do with them.

That’s all investing is too, really. The market is honest the same way. It doesn’t care about your narrative or your confidence or how good the thesis sounded when you wrote it. It shows you, over time, whether the preparation was real.

I came back wanting mine to be real. Still working on it.

---

Sissu · April 2026

The Memo covers the competitive landscape, demand drivers, all the risk variables — and makes the conclusion.

Its the one of the best bets when all the risk are playing out for it and market is discounting the worst….The article starts here…..

Airlines are capital-intensive, fuel-exposed, cyclical, regulated, and compete on price in markets where every carrier is trying to go bankrupt more slowly than the next one. The history of aviation is a history of capital destruction on a civilisational scale.

And yet: Ryanair has compounded at 18% annually for 25 years. Southwest was profitable for 47 consecutive years before COVID finally broke its streak. The low-cost carrier model — when executed with genuine cost discipline in a structurally growing market — produces returns that completely defy the industry’s reputation.

IndiGo holds 63% of India’s domestic aviation market. It has 434 aircraft in the air and 900+ more on order. It’s sitting on ₹37,000 crore in cash with zero net debt. And right now, after an operational meltdown in December and its CEO walking out the door, the stock sits roughly 30% below its 52-week high.

So: buying opportunity, or value trap? I’ve spent the last month going deep on the financials, the competitive dynamics, and a few numbers buried in the MCA filings that I don’t see anyone talking about. Here’s everything.

A Duopoly Where Only One Side Knows How to Make Money. IndiGo Earns ₹7,000 Crore. Air India Loses ₹15,000 Crore. Same Market. Same Routes. Same Passengers.

Air India has lost ~₹50k Cr since its divestment in Jan 2022— entirely funded by Tata Sons equity injections

IndiGo is the dominant, cash-generative operator in India’s fastest-growing aviation market, while Air India Group — despite scaling revenue rapidly — is still burning ₹16,000+ Cr annually in combined losses funded by Tata Sons’ balance sheet, with Air India Express now EBITDA-negative, creating a window of competitive advantage that could last through FY28.

India's Aviation Market Is the Easiest Opportunity and the Hardest Business — At the Same Time. By the Time Any New Airline Figures This Out, IndiGo Will Have Flown Another Billion Passengers.

Let me give you the one number that frames this entire thesis.

India has 0.22 air trips per capita annually. The United States has 2.7. Europe averages 1.8. China — a country at roughly twice India’s per-capita GDP — is at 0.44.

If India simply reaches China's current penetration rate by 2030 — not Europe's, not America's, just China's — the market nearly doubles. That's 300 million additional passengers annually in four years.

No other aviation market of this scale is simultaneously at this early a stage of penetration. China was here in 2012. India is here in 2026, with a growing middle class of 400 million new consumers, digital payments making ticket purchase frictionless, and an LCC that has trained an entire generation of first-time flyers to think domestic air travel is affordable.

That’s the macro. Now let’s get into the business.

IndiGo is three businesses. Most people only price in one.

This is where most aviation analysis goes wrong. People model IndiGo as a single P&L. It isn’t. There are three distinct businesses running under the same livery, with very different economics, growth profiles, and risk factors.

Business ① — The Domestic LCC Machine (~78% of revenue)

This is the core, and it’s exceptional. IndiGo earns 28–30% EBITDAR margins on the world’s fastest-growing large aviation market. The moat isn’t brand. It’s frequency.

On Delhi–Mumbai, IndiGo operates 30+ daily frequencies. Even if a competitor undercuts on price by ₹500, the business traveler takes IndiGo because there’s a flight every 30 minutes. That lock-in is structural, not marketing-driven. It’s the Ryanair play — but in a market with demographics Ryanair never had.

Business ② — The International Expansion Bet (~15% of revenue)

IndiGo has ordered 69 Airbus A321XLRs — a narrow-body with 8,700 km range that can fly Delhi to Athens nonstop. The unit economics are genuinely competitive: single-aisle CASK of ~$0.05–0.06/ASK versus ~$0.07–0.09 on the wide-bodies that Emirates and Turkish operate.

The problem is product and loyalty. Emirates has 270 wide-body aircraft, a 5-star cabin, and 30 million Skywards loyalty members. The India–London business traveler paying ₹2–4 lakh for a flat-bed will not move to a 7-hour narrow-body to save ₹30,000. IndiGo will almost certainly win the price-sensitive leisure segment — visiting family in London from Amritsar, that sort of thing. Whether it cracks corporate is a 5-year open question. The committed investment is ₹18,000 crore.

Business ③ — The Hidden Financial Company (₹5,200 Cr income, FY25)

This is the one nobody talks about, and it changes the entire valuation story. I’ll give this its own section because it deserves it.

The number buried in the FY25 filing that changes everything

IndiGo reported ₹7,258 crore in PAT for FY25. That headline number is real. But look at what generated it.

IndiGo is sitting on ₹36,945 crore in cash and equivalents — roughly $4.2 billion — earning 13–14% in Indian treasury instruments. That cash hoard generated approximately ₹5,200 crore in treasury income in FY25 alone.

Do the arithmetic: ₹7,258 Cr reported PAT minus ₹5,200 Cr treasury income = ~₹2,058 Cr from actually flying aeroplanes.

At ₹4,388/share, the reported P/E is 53x. But the airline's operating P/E — stripping out what the cash pile earns — is closer to 80–100x. You are partly paying for a treasury operation. That's not inherently bad (the cash is real, it compounds, and it will eventually fund the IFSC aircraft ownership programme that permanently lowers CASK). But you need to know it exists before you decide what IndiGo is worth.

