Why Moore’s Law Is Coming for Indian Healthcare

Healthcare costs do not fall because time passes. They fall when volume compounds, competition arrives, scarce expertise becomes distributable and expensive assets stop sitting idle. India is beginning to assemble all four.

At the beginning of this century, keeping a person alive with HIV cost roughly $10,000 to $15,000 a year. A preferred first-line regimen can now be procured for under $45 annually. That is a decline of more than two hundred-fold, and Indian generic manufacturers were central to making it possible.

HIV was not an isolated miracle. A Hepatitis C cure launched at $84,000; Egypt drove the price below $100 as it licensed local manufacturers and multiplied suppliers. An imported intraocular lens cost more than $100 in the early 1990s; Aurolab in Madurai now produces one for around three dollars. The cost of sequencing a human genome fell from tens of millions of dollars to a few hundred.

These examples come from different diseases, technologies and health systems, but they share a common shape. Once a unit of care becomes standardised, once cumulative volume begins compounding and once suppliers are forced to learn, its cost can fall at a rate that appears impossible at the beginning of the curve.

This memo is about that machine: why it has worked in some parts of healthcare, why it has stalled in others, what is now changing, and which parts of Indian healthcare are likely to reorganise first when the curve begins to move.

Why costs fall

Nobody describes the collapse in HIV-treatment costs as Moore’s Law. Strictly speaking, they should not. Gordon Moore’s 1965 paper was about the number of transistors that could fit on an integrated circuit, not the price of a medicine or a medical procedure.

The more useful law was described almost thirty years earlier by an aeronautical engineer named T. P. Wright. In 1936, Wright observed that every time cumulative aircraft production doubled, the labour required to make each aircraft fell by roughly 20%. The reduction was not tied to the calendar. It occurred when the number of units ever produced crossed another doubling.

Later research tested Wright’s formulation against several competing forecasting laws across dozens of technologies and found that it was generally the most reliable. The underlying idea is simple: cost falls on a counter, not a clock.

The first unit carries the cost of every mistake, every manual process, every underused machine, every untrained worker and every uncertain supplier relationship. The millionth unit does not. By then, the process has been standardised, waste has been removed, workers have become faster, machines are fuller, suppliers compete and fixed costs have been spread over enormous volume.

This is why the unit matters so much. “Healthcare” cannot be pushed down a learning curve because healthcare is not one thing. A tablet can be. A cataract operation can be. A chest X-ray interpretation, a diagnostic panel, a knee replacement or a diabetes-care pathway can be.

Before cost can compound downward, the unit must become visible. It must then become repeatable. Only after that can its counter advance.

At Wright’s canonical learning rate, a 20% decline with every doubling produces dramatic results when the doublings occur frequently. To reduce cost by roughly 90% over a decade, cumulative volume in the standardised unit needs to double approximately once a year.

That sounds implausible until one looks outside healthcare. India has already built two national-scale volume engines in the past decade. Mobile-data prices fell from ₹269 per GB in March 2014 to ₹8.31 by June 2024. UPI grew from two crore transactions in FY17 to more than 24,000 crore in FY26, while the underlying cost of processing a merchant payment remained extraordinarily low.

Neither transformation depended on a single scientific breakthrough. In both cases, a standardised rail attracted volume; volume justified investment; investment improved reliability and reduced unit cost; lower cost brought in more users; and the counter accelerated again.

Healthcare has generally failed to create the same loop. It has often invested in better technologies without building the operating system required to push enough repeated units through them.

Aravind Eye Care illustrates the distinction. It remains one of the most efficient surgical systems in the world, yet its cost per cataract operation rose from roughly ₹15,000 in 2012 to ₹23,000 in 2023. Narayana Health’s revenue per occupied bed has been rising. The GeneXpert tuberculosis cartridge took eleven years to fall from $16.98 to $7.97. Some biologic medicines declined in price when the first biosimilar arrived and then stopped moving, even as more products entered the market.

