Replacing one Keytruda takes ninety average launches. Life sciences isn't built for that

Dushyant SapreFounder & CEO, SwishX·9 min read
Contents
  1. Introduction
  2. The revenue at risk sits in a few very large brands
  3. Replacement revenue arrives in much smaller pieces
  4. More of those launches start in someone else's lab
  5. The IRA adds a second clock, and it pushes the same way
  6. A launch has to cost a fraction of what it costs today
  7. What I would do before 2028

The patent cliff is usually told as a pipeline and M&A story. For commercial teams it is an operating model problem: the replacement revenue arrives in many smaller launches, and the cost of a launch has to fall to match.

Almost everything written about the patent cliff ends at the pipeline and the deal desk: which assets will fill the hole, and who will buy whom. That framing is right, and it is incomplete. For the commercial side of a life sciences company, I think the cliff is an operating model problem, and the companies that treat it as one will spend this decade very differently from the ones that don't.

The reason is arithmetic. The revenue that expires before 2030 sits in a handful of very large brands. The revenue that replaces it will arrive in many smaller ones, and a growing share of them will have been discovered by someone else. A commercial organization built to launch a few big brands a decade cannot launch dozens of small ones at the same cost per brand. Something has to give. My bet is that it is the cost of a launch.

The revenue at risk sits in a few very large brands

The size of the cliff is not really in dispute, only its edges. Evaluate's World Preview, published in June 2025, put more than $300 billion of sales at risk from patent expiry by 2030. GlobalData estimated that the US market alone would lose more than $230 billion between 2025 and 2030. The methods differ. The order of magnitude does not.

What matters more than the total is its shape. Keytruda brought Merck $31.7 billion in 2025, out of company sales of $65.0 billion. That is one brand carrying almost half of one of the largest companies in the industry, and Merck's own annual report says it expects biosimilar competition to hit Keytruda materially between 2028 and 2029. At Bristol Myers Squibb, Eliquis ($14.4 billion) and Opdivo ($10.0 billion) made up about half of 2025 revenue of $48.2 billion, and the company lists 2028 as the expected US exclusivity date for both.

Stelara shows what the far side of the edge looks like. Biosimilar competition arrived in 2025, and J&J reported that Stelara's decline took about 1,040 basis points off its Innovative Medicine growth for the year. The segment still grew 5.3% operationally, because Darzalex, Carvykti, Erleada and Rybrevant were already in market and growing. That is the lesson I take from it. The cliff is survivable when the replacements are already launched. It is much harder when they are still being launched.

The standard responses are well known. Merck won FDA approval for a subcutaneous form, Keytruda Qlex, in September 2025. BMS sold $238 million of its subcutaneous Opdivo Qvantig in 2025. BMS also bought Karuna for $14.0 billion in 2024. All of that is the pipeline and M&A story, and it is where most analysis stops.

Replacement revenue arrives in much smaller pieces

The best public view of what the pipeline is worth comes from Deloitte's annual study of the 20 largest pharma companies. Its 2026 edition, published in May, puts the average forecast peak sales of a late-stage asset at $598 million. Take out the GLP-1 class and the figure drops to $353 million. More than half of pipeline programs are forecast to peak below $250 million, and about 9% of late-stage projects account for roughly 70% of forecast revenue.

Now do the division. Keytruda's 2025 sales equal about ninety average non-GLP-1 assets at peak. Nobody will literally replace Keytruda with ninety launches, and Merck will not try. But the number is a useful way to see the problem. I think of it as the replacement ratio: how many typical launches it takes to refill one dollar of expiring revenue. For most of the industry, that ratio is going up.

The approval mix points the same way. Of the 46 novel drugs FDA's drug center approved in 2025, 23 had orphan designation. Rare and specialty indications are where much of the science is going, and by definition they treat fewer patients per brand.

BMS's own numbers make it concrete. Nine of its newer products, Reblozyl, Breyanzi, Opdualag, Camzyos, Zeposia, Abecma, Sotyktu, Krazati and Cobenfy, sold about $7.6 billion together in 2025. That is roughly half of what Eliquis sold on its own. Several are early in their curves. Cobenfy, which came with Karuna, sold $155 million in 2025 and may well become very large. I am not judging any one of them. The point is structural: the portfolio that replaces a blockbuster looks like a portfolio, not like another blockbuster.

More of those launches start in someone else's lab

The second shift is where the assets come from. An IQVIA Institute report from October 2025 found that emerging biopharma companies originated 65% of the 394 new active substances launched worldwide from 2020 to 2024. In the US, emerging companies also commercialized 46% of new active substance launches in that period. Of the 197 drugs they launched over the past decade, 58 later moved with an acquisition of the company and 21 were out-licensed.