The seven risks — ranked by what actually matters

Risk ① — Crude oil (biggest single variable)

ATF is 28–32% of IndiGo’s total cost. Every $10/barrel move in Brent crude translates to roughly ₹800–900 crore in annual PAT — which is 10–12% of FY25 profits. Brent sits at $82 today, which is base-case territory. A move to $65 (plausible given Trump’s Iran rhetoric) would be a ₹3,500 crore annual tailwind. A move to $105 from a Middle East escalation would be a ₹2,000+ crore headwind.

The critical detail: IndiGo does not hedge fuel systematically. Ryanair hedges aggressively. IndiGo’s founders have always believed that hedging costs more than it saves at scale over a full cycle. That’s a legitimate position — but it means IndiGo’s PAT is a direct function of crude, more so than almost any airline of comparable size.

Risk ② — The pilot shortage (most underweighted)

December 2025 wasn’t a scheduling failure. It was arithmetic catching up with reality.

IndiGo adds ~57 aircraft per year. Each aircraft needs roughly 14 pilots. That’s 800 new pilots just for new planes — before replacing the 8–10% of the existing 9,000+ pilot workforce that leaves annually. Total annual requirement: 1,500–1,700 pilots.

India’s DGCA-approved flying schools produce about 900–1,000 graduates a year. IndiGo’s own IGRUA academy contributes 300 at full capacity. International hiring requires a 2-year DGCA validation window. Net annual supply available to IndiGo: roughly 600–750 pilots.

Annual deficit: approximately 900–1,000 pilots. Every year. This is not solved by one management intervention. Unless DGCA accelerates international licence recognition, every time IndiGo pushes growth, it runs into this ceiling and DGCA forces schedule cuts. This will recur.

Risk ③ — Air India (the 5-year threat, not the 2-year one)

Tata Sons has committed fully — 470 aircraft ordered, Vistara merged in, Singapore Airlines as a strategic partner with its operational expertise. Air India is burning roughly ₹4,000 crore at the EBITDAR level annually right now. Tata is absorbing it. The question is not whether Air India recovers; it’s when, and what happens to IndiGo’s pricing power when it does.

Risk ④ — SpiceJet exit (this one’s actually a positive for IndiGo)

SpiceJet has negative net worth, ongoing lease defaults, and lessor repossessions. The probability of formal exit or grounding within 12 months sits at 40–50%. If it goes: IndiGo and Air India absorb 3–4% market share, and IndiGo — with its cash and order book — gets first pick of SpiceJet’s 50+ metro airport slots. A SpiceJet exit is worth an estimated ₹1,000–1,500 crore in annual PRASK uplift for IndiGo. Watch lessor repossession news as the leading indicator.

The competitive moat — is it real?

IndiGo’s CASK of ₹2.96/ASK versus Air India’s estimated ₹4.05–4.30/ASK is a 27–35% structural cost advantage. Here’s where it comes from — and which parts are durable.

The lease cost advantage is the most structural and most underappreciated. IndiGo locked its Airbus order pricing before the 2020s inflation cycle. Air India is ordering at current market rates. That gap doesn’t close until IndiGo’s older aircraft cycle off the books — which won’t happen fully until mid-2030s.

The Akasa comparison is worth noting: Akasa’s CASK (₹2.85–2.95) is similar to IndiGo’s. Akasa has the same single-family fleet DNA, no legacy costs, and fresh aircraft. IndiGo’s advantage over Akasa is scale, network density, and slot access — not unit economics. That matters for how you think about long-term pricing power.

The CEO question is the most important thing right now

Rahul Bhatia co-founded IndiGo at 60. He built it from zero to India’s dominant carrier in a decade, created a cost culture unlike anything else in Asian aviation, and knew every regulator, supplier, and operational lever in the business. He stepped back in after December 2025, and for the phase the company is in — restoring discipline, rebuilding DGCA trust, reasserting operational authority — he is the right person.

The question is what happens after. Running a 434-aircraft operation with 2,200 daily flights and 35,000+ employees requires world-class professional management infrastructure. Every month IndiGo goes without a permanent CEO appointment is a governance discount on the stock.

The CEO wildcard: if IndiGo appoints someone with genuine LCC discipline — Ryanair, Wizz Air, AirAsia pedigree — the stock could re-rate 15–20% purely on management premium, independent of any operational improvement. That's the single most asymmetric near-term catalyst in the stock, and it costs nothing to wait for.

So should you buy it?

Let me be direct about the bear case first, because intellectual honesty requires it.

The pilot deficit is structural and won’t resolve in 12 months. Air India is on a recovery trajectory that will eventually compress IndiGo’s pricing power. The international expansion is a ₹18,000 crore bet on a product category (long-haul leisure on narrow-bodies) that has never been proven in India. And at 53x reported P/E — or 80–100x stripped of treasury — this is not a cheap stock by any conventional metric.

Now the bull case.

NMIA opened March 29. Mumbai’s slot constraint — the binding ceiling on IndiGo’s highest-yield routes for the last decade — just broke. Navi Mumbai is a 90 mppa airport at full build-out. SpiceJet’s exit, if it comes in the next 12 months, hands IndiGo another 50+ metro slots and removes a source of low-fare competitive pressure. Crude at $82 and trending lower is a tailwind.

ATF’s inclusion in GST — a policy the industry has lobbied for since 2017 — would deliver ₹1,500–2,500 crore in structural annual savings.

And a quality permanent CEO appointment, the most immediately actionable catalyst, could re-rate the stock 15–20% before any of the operational improvements show up in results.

The reverse DCF is the clearest framing: at ₹4,388, the market is pricing in IndiGo achieving roughly 12% EBITDAR CAGR for 10 years and reaching ~480 million passengers annually by FY35. That implies IndiGo maintaining 63% share of an Indian market of ~750 million passengers. It is plausible — it is the base case. The bull case isn’t in the price.

The question is not whether India's aviation market becomes enormous. It will. The question is whether this is the right price to get in. At ₹4,388, with everything I know about this business, I think it is — with appropriate position sizing for the risks above.