These examples do not prove that healthcare is exempt from learning curves. They show what happens when the conditions required to sustain the curve disappear.

GeneXpert spent years without a serious second supplier. Aravind reached the frontier of what its existing operating model could achieve and had nowhere to send enough additional volume. In each case, the counter slowed, so the cost stopped falling.

Competition is particularly important. The United States Food and Drug Administration studied prices across 181 medicines and found a striking relationship between the number of generic manufacturers and the decline from the branded price. One generic entrant reduced the price by approximately 39%. Two reduced it by 54%. Four reduced it by 79%. Six or more reduced it by upward of 95%.

The first competitor breaks the monopoly, but it does not necessarily create a competitive market. The structure changes more fundamentally when the fourth or sixth entrant arrives and each supplier knows that another credible manufacturer can take meaningful volume away from it.

Competition, however, is not sufficient by itself. A market can admit six suppliers and still fail if none has enough predictable demand to invest, learn and survive. Competition without volume produces fragile companies. Volume without competition protects incumbent margins.

A durable cost curve needs both. It needs a buyer or aggregator capable of concentrating enough demand to make the next doubling plausible, and it needs enough credible suppliers to ensure that the benefit of learning reaches the patient rather than remaining inside the manufacturer’s margin.

The missing ingredient in healthcare has therefore rarely been technology alone. More often, it has been a system capable of manufacturing the doublings.

What makes the curve move

A learning curve does not run on optimism. Someone has to organise the volume, competition has to be credible, scarce expertise has to become distributable and expensive assets have to be used far more intensively than they are today.

China turned the first requirement into procurement policy. Under its volume-based procurement programme, suppliers receive visibility into how much participating public hospitals are expected to purchase before they submit a price. The exact allocation depends on the number of winning manufacturers, but the principle is consistent: a substantial share of prior annual demand is contractually attached to the tender.

This is different from the way most healthcare companies forecast demand. A forecast tells a manufacturer what the buyer hopes will happen. A procurement commitment changes what the manufacturer can afford to do before it happens. It allows the supplier to plan capacity, negotiate raw-material costs, remove distribution uncertainty and accept a smaller margin per unit in exchange for much larger and more predictable volume.

The results have been dramatic. Coronary-stent prices fell by more than 90%, from roughly ¥10,242 to ¥676, while procurement volume increased. Hip and knee implants experienced similarly large reductions. Across successive procurement rounds, hundreds of medicines were included and some individual drug prices fell by more than 90%.

China also demonstrated the risks of running this mechanism without sufficient safeguards. Physicians raised concerns about the quality of some ultra-low-price products. Research suggested that part of the savings could leak into substitute prescribing or be offset by higher utilisation. Beijing subsequently introduced stronger price floors and cost-justification requirements for unusually low bids.

The lesson is not that the mechanism failed. It is that a volume guarantee needs a quality floor beside it. The guarantee is the engine; quality assurance and economically credible bids are the brake. A system designed around only one of the two will eventually fail.

Egypt reached the same underlying mechanism through a different route. It refused a patent for sofosbuvir, licensed multiple domestic manufacturers and allowed those suppliers to compete. The cost of a Hepatitis C cure moved from an international list price of $84,000 to a negotiated price of $900, then below $300 as local production began and eventually below $100 as the supplier base expanded.

Egypt then organised the demand required to sustain those economics. It screened more than 60 million people through thousands of fixed and mobile sites and treated millions of patients who tested positive. In 2023, it became the first country to receive the World Health Organization’s gold-tier validation on the path to eliminating Hepatitis C.

The achievement matters not only because of the price decline or the health outcome. It demonstrates that a country can deliberately create a healthcare cost curve by combining procurement, standardisation, competition and national-scale demand.