Read that from the brand team's chair. A growing share of large-company launches will be products the company did not discover, often acquired late, sometimes months before a filing or after one. The evidence was generated by someone else's clinical team. The claims were argued over in someone else's meetings. The early content, if any exists, was written to someone else's standards. The commercial organization inherits all of it and has less time than it used to have to make it its own.

There is a practical cost here that rarely shows up in a deal model. An acquired asset arrives with its own agencies, its own document conventions and its own references stored in its own systems. Before the new owner can put a single promotional piece in front of a physician, someone has to rebuild the claims and their support inside the new owner's review process. On a large brand that work is a rounding error. On a small one it can consume months the launch does not have.

China accelerates this. Evaluate expected Chinese-origin assets to make up nearly 40% of global licensing deals in 2025. I will come back to that in a separate essay. For now, the relevant point is that the origin of a launch and the owner of a launch are drifting apart, and the gap lands on the commercial side.

The IRA adds a second clock, and it pushes the same way

In the US, patent expiry is no longer the only clock. Under the Inflation Reduction Act, Medicare's negotiated price for Eliquis took effect on January 1, 2026: $231 for a 30-day supply against a list price of $521. That cut arrived about two years before US generics are permitted. In January 2026, CMS selected 15 more drugs for prices that take effect in 2028, and for the first time the list included drugs paid under Medicare Part B, the physician-administered side of the program.

The rules also shape what gets developed and in what order. A small molecule becomes eligible for selection seven years after approval and a biologic after eleven. In July 2025 Congress widened the orphan exclusion to cover drugs with one or more orphan designations, and moved the start of the clock for a former orphan drug to its first non-orphan approval. If you run a portfolio, the incentives now favor biologics, rare indications first and broader indications later.

Each of those choices produces more, smaller launches. An orphan-first strategy means a first label for a small population, then a second launch into a larger one, each with its own evidence, materials and review. I am less sure how durable these rules are, since Congress has already changed them once. But for planning purposes, the direction is clear, and it compounds the replacement ratio rather than offsetting it.

A launch has to cost a fraction of what it costs today

Here is where I think the commercial conversation has to change. A launch carries a large fixed cost that does not shrink much with the size of the opportunity: a field strategy, a medical plan, a payer value story, advisory boards, a content library for every channel and audience, and the review work behind each piece. In a blockbuster world that cost was spread over a brand selling $10 billion or more a year. Spread over a $350 million peak, the same cost structure eats the brand.

I spent years in ad tech before life sciences, and the closest parallel I know is the long tail of small advertisers. Those businesses always existed. They became a market only when the cost to set up and serve one account fell close to zero, through self-serve tools and automation. The companies that won did not sell harder to small advertisers. They rebuilt the cost structure so that a small account was worth serving at all. I think launches are heading the same way.

Content and review are where the fixed cost hides. A rare-disease brand with a few thousand treating physicians still needs HCP, patient, payer and digital materials, and every one of them still passes medical, legal and regulatory review. A company that ran five large brands through one review team may now need to run fifteen smaller ones through the same team. The cost of slow review used to be a nuisance on a big brand. On a small one it is a material share of the margin. Outside the US the load grows too: since January 2025, new cancer medicines and advanced therapies in the EU go through a joint clinical assessment, and orphan medicines join in January 2028.

The strongest argument against all this is the GLP-1 class. Deloitte finds that GLP-1 assets make up about 38% of forecast late-stage revenue at the top 20 companies, and obesity alone about 25%, up from 1% three years earlier. If a few mega-brands refill the cliff, the blockbuster operating model survives. I think that is true for the companies that own those assets. For everyone else, the number that matters is $353 million.

What I would do before 2028

First, measure the fixed cost of a launch, not just the launch budget. Most companies know what they spend on a brand. Far fewer know what it costs to take any brand from approval to a fully reviewed, multichannel content set and a trained field team, regardless of its size. That number is the floor under every small launch, and it should have an explicit target to fall.

Second, build launch capability once and share it across brands. That means one claims infrastructure where every claim is tied to its source, review capacity planned across the portfolio rather than brand by brand, and content built in modules so that a new indication reuses what is already approved. This is a different operating model for content, and it has to exist before the wave of small launches, not during it.

Third, put commercial readiness into the deal. When business development evaluates an asset, someone should ask how long it will take, after signing, to get a reviewed core content set and a credible evidence story in front of physicians. That time is now part of the value of the asset, even though it appears on no term sheet.

Three questions are worth asking at the next leadership offsite. What does our cheapest credible launch cost? How many launches a year can our review and medical capacity actually carry? How many days does it take us to go from a signed deal to approved materials in the field? If the answers are "we don't know", the cliff is closer than the patent dates suggest.

The companies that do well this decade will not be the ones that found the next Keytruda. They will be the ones that could afford to launch ninety things that weren't.

Filed underPerspectives

Keep reading

All articles

Want to see Magic Studio on your brand?

Video, visual aids, ads, approved email and web, made from your approved claims and ready for MLR review.