Disclaimer: This article is based on publicly available data — MCA filings, exchange disclosures, DGCA traffic data, and company annual reports. It does not constitute investment advice. All figures in Indian Rupees unless stated otherwise. The author may hold positions in securities mentioned.

If you found this useful, the full institutional version covers all seven key variables in detail — complete crude and FX sensitivity tables, Air India MCA forensics, the full pilot supply model, airport slot analysis, and a complete financial forensics section with every adjustment explained. Hit reply if you’d like that version.

Everything about GLP Market

Published February 2026 · 12 min read · Covers 17 countries, 30+ companies, and the next decade of the most competitive pharmaceutical race in history

If you’ve heard of Ozempic, Wegovy, or Mounjaro, you already know the pitch: injectable drugs that help people lose dramatic amounts of weight, control blood sugar, and maybe prevent heart disease along the way. Novo Nordisk and Eli Lilly became two of the world’s most valuable companies because of them.

What you might not know is that the walls protecting those blockbuster drugs are starting to crack — and in some countries, they’re already gone.

This is the story of what happens next.

📊 Key Global Metrics — 2026 Snapshot

Why 2026 Is the Year Everything Changes

Novo Nordisk’s semaglutide — the molecule behind Ozempic and Wegovy — earned the company roughly $26.7 billion in 2026 alone. For years, airtight patent protection in every major market kept generic competition completely off the table. Challenging those patents was expensive, time-consuming, and legally treacherous.

But patents expire. And in four major markets — Canada, India, China, and Brazil — semaglutide’s core patent either lapsed or is expiring in early 2026. That’s not a minor technicality. That’s the starting gun for a generics race unlike anything the pharmaceutical industry has seen in this century.

The Three Patent Cliff Waves:

The US market — $20+ billion in annual GLP-1 sales and counting — stays protected until at least 2031. That’s not a mistake. Novo Nordisk has constructed a complex web of secondary patents around the device, the formulation, and the delivery mechanism. The core molecule patent may expire, but the practical barriers to a generic launch in the US don’t disappear until the entire stack unravels.

Europe is in the same boat. So is Japan. Generic companies are preparing, filing dossiers, and positioning — but none of them are actually launching anything meaningful in those markets until the early 2030s.

Canada, though? That’s happening right now.

The Canada Story — Why Every Generics Company Is Watching It Closely

Canada has become the most strategically significant generics market in the world for semaglutide, and the reason is almost paradoxical: it’s a regulated, Western market with real pricing power — and yet its core patent lapsed in January 2026. No other market of this calibre is accessible this early.

That combination — regulated-market prices paired with early generic entry — is why you have Sandoz, Dr. Reddy’s, Hikma, Apotex, and even US telehealth company Hims & Hers all chasing the Canadian approval simultaneously. The current branded price runs CAD $200–300 per month. Generics are expected to come in at CAD $30–80 per month. That’s a 50–80% price collapse, and it creates a massive access opportunity for the roughly 9 million Canadians living with obesity and 3.7 million with type 2 diabetes.

There’s one more wrinkle that makes Canada particularly interesting from a manufacturing standpoint: the dominant dosage format here isn’t a standard pre-filled pen — it’s a DDC (Drug Device Combination) cartridge that loads into a reusable pen device. That requires a very specific type of manufacturing facility called a fill-finish CDMO with DDC capability. And that brings us to arguably the most interesting company in this entire landscape.

The Manufacturing Race No One Is Talking About

Making a semaglutide generic isn’t like making a generic painkiller. The drug is a peptide — a chain of amino acids that requires highly specialized synthesis, refrigeration throughout its supply chain, sterile fill-finish manufacturing, and — for the Canadian DDC format — assembly into a cartridge compatible with a re-usable pen device. Getting one of those steps wrong means a contaminated batch, a regulatory rejection, or a lawsuit.

This is why the CDMO (Contract Development and Manufacturing Organization) landscape matters enormously. The companies building this infrastructure are as central to the GLP-1 story as the generics companies themselves.

That $16.5 billion Catalent acquisition deserves a longer look. Catalent was one of the world’s largest drug manufacturing service companies — a neutral CDMO that worked with dozens of pharma clients. When Novo Nordisk’s parent holding company bought it in March 2024, they essentially took a massive piece of global fill-finish manufacturing capacity off the open market and locked it exclusively for Novo products. That’s not a supply chain decision. That’s a competitive moat strategy at industrial scale.

The Global Generic Race, Country by Country

Here’s a picture of how the competitive landscape looks across the major generics-accessible markets right now.

India — 101 Million Diabetic Patients and a Court Battle

India is one of the most dramatic markets in the GLP-1 story. Dr. Reddy’s Laboratories received CDSCO approval for a generic semaglutide, celebrated briefly — and then got hit with an interim injunction from the Delhi High Court at Novo Nordisk’s request. So the drug is technically approved but legally blocked from sale domestically, at least for now.

Meanwhile, Biocon, Cipla, Sun Pharma, Lupin, and Alkem are all in various stages of development or filing. Lupin’s CFO said the Indian market alone could be worth $1 billion per year at generic prices. At current branded prices of ₹8,000–15,000 per month, the 101 million diabetic and 180 million+ obese patients in India are almost entirely priced out. A generic at ₹700–1,200 changes that equation completely.

China — 15+ Generics Fighting Over 140 Million Diabetics

China is the most crowded competitive battlefield on earth for generic semaglutide. Fifteen or more domestic generic programs are in late-stage clinical trials or preparing regulatory submissions. Jiangsu Hengrui — China’s largest domestic pharma — had Novo’s patent overturned at the Supreme People’s Court, though Novo is appealing. The price is expected to collapse from CNY 2,000–4,000 per month down to CNY 150–350 within five years of generic entry.