Brazil’s Farmácia Popular programme and Japan’s national fee schedule offer variations on the same idea. Brazil used a large private-pharmacy network to expand utilisation while lowering government expenditure per hypertension and diabetes treatment. Japan repeatedly reset reimbursement prices while allowing imaging capacity to proliferate, creating one of the highest densities of MRI machines in the world alongside some of the lowest prices in a developed market.

Different political systems arrived at the same economic truth: price is downstream of a promise. Before a supplier can reliably lower the unit price, someone has to make the volume credible.

The second major shift is that the scarce human being no longer has to stand inside every unit of care.

Indian healthcare’s binding constraint has never been only the number of machines or the availability of medicines. It is the uneven distribution of qualified people. India has roughly one radiologist per 100,000 people, compared with eight to ten across much of the OECD. Rural community health centres have 913 surgeons in post against 5,491 required, an 83% shortfall that has barely changed despite the expansion of medical education.

Training more clinicians remains essential, but training alone will not resolve this distribution problem in a relevant timeframe. Specialists do not distribute evenly simply because more of them graduate. Doctors distribute more like real estate: they remain concentrated where the economic and professional conditions are attractive. Software distributes more like electricity.

That difference is beginning to change the cost structure of screening and routine interpretation. In Sweden’s MASAI trial, more than 100,000 women were randomised between AI-supported and conventional mammography screening. AI support reduced radiologists’ reading workload by approximately 44% while improving sensitivity and reducing interval cancers. Large-scale chest X-ray research has similarly suggested that a substantial share of routine reporting can be automated while reserving clinician attention for abnormal or uncertain cases.

Regulators have already accepted this principle in specific, bounded domains. The World Health Organization permits computer-aided detection as an alternative to human interpretation of chest X-rays for tuberculosis screening in people aged 15 and older. The United States has approved autonomous diabetic-retinopathy screening. The United Kingdom permits limited AI-led decisions in parts of the skin-cancer pathway.

This does not mean that software is about to replace doctors across medicine. It means something more practical and more economically important: a scarce specialist no longer needs to review every routine case for the system to deliver a safe service.

The radiologist can focus on ambiguous images rather than every normal scan. The ophthalmologist can focus on patients requiring treatment rather than screening every retina. The pharmacist can spend more time resolving complex interactions and adherence failures rather than repeatedly delivering standard instructions.

Once scarce human time is removed from every unit, the cost structure changes. An AI chest X-ray interpretation can cost a fraction of a conventional radiologist read. Retinopathy-screening systems have demonstrated lower per-patient costs than trained human graders. Tuberculosis-screening software has already declined sharply in price as throughput expanded.

The intelligence inside a screening programme can now become cheaper on a semiconductor cost curve rather than more expensive on a wage curve. Inference remains only a small part of the cost of delivered care, and the frontier models will continue to command a premium. But the direction is new. The cognitive component of a repeatable healthcare task can now improve and deflate independently of the local supply of specialists.

The third opportunity is hidden inside assets the health system already owns.

Intuitive Surgical reported approximately 3.15 million da Vinci procedures in 2025 across a year-end installed base of 11,106 systems. As a rough global average, that is little more than one procedure per system per day.

An airline could not survive by flying an aircraft for one hour and parking it for the remaining twenty-three. Aviation economics depend on keeping expensive metal in the air. Hospitals routinely purchase operating theatres, scanners and surgical robots, use them for a small portion of the day and treat the idle time as an unavoidable feature of medicine.

Much of it is not unavoidable. English NHS surgical hubs increased high-volume, low-complexity surgery substantially by separating elective work from emergency care. Narayana’s surgeons have historically completed several times the annual procedure volume of many American counterparts through sequential scheduling, six-day operating weeks and reduced downtime between cases.

The immediate opportunity in surgical technology is therefore not autonomous surgery. It is utilisation. A robot or operating theatre running multiple predictable shifts can spread its fixed cost across several times as many procedures without waiting for another scientific breakthrough.