What makes China even more extreme is the Volume-Based Procurement (VBP) mechanism — China’s government-mandated competitive tendering process that compresses drug prices to the floor rapidly. In other categories, VBP has driven price erosion of 90%+ in just a few years. The same dynamics will apply here.

Brazil — Novo Tried to Stop the Clock and Couldn’t

Novo Nordisk went to Brazil’s Supreme Court to seek a patent extension. The court refused. That single ruling opened the door for March 2026 generic entry, and Hypera Pharma — Brazil’s largest generic company — has been preparing a first-mover launch since the decision came down. Biocon is partnering with local company Biomm SA. EMS Pharma is filing. The current price of BRL 1,500–3,000 per month is expected to fall 82–90% over five years.

Saudi Arabia — Already Happened

Here’s something that barely got coverage outside of specialist circles: the first approved generic semaglutide in any regulated market globally launched in Saudi Arabia in February 2026. Hikma Pharmaceuticals, working with its CDMO partners, received SFDA approval and got to market. It’s not the biggest market — but it’s a proof point that the generics pathway works, and it’s a live commercial precedent that every other generic company is now citing in their regulatory submissions.

What Happens to Price — And What That Means for Patients

The price erosion data is where this story gets truly consequential. These aren’t just competitive dynamics between pharma companies — they’re about whether billions of people can afford drugs that could transform their health.

The US number deserves emphasis: Americans currently pay $900–1,100 per month for Ozempic or Wegovy without insurance. Even with insurance, co-pays can be brutal, and many plans don’t cover obesity medications at all. The 37 million Americans with type 2 diabetes and 108 million living with obesity won’t see meaningful generic relief until at least 2031. That’s five more years of a situation where access is almost entirely determined by insurance status and income.

The Next-Gen Wave — Just When You Thought You Understood the Market

Here’s the complication: just as semaglutide generics are finally reaching patients in Canada, India, China, and Brazil, the pharmaceutical industry is already rolling out the next generation of these drugs — and some of them are genuinely better.

Tirzepatide (Mounjaro/Zepbound), Eli Lilly’s dual GLP-1 + GIP agonist, is already growing faster than semaglutide in the US. Clinical trial data shows meaningfully superior weight loss. It’s a different molecule with a different manufacturing supply chain, and its patents won’t expire until 2031+ either.

But the one that could genuinely reshape everything is orforglipron — Eli Lilly’s oral, non-peptide, small molecule GLP-1 agonist. Phase 3 trials for obesity completed in August 2025. If FDA approval comes in 2026 as expected, orforglipron would be the first oral weight loss drug with GLP-1-class efficacy. No injection. No cold chain. No DDC device complexity. Just a daily pill.

Novo Nordisk isn’t sitting still either. CagriSema — a combination of semaglutide and cagrilintide (an amylin analogue) — had Phase 3 topline data in March 2025, and Amycretin, a single-molecule GLP-1/amylin dual, is entering Phase 3 in 2026. Novo is essentially trying to innovate past its own product before the generics get there — upgrade the semaglutide franchise into something new enough that a generic of the old thing becomes less relevant.

One molecule worth watching carefully: Retatrutide. In clinical trials, the triple agonist (GLP-1 + GIP + glucagon) produced the highest weight loss ever recorded in a drug trial. We are potentially talking about pharmacological tools that can produce 25–30% body weight reduction. The implications for obesity medicine — and for the CDMO supply chain required to manufacture these molecules — are profound.

What This All Adds Up To

If you zoom out, the GLP-1 story is fundamentally about the collision of three forces that rarely intersect at the same moment in pharma history.

First, a medical need of almost incomprehensible scale. Over one billion people globally meet the clinical definition of obesity. Around 537 million have type 2 diabetes. The drugs addressing these conditions work better than almost anything medicine has previously produced. The demand isn’t manufactured — it’s real, it’s massive, and it’s in every country.

Second, a patent cliff that’s already happening in four major markets simultaneously, with the biggest prizes (US, EU) waiting just five years behind. The generic companies, the CDMOs, the API suppliers — they’ve all been building to this moment for years. The infrastructure race is as intense as the drug race.

Third, an innovation pipeline that could make today’s drugs look like the first generation of smartphones — functional but primitive compared to what’s coming. Orforglipron alone could add hundreds of millions of addressable patients globally. Retatrutide could make the current obesity treatment conversation look like we were practicing. And nobody has yet figured out what happens when genuinely effective obesity drugs become genuinely affordable at scale.

For investors, this means the GLP-1 space isn’t a trade on a single company anymore — it’s an ecosystem, with value distributed across innovators, generic challengers, CDMO manufacturers, API suppliers, device makers, and the distribution infrastructure connecting all of them. The companies building the picks-and-shovels of this gold rush may prove as important as the miners.

For patients — especially the hundreds of millions priced out of these drugs today — the next five years represent the best realistic chance in medical history that transformative obesity and diabetes treatment becomes accessible at the population level.

The patent wall is cracking. The generics are ready. The next-gen pipeline is extraordinary. And the world has never needed these drugs more.

Watch this space closely.

The Indian AC market presents a rare combination of structural growth drivers:

(1) 12x penetration arbitrage vs developed markets creating long runway,

(2) Climate crisis extending AC season by 40%,

(3) Government PLI scheme enabling margin expansion for component manufacturers,

(4) Premiumisation driving ASP growth 3-5% above volume growth.

Value capture is bifurcated: Brand OEMs face margin compression from competition while ODM/component players (Amber Enterprises, PG Electroplast) capture disproportionate value through backward integration and PLI incentives.

AC PENETRATION: INDIA VS GLOBAL MARKETS

COMPETITIVE LANDSCAPE & MARKET SHARE ANALYSIS

VALUE CHAIN ANALYSIS & BILL OF MATERIALS

TECHNOLOGY EVOLUTION & ESG COMPLIANCE

INVESTMENT RISKS

CONCLUSION: The Indian AC market offers a 20-year structural growth opportunity with 14-17% CAGR driven by climate crisis, urbanization, and 12x penetration arbitrage. However, value capture is BIFURCATED:

(1) Brand OEMs (Voltas, LG, Daikin) will grow revenues but face margin compression from intense competition. Premiumisation favors tech leaders (Daikin) over volume leaders (Voltas).

(2) ODM/Component manufacturers (Amber Enterprises, PG Electroplast) will capture disproportionate value through PLI incentives (4-6% EBITDA boost), backward integration (300-400 bps margin advantage), and multi-brand scale aggregation.

Why the Market Says “Yes” to a Sprint

• Inventory as an Asset: The record 2.5 million units in channel inventory (left over from the 2025 rain) is now seen as a hedge against potential early-summer supply shocks in 2026.

• The GST Catalyst: The cut from 28% to 18% is the “secret sauce.” It effectively offsets the ₹500–2,500 price hike caused by the new Jan 2026 BEE Star Labeling norms. Consumers get better tech at the same (or lower) price point.

• 12-Month Demand: The aggressive push into “Hot & Cold” ACs by players like Blue Star and Panasonic is finally breaking the industry’s addiction to the April-June window, making it a more resilient, year-round play.

Ping me for a detailed financial model..

Share piyush

The investment case for Kaynes Technology is rooted in a fundamental structural shift in the Indian Electronics System Design and Manufacturing (ESDM) sector: the transition from low-value assembly to high-value component and product ownership.

1.1 Total Addressable Market (TAM) Architecture

The Indian ESDM sector is currently undergoing a non-linear expansion, driven by domestic consumption and import substitution. The market is projected to triple over the next five years, creating a massive vacuum for domestic players with technical capabilities.

Table 1: India ESDM and Component Market Projections (USD Billion)

ELECTRONICS PRODUCTION & SEGMENT BREAKDOWN

Table 2: India PCB Market (USD Billion)

TAM is big enough to accommodate multiple players. Next question – Which player with certain traits can capture the market?

1.2 The Value Capture Curve

The global electronics industry is governed by the “smile curve” of value addition, where the highest value is captured at the ends of the chain (Design/IP and Services/Branding), while the lowest value resides in the middle (Assembly/Manufacturing).

SMARTPHONE COST BREAKDOWN

• Global Benchmark: OEM/Product Owners (e.g., Apple, Cisco) capture ~50-75% of the industry profit pool.

• Service Providers: Pure-play EMS companies typically capture only ~2-3% of the profit pool.

Strategic Imperative:

1. For India to create domestic OEMs, a robust Tier-1 manufacturing ecosystem must first exist to support them. Kaynes is positioning itself not just as a service provider (EMS) but as an integrated enabler (ODM + Components) to capture a larger share of this value. By moving into OSAT (Outsourced Semiconductor Assembly and Test) and bare PCB manufacturing, Kaynes attempts to migrate from the “2% bucket” toward the “15-20% bucket” of component value addition.

2. India at some point of time will look at Strategic Autonomy. Building this ecosystem is critical for India.

2.1 The “Chip-to-Product” Ecosystem Feedback Loop

By co-locating near the Tata Fab, Kaynes secures a supply chain advantage. The ecosystem allows for “Design in India” where domestic product companies (the future OEMs) can source chips, boards, and assembly from a single geography, reducing time-to-market. This integration acts as a moat against purely import-dependent competitors.

2. Competitive Benchmarking: Margin Profiles and Business Models

Kaynes Technology operates a High-Mix, Low-Volume (HMLV) model, which is fundamentally different from the Low-Mix, High-Volume (LMHV) model of peers like Dixon Technologies or Foxconn. This distinction is visible in the EBITDA margin profile.

To understand the end-state of Kaynes’ backward integration strategy (into OSAT), we must benchmark against global leaders like ASE Technology Holding (Taiwan), the world’s largest OSAT player.

Table 3: Technical Capability Benchmarking – OSAT

Table 4: Technical Capability Benchmarking – OSAT

Table 5: Operational Benchmarking – EMS vs. OSAT vs. Kaynes

Source: Derived from ASE Financials, Jabil Financials, Kaynes Financials. * GM includes all the cost incl. COGS, Labour, R&D & Depreciation; hence not comparable

Analysis:

• Traditional EMS operates on thin margins, relying on massive scale.

• By focusing on mature technologies like Quad Flat No-leads (QFN) and Ball Grid Array (BGA), Kaynes minimizes execution risk. Yields on these mature packages are easier to stabilize compared to the exotic yield challenges of 2.5D packaging.

• ASE Technology (Benchmark): Consistently delivers EBITDA margins between 16% and 18%. This validates Kaynes’ strategy: as the revenue mix shifts toward OSAT and HDI PCBs (starting FY27/28), the blended EBITDA margin ceiling elevates from the current ~15% to potentially ~18%.

• Kaynes currently commands a “scarcity premium” because it is the only Indian player aggressively constructing this high-margin vertical stack.

The player who will be on the top of the game w.r.t technology and scale will capture the market. Next question – If Kaynes can be 1 of the front runners, then do we have a valuation comfort?

3. Financial & Operational Outlook (FY25-FY28)

Kaynes is currently in a heavy investment phase (Capex > Cash Flow), typical of infrastructure companies before they hit operating leverage. The financials below factor in a strategic reduction in the smart meter business to correct working capital bloat.

3.1 Forecasted Financial Trajectory

Table 6: Consolidated Financial Estimates (INR Million)

3.2 Segmental Shift: De-risking the Revenue Mix

The company is actively reducing its dependence on the Industrial (Smart Meter) segment to improve balance sheet health (Working Capital Days). Growth is being redirected toward high-precision verticals.