Autonomy may eventually extend the curve, but it remains early. Experimental systems have demonstrated portions of surgical procedures in animals or controlled laboratory settings. The overwhelming majority of cleared surgical robots operate at low levels of autonomy, and fully autonomous surgery has not yet been tested in humans.

At the same time, robotic surgery currently tends to cost more than conventional laparoscopic surgery. Consumables remain expensive, and a dominant razor-and-blade model has had little reason to reduce its own margin.

The mechanism that is more likely to make robotic surgery affordable in the near term is familiar: competition and utilisation. New global entrants and lower-cost Indian systems are beginning to challenge the incumbent on both capital cost and consumables. As more suppliers arrive and installed machines are used more intensively, the counter can begin to move.

This is the central point of the second act. The curve is not unlocked by technology in isolation. It moves when volume becomes certain, competition becomes credible, scarce expertise becomes distributable and installed assets are pushed through far more repeated units.

What happens when the curve moves

The consequence is an inversion that healthcare has historically struggled to believe.

Healthcare does not become affordable because an operator agrees to accept permanently lower profits. It becomes affordable when the high-throughput model becomes the economically stronger business.

The cost leader attracts more demand. More demand fills capacity. Fuller capacity reduces the cost per unit. Lower cost creates room to reduce price again. The resulting volume produces more purchasing power, more process data, more repetition and faster organisational learning.

The curve begins to feed itself. At that point, low price is no longer a charitable concession. It becomes the visible output of a better operating system.

This is why the cheaper operator can eventually become the more profitable one. Its advantage is not simply that it charges less. It is that it runs the fastest counter.

Pharmacy is likely to reorganise first because India’s problem was never primarily the cost of manufacturing medicine. India already manufactures medicines for the world. The opportunity lies between the factory and the patient.

The trade-markup stack persists because demand remains fragmented across hundreds of thousands of pharmacies purchasing in small lots. No individual store can create enough volume to negotiate meaningfully, standardise assortment, optimise substitution or spread inventory risk across a larger network.

Aggregate that demand and the first cost reduction is available before a single manufacturing process improves. Consolidated procurement creates purchasing power. Standardised assortment reduces complexity. Better substitution directs demand toward equally effective lower-cost products. Digital routing reduces inventory mismatch, and network-level data makes it possible to place stock where it is most likely to move.

The winning pharmacy model will therefore not merely sell cheaper medicines. It will learn faster because every transaction improves purchasing, substitution, inventory placement and demand prediction. DawaaDost is one attempt to build that aggregated demand pool across Indian pharmacy, and I founded it.

Diagnostics is already moving in the same direction. The neighbourhood laboratory once won because it was nearby. Reliable collection and transportation allowed the analytical work to move into central hubs, where machines could run for longer, reagents could be purchased at scale and quality systems could be spread across far more samples.

Thyrocare is a clear Indian expression of this model: a wide collection network feeding concentrated processing infrastructure. It has historically generated lower revenue per patient than some peers while sustaining strong margins because the economics are driven by density, automation and machine utilisation rather than by maximising the price of each encounter.

A blood sample does not care whether the analyser is ten streets away or several hundred kilometres away. Once local collection and transportation are reliable, the advantage shifts toward the operator capable of processing far more samples through the same hub.

Software adds another layer to this concentration. The same network that centralised the machine can increasingly centralise, prioritise and automate portions of interpretation. Proximity first lost to logistics. Routine interpretation is now beginning to lose to software.

Hospitals are the hardest part of the system because they contain too many different units. A general hospital may perform hundreds of procedures, serve dozens of specialties and mix unpredictable emergency work with scheduled care under one roof. That variety makes repetition difficult and leaves expensive capacity exposed to interruption.

The best food in India often comes from businesses that make one thing exceptionally well: the idli shop, the biryani shop, the kebab shop with four items on the board. Nobody expects the best meal from a restaurant with a two-hundred-item menu. Indian hospitals have spent decades running two-hundred-item menus and calling the result comprehensive care.

Focused healthcare factories behave differently. Narayana built cardiac surgery around high procedure volumes, standardised teams and unusually high asset utilisation. It has delivered surgery at a fraction of American prices while reporting strong outcomes and attractive profitability. Aier Eye used a hub-and-spoke model to build scale across ophthalmology. Clínicas del Azúcar created a standardised annual diabetes-care pathway at a price accessible to Mexican families and achieved materially better disease control than the public system.

The lesson is not that every hospital should perform only one procedure. It is that every repeatable pathway inside a hospital should be treated as a production system.

Scheduled care should be separated from unpredictable work wherever possible. The unit should be defined clearly. The pathway should be standardised. Volume should be concentrated. Expensive equipment should remain active, and specialists should spend their time on the elements of care that are genuinely variable.

The hospital that reorganises first will not necessarily be the one with the newest tower, the largest lobby or the widest specialty menu. It will be the one whose counters advance fastest.

Across pharmacy, diagnostics and hospitals, the likely winners share one trait: each identifies a unit and organises the business around compounding its cumulative volume.

In pharmacy, the winner is the aggregator capable of turning fragmented storefronts into one demand pool. In diagnostics, it is the hub-and-spoke network that understands logistics can outperform proximity. In hospitals, it is the focused operator performing one pathway ten thousand times rather than dozens of pathways occasionally.

Underneath all three sits an intelligence layer that reads the image, flags the interaction, triages the queue, predicts demand and removes scarce professional time from routine work.

The losers will not always look obviously obsolete. They will include the chemist whose economics depend on a trade markup that consolidated buying can compress; the standalone laboratory whose principal advantage is being nearby; the nursing home performing too many procedures too infrequently to become excellent or inexpensive at any of them; and the mid-tier hospital whose prices were built around scarcity rather than throughput.

They may not be defeated by visibly superior technology. They may simply discover that another operator has been learning faster on a counter they were never running.

There is, however, one final qualification. A low unit cost is not sufficient to create a durable institution.

Jan Aushadhi has built a powerful affordability programme, with thousands of stores selling medicines substantially below branded prices. Its franchise economics nevertheless demonstrate the difficulty of reducing price without simultaneously creating enough throughput, inventory flexibility and consumer demand to make the retail model self-sustaining.

The price can fall while the business engine remains incomplete.

Babyl demonstrated the same problem from a different direction. Its Rwanda service reached millions of users and showed that digital consultations could reduce time, unnecessary prescribing and laboratory use. Yet the wider company collapsed. The consultation represented only the front door of healthcare; much of the economic value remained in diagnostics, medicines and downstream treatment.

Babyl had made the front door extraordinarily efficient without owning enough of the house.

These examples do not weaken the low-cost thesis. They clarify it. Low unit cost without sufficient volume is not a business. Volume without durable economics is not a business. And volume without a meaningful claim on the surrounding value chain may create social value without creating an institution capable of sustaining it.

The design problem is now visible.

Choose a unit and standardise it until the process becomes boring. Organise credible demand before demanding a lower price. Ensure that the fourth and sixth competitors can arrive and survive. Put software where scarce human expertise is limiting distribution. Run the expensive assets already installed for more hours each day. Then trade margin for throughput and resist the temptation to take the margin back when scale arrives.

Moore’s Law will not arrive in Indian healthcare automatically. Cost curves never do. They have to be manufactured through procurement, competition, standardisation, software and operational discipline.

For the first time, every major component required to manufacture one is sitting on the table.

The run has just begun.

Disclosure: I founded DawaaDost, a pharmacy retail chain operating in the segment described above.

Authored by: Amit Choudhary

Linkedin Profile: https://www.linkedin.com/in/amitchoudhary1/

Publishing Support:  Namrata