Table 7: Revenue Contribution Shift (Strategic Re-alignment)

4. Strategic Risks & Governance Context

Recent scrutiny regarding Related Party Transactions (RPT) and accounting standards represents the friction of transitioning from a family-run mid-cap to an institutionally owned large-cap.

• Accounting: The lack of immediate RPT reconciliation in standalone vs. consolidated books indicates weak internal financial controls (IFC), though not necessarily financial misstatement, as consolidated numbers eliminate these entries.

• Working Capital: The spike in WC days to ~116 in 1HFY26 was driven by the specific “Service Provider” model in smart meters. The pivot to a “Device Supplier” model (selling hardware only) is a corrective measure to bring WC days back to <90.

• Intangible Assets Capitalization: Kotak flagged the capitalization of ₹1.8 billion (6.5% of revenue) as “technical know-how.” Management clarified that this figure includes ₹1.15 billion in customer contracts acquired through the Iskraemeco deal.

  • Insight: In the EMS/ODM world, long-term, sticky customer contracts are the primary asset. Capitalizing them is an aggressive accounting stance but is defensible under Ind AS 103 (Business Combinations) if they represent distinct, measurable economic value. This is a common practice in acquisitions to reflect the value of the acquired order book.

5. Valuation Framework: Sum-of-the-Parts (SOTP)

Given the disparate nature of the core EMS business (cash cow) and the new semiconductor initiatives (high growth/capex), a SOTP approach is required to accurately value the entity.

Table 8: SOTP Valuation Model (FY27E Basis)

Kaynes has the valuation comfort at INR 22,000 Market Cap. But if any of the major risk plays out (delay in CAPEX or delay in lines coming on stream, stock will continue to suffer).

Governance risk and WC risk are not major risk for the company in longer term

6. Market Sentiment and Investor Psychology

The price action in Kaynes Technology is a quintessential case study in behavioral finance, illustrating the interplay between FOMO (Fear Of Missing Out), Narrative Fallacy, and Loss Aversion within the Indian equity market.

6.1 The FOMO Phase

• The Narrative: “India is the next China,” and “Semiconductors are the new oil.” Kaynes became the poster child for this narrative.

6.2 The Fear Phase

• Current Sentiment: The market is currently in a “show me” phase. The “governance discount” has compressed multiples, but the “growth premium” remains latent. Investors are waiting for tangible execution proof points—specifically, the normalization of working capital cycles and the first commercial chip shipments from Sanand—before re-rating the stock.

7. Conclusions & Strategic Outlook

Kaynes Technology is currently an “infrastructure play” on the Indian electronics ecosystem. The near-term volatility regarding WC and governance is the price of admission for a company aggressively scaling capital-intensive verticals (OSAT/PCB) ahead of the market. The thesis relies on the successful execution of the Sanand (OSAT) and Mysore (PCB) facilities, which will structurally alter the company’s margin profile and align it with global technology majors rather than contract manufacturers

Key Conclusions:

1. The Pivot is Real but Risky: The move into OSAT and HDI PCB is strategically necessary to escape the “low-value trap” of EMS.

2. Technical Credibility via Partnership: While not competing with TSMC/ASE on the bleeding edge (3nm/AI logic), Kaynes has carved a viable niche in power modules and legacy packaging (QFN/BGA). Its partnerships with AOI Electronics (for PLP) and Globetronics (for sensors) provide the necessary technical scaffolding to execute this “trailing edge” strategy effectively.

3. Governance as a Rite of Passage: The company must transition from “family-run” opacity to “institutional” transparency to retain investor trust.

4. Strategic Entry Point: For the long-term investor, the governance-induced correction offers an entry point into a scarcity asset.

The views expressed in this research report accurately reflect personal views about the subject securities and issuers. No part of the analyst’s compensation was, is, or will be directly or indirectly related to the specific recommendations or views expressed in this report.

Annexure

Case Study 1: Luxshare Precision (The “Acquisition & Integrate” Model)

• Origin: Luxshare started as a simple connector manufacturer (PC cables) in 2004, supplying to Foxconn.

• The Pivot: The company aggressively acquired entities to move up the value chain—acquiring Lanto (cables), Suin (audio), and eventually Wistron’s iPhone assembly plant.

• Governance & Skepticism: Throughout 2011-2015, analysts expressed deep skepticism regarding Luxshare’s ability to manage cash flows during its heavy capital expenditure phase. The company faced scrutiny over negative free cash flow (FCF), rising debt, and aggressive accounting as it integrated acquisitions.

• The Outcome: Luxshare proved that negative FCF is often a feature, not a bug, of rapid infrastructure scaling in the manufacturing sector. By FY2024, Luxshare had become a dominant player in the Apple ecosystem, challenging Foxconn.

If you’ve spent even a few years in the market, you’ll notice one pattern repeat endlessly:

The market over-reacts to good news,
and it over-reacts to bad news even more.

There’s a strange habit of the market :

Step 1: Price a company as if the next five years will go perfectly.
Step 2: Get disappointed the moment life behaves like… life.

This loop plays out again and again — across industries, across geographies, across cycles.

Management lays out a plan, gives a growth roadmap, maybe adds a few ambitious targets — and suddenly the stock price starts behaving as if everything from here will be flawless.

And the moment the company hits a rough patch — even a short one — the same stock gets punished far more than the situation deserves.

It’s not new.
It’s not surprising.
But it’s still fascinating how quickly the market forgets its own optimism.

NVIDIA: A Reminder of How High the Bar Has Become

NVIDIA didn’t post “bad” numbers.
They posted “not-as-insanely-good-as-expected” numbers.

And that was enough for tens of billions of dollars to get wiped out within minutes.

Anyone who has been investing long enough has seen this before.
It’s not about the earnings — it’s about expectations.
And expectations today sit at extremes.

In India, It’s the Same Story — Just a Different Set of Names

Companies like:

• PG Electroplast / EPACK Durables / Kaynes Technology / TRIL / Zaggle and list goes on….…are all operating in strong, long-term themes.

• Strong theme → Market prices perfect execution.

• 1–2 weak quarters → Market acts betrayed.

• Stock gets derated far more than the actual risk.

These are businesses you expect to have good years and bad years, but overall, the direction is clear.

Yet the moment they hit 2–3 weak quarters — which is absolutely normal for any growing business — the stocks get derated sharply.

Not because the story broke.
Not because demand disappeared.
Simply because the market wanted a smooth line.
And real growth is anything but smooth.

The Market Understands Cyclicals Better Than It Understands Secular Growth

Cyclicals are easy for the market to process:

• The cycle turns

• Volumes drop

• Prices fall

• Everyone expects pain

But with secular growth companies, the market behaves strangely.

Even when the long-term runway is obvious, it still prices the business as if growth will be perfectly predictable.
And when a quarter or two disappoint?

The reaction is far harsher than necessary.

This is the gap between expectations and execution.
This is where the mispricing happens.

And honestly — this is where patient investors earn their returns.

The Truth: A Few Rough Quarters Don’t Break a Secular Story

Anyone who has actually built or run a business knows this:

Growth is never linear.
Margins don’t expand in a straight line.
Demand doesn’t behave like a forecast model.

But the market still prices companies as if they will execute with machine-like precision.

So when a company enters a 2–4 quarter rough patch — which is almost a rite of passage — the market behaves like the entire thesis has collapsed.

And yet, most of these businesses bounce back quickly.
Because the underlying structural tailwind didn’t disappear.

Cyclicals?
You genuinely don’t know how long the pain will last.

Secular names?
Usually, the recovery is visible much sooner.

This Volatility — As Irritating As It Feels — Is Actually a Gift

When you’re holding a stock that is down 30–50%, it never feels like a gift.
Let’s be honest about that.

But with time, you realize:

Volatility is the only reason good companies become available at fair prices.

If the market stayed rational, patient, and calm —
you’d never get an opportunity to accumulate anything quality at a discount.

This impatience is what creates the gap between price and value.
And that gap is where long-term returns come from.

A Video You Might Like

A Simple, Human Takeaway

The market doesn’t have to become patient.
It won’t.
It never has.

But you can be.

And that alone — not IQ, not spreadsheets, not models —
is often the difference between average returns and great returns.

If the story is intact, if the runway is long, if the management is executing over the medium term…

…then surviving a few rough quarters is simply part of the journey.

Price reacts fast.
Businesses recover slowly.
And investors who understand this difference — they win.

In the global solar photovoltaic (PV) market, cycles are typically short (2-3 years) due to rapid capacity additions not constrained by pricing discipline. However, the Indian government has created a “walled garden” through progressive tariff and non-tariff barriers, effectively suspending global demand-supply dynamics. This is expected to elongate the solar PV cell upcycle in India, possibly lasting until FY29-30E, longer than the consensus view.

The longevity of the cycle is secured by inherent and forced structural constraints that cause a delay and reduction in actual output compared to theoretical capacity.

The Hidden Constraints on Domestic Supply

Indian manufacturers face substantial technical hurdles that stretch the timeline for supply realization:

-------------------------------------------------------------------------------
Section 2: The Regulatory Fortress: DCR and ALMM Guarantee Demand

The elongated supply cycle is met with a fiercely protected domestic demand guaranteed by policies like the Approved List of Models and Manufacturers (ALMM) and the Basic Customs Duty (BCD). These measures act as non-tariff barriers, ensuring high-margin demand remains captive domestically. As of Fiscal 2025, prices of domestic modules assembled using domestic cells (DCR) were approximately 60% higher than prices of domestic modules assembled using imported cells (non-DCR).

Compulsory Cell Mandates: A Timeline for Domestic Dominance

Guaranteed Domestic Cell Demand (Fiscals 2026 – 2030)

India’s total solar PV cell manufacturing capacity increased rapidly, from approximately 3.2 GW in March 2022 to an estimated 23 GW by March 2025. India’s total module demand is projected to average 38 GW to 43 GW annually between Fiscals 2026 and 2030. This demand is segmented across key protected verticals:

--------------------------------------------------------------------------------

Section 3: The Next Great Leap: Vertical Integration into Wafers and Ingots

As the cell supply bottleneck is addressed by ALMM List II, the constraint is expected to shift upstream to wafers and ingots. Raw material costs (primarily imported wafers) constitute 80% to 85% of the operating costs for Indian solar cell manufacturers.

Regulation to Force Integration: The final layer of protection is the proposed ALMM List III for wafers and ingots.

• Implementation Date: ALMM List III is slated to take effect on June 1, 2028, conditional upon a minimum of three separate domestic wafer/ingot manufacturing facilities collectively achieving at least 15 GW of annual production capacity.

• Capital Imperative: The cumulative capital expenditure (capex) required to establish the necessary ∼75 GW ingot and wafer capacity by FY30E could entail a cumulative outlay of approximately Rs500 billion.

• Capacity Targets: Large-scale wafer and polysilicon facilities are expected to be established, targeting capacities of 35 GW to 45 GW and 15 GW to 25 GW, respectively, by Fiscal 2030.

--------------------------------------------------------------------------------

Section 4: Identifying the Winners: Strong Balance Sheets and High Tech

The structural environment of protected demand, high capital investment, and technological complexity mandates that profitable manufacturers excel in three areas: financial discipline, integration, and high technology.

1. Financial Discipline Ensures Downcycle Margins

Due to higher plant capital costs (∼25-30% higher than in China, partially due to costlier European equipment) and high Indian interest costs (estimated around 10–11%), Indian manufacturers carry a larger debt burden than their Chinese counterparts.

• Downcycle Margin Floor: This leverage forces pricing discipline, ensuring that the estimated downcycle EBITDA margin floor for Indian manufacturers will be at least 18% to 20% (approximately Rs3.4/Wp), significantly higher than the sub-10% margins of subsidized Chinese competitors.

• Early Mover Advantage: Companies that commissioned cell capacity early, such as Premier Energies (2011), gained 14 years of experience. This operational history allowed them to capture high-margin demand earlier and write off assets ahead of peers. Premier had depreciated approximately 39.8% of its Gross Fixed Assets (GFA) by FY25, substantially higher than the peers’ median (18.9% for Waaree, 9.8% for Emmvee).

2. High Technology is Non-Negotiable (TOPCon)

The transition to high-efficiency N-type cells, primarily TOPCon (Tunnel Oxide Passivated Contact), is critical.

• Market Shift: The share of cell manufacturing capacity configured for or fungible with TOPCon reached an estimated 60% to 65% in Fiscal 2025. By Fiscal 2030, nearly 80% of India’s cell manufacturing capacity is expected to be TOPCon-capable.

• Early Adoption: Companies like Emmvee are noted as one of the first in India to adopt TOPCon cell manufacturing, leveraging this advantage to meet demand arising from ALMM II. Emmvee’s 2.94 GW facility is recognized as one of India’s largest TOPCon cell manufacturing facilities in terms of installed capacity as of May 31, 2025.

3. Integration Drives Superior Returns

Vertical integration provides superior control, efficiency, and margin capture. Operating margins for pure-play cell manufacturers (∼19% in Fiscal 2024) were materially higher than those for module manufacturers (∼13%).

Sources - Investor presentations, DRHPs and IIFL report “India Energy Transition” published on 24th Oct 2025

Subscribe now

Share

That narrative misses the fundamental transformation: Each underlying business is demonstrating genuine organic growth that would justify premium valuations as standalone entities.

The insight: You're not just playing corporate action catalysts - you're buying into India's growing financial services segments at distressed to average prices.

Question 1 - Is is worth even thinking that turnaround will happen ?

• Management has turnaround with Nuvama earlier and have clear roadmap now ahead

• EAAA IPO being the next leg of monetization with Mutual Fund following.

  Conclusion - Management can walk the talk going forward

Question 2 - Does the investment offer a sufficient Margin of Safety, and is there an unusually high probability of outsized upside returns?

Current SOTP Value: ₹161-180 bn vs Market Cap ~₹110 bn (45-55% discount)

• FY27E SOTP Value: ₹217-240 bn (100% upside potential). At 50% holding discount persisting, it might be ~₹110 bn - ₹120 bn

Complete Investment Framework: Growth + Value Unlocking Analysis

Master SOTP Valuation: FY25 Base vs FY27E Projections

So, ultimately, there isn’t a downside to the stock price from the current levels and probably higher chances of returning 20-25% kind of upside in 9-12 months time.

Question 3 - What is the Strategic outlook for FY26 and is the company on track to achieve that?

Company seems to be on tract to achieve all of these.

Conclusion - A High-Conviction Investment in India's Growing Financial Services at an Attractive Valuation, Driven by Proven Management and Strategic Organic Growth.

Sources - SOIC Video &

Keynote Sees 95% Upside In Edelweiss Financial Services Special

1.19MB ∙ PDF file

Download

Download

Imagine this: a battery discarded in Ghana, a shredded plastic drum in Jaipur, and an aluminum scrap yard in Sri Lanka — all feeding into a single global engine that transforms waste into value. This isn’t a futuristic concept. It’s happening right now, and Gravita India Limited is at the center of it.

What started in 1992 as a single-unit lead recycler in Rajasthan has evolved into one of Asia’s leading integrated recycling companies, with a presence across four continents and a robust portfolio spanning lead, aluminum, plastic, and rubber recycling, alongside turnkey solutions.

Let’s trace how Gravita became the poster child for India’s circular economy — one that doesn’t just recycle material, but recycles value, vision, and trust.

A Vision Built on Purpose — And Execution

Founded by first-generation entrepreneur Mr. Rajat Agrawal, Gravita’s mission was always more than just business — it was about giving back more than we take. The company’s Vision 2028 is ambitious yet focused:

• 25%+ volume CAGR

• 35%+ profitability growth

• 30%+ renewable energy usage

• 50%+ revenue from value-added products

• 30%+ business from non-lead verticals

• 10%+ reduction in energy intensity

Gravita sees itself not just as a recycler — but as a circular economy leader, offering solutions that reduce dependency on virgin resources and carbon-heavy industrial processes.

A Multi-Metal, Multi-Market Recycling Powerhouse

Gravita is one of India’s few fully integrated recyclers across multiple verticals. Its product and geographic diversification reduces concentration risk and future-proofs the business.

Core Recycling Verticals:

From Chittoor to Chile, Mundra to Mozambique, Gravita’s 11 manufacturing units and multi-continent reach make it a truly global materials recovery player.

Turnkey Projects: India’s Export of Circular Expertise

Beyond recycling, Gravita is also a global supplier of turnkey recycling solutions. Its in-house expertise has led to 70+ projects delivered in nations like Qatar, Chile, Saudi Arabia, and Poland.

These are not just equipment sales — they’re fully integrated plant setups for aluminium, plastic, rubber, and lead recycling, complete with design, installation, commissioning, and training.

This business vertical:

• De-risks the core operations

• Adds a high-margin consulting revenue stream

• Strengthens Gravita’s brand as a global circular economy thought leader.

Hedged for Stability, Driven for Growth

One of Gravita’s core strengths lies in its risk management DNA:

• Back-to-back hedging for all scrap procurement and LME pricing

• Natural hedging for exports (50%+ revenue in USD)

• Strict working capital control — reduced net working capital cycle to 83 days

• Stable access to low-cost credit from 10+ lenders due to high ICRA ratings