Why there are no trillion-dollar biotechs

Nvidia is worth more than every pharmaceutical company. Some say biotech CEOs should see this as a challenge. But if they try for the big T, markets will constantly signal that it's the wrong goal.

This essay is a ‘yes, and…’ response to Lada Nuzhna’s piece on missing trillion-dollar biotechs. Lada concluded that neither genetically validated targets, drug repurposing, nor AI will change biotech efficiency or success rates enough to create extremely valuable companies. Instead, she proposes targeting larger markets, primarily in diseases of aging. This essay argues that even with that strategy, key features of very successful companies are missing from how biotechs are currently being built.

To start off, let’s quickly review why value creation in biotech is harder than most industries. This creates a baseline for the scope of the challenge.

Bioscience is harder than rocket science

The first reason is that ‘science is hard’. Finding the right approach to curing a disease is like finding a specific bug in millions of years of vibe-code. We start with an idea of what we want to do (cure Alzheimer’s, say), but we’re not sure how to make that happen or what the rules of our system are. Even sending rockets to space is a much more predictable kind of engineering.

The science is particularly hard because of the limited tools available to biotechnologists. We can’t take a person with Alzheimer’s and with the press of a button make a change to their brain. In fact we are not allowed to do anything to a human until we’ve done a lot of preliminary work. For both practical and legal reasons, we instead have to go work on something that is not Alzheimer’s but hopefully relevant (individual brain cells, or a mouse engineered to mimic parts of the disease) to be able to try things at all. And even here, we often struggle to manipulate the model system in the ways we want. Cells are crowded and bustling with activity. Testing a biological hypothesis usually means a targeted change to one protein inside the cell. Before we test our idea for (hopefully) treating disease, we often need to engineer a way to make that targeted change. If you’re not a biologist, think of this as trying to add chili powder to a finished burrito, but only on the onions. It can be done, but only with considerable ingenuity and perspiration.

Biotech means building for a customer who needs a very very specific physical product, but will never tell you what the specs are.

Similar problems can occur in other industries. For example, you know you want a user to get the search results they care most about, but not how. Or you’re building an engine before you have a spaceship to launch. But in general it is much easier to work up an idea and get direct feedback on whether it is good: write and run the code quickly, or test the engine on the ground. Biotech only gets incontestible feedback from human testing. This is akin to not being allowed to test your rocket engine before the final launch [1].

Making the experiments of early biotech faster and cheaper is good and useful. But it won’t speed things up very much as long as real feedback comes from human trials. Predicting human outcomes better would be far more useful, because it would both increase eventual success rates and allow more iteration with strong feedback.

How biotech works: Solve a series of different problems, each difficult: Generate insight into how a disease works and how we can treat it, then create a physical product that does what you want, then test it in human trials.

Biotech does not want to become a trillion-dollar company

But the reason there is no trillion-dollar biotech is not just because ‘science is hard’. Rather, the ecosystem we have created to create and capture value in biotech deters and destroys would-be trillion-dollar companies.

Extremely valuable companies tend to have two features:

1. At least one flywheel or network effect where growth makes future growth easier, because more users makes the product more valuable or because multiple offerings boost each other’s success.

2. Near-monopolistic distribution of something that’s valuable to a very large market. This could be the best car or phone, essential tools for a growing industry (like chips or payment systems), or the attention of billions.

The first of these is very rare in biotech. The second is common, but not robust. Let’s explore:

Turning flywheels into firewood

Enough has been said about flywheels in business. When you are good at something and your output makes you even better, growth can take off. Unfortunately, the standard biotech business model actively destroys flywheels as a way to minimize risk.

For most businesses, scaling to the next stage happens when you have growing revenue.

In biotech, you scale to the next stage without revenue but with highly concentrated risk.

For example: When a biotech seeks funding to launch clinical trials, it is naturally encouraged to derisk in the most cash-efficient way. Get the key data, and we’ll give you money for big trials. Regardless of AlphaFold, Nobel Prize winners, and whatever else, success rates when drugs first go into humans sit at 10-15%. Investors know this, and so prefer to wait for the human readouts before piling on more cash.

But by that time the company will have ‘molted’, shedding its early discovery team to conserve cash during the years of clinical trials. We are a humane industry, and the scientists are not condemned to death or even exiled. They disseminate out into other biotechs and do good work there. But in most cases they are no longer working together - the superorganism that created a new medicine has died.

This phenomenon occurs because 1) once clinical trials are launched the product and design are locked in by FDA filings, and cannot benefit from new insights or improvements, and 2) successful clinical trials reward successful early discovery and drug development, but not until many years after the work is done. So the stages of value creation and capture are de facto decoupled, and unable to form a flywheel.

Putting a medicine on the market involves many steps that each must succeed, but all rely on clinical trials to capture value. And so molting can happen at multiple points of a biotech’s life: When gleaning biological insights about disease shifts to drug-making medicinal chemistry, when new insights shift focus to a different disease area, when drugs eventually go into clinical trials…

Each time it happens, it makes sense as a way to improve ROI by limiting the cost of failures. And each time it happens, we pull up the tracks behind us and feed them to our train’s furnace: We will get where we’re going, as fast as we can! But no railroad system will be left behind for future efforts.

For my next trick, I’ll make my TAM disappear

At least biotech get monopolies, right? Sort of. When a new drug is approved there’s a period of patent protection (typically ~a decade) where competitors aren’t allowed to copy your product. This what gets investors to fund the science and clinical trials. But it is explicitly temporary. When the protection ends, companies all over the world will start selling the same product and profit margins will fall dramatically.

Because the value we create in biotech is not the physical molecules, but the knowledge that this molecule is an effective way to treat a disease. This is great for society: the knowledge we generate stays with humanity forever. It’s the only part of healthcare where the same treatment gets cheaper over time. (footnote Peter Kolchinsky’s book).

Longevity is a rare area of medicine where success expands your market instead of shrinking it.

So the innovative company must innovate again, make a new product that’s even better. This too is great. It’s how a free market society makes progress. But companies treating disease have a unique obstacle for how they innovate: If the first product worked, the disease is gone or diminished. Once you’ve brought someone’s cholesterol back to normal, they don’t need another cholesterol drug. Maybe you can reduce some side effects or make it more convenient, but the largest markets will be doing something totally different.

This is very different from providing a way to handle online payments. Or even sending rockets to space (at least with resuable rockets). Companies that keep growing are able to use what they’re good at to make better and better products. As the products get better, more and more people want to use them. It’s good for society that this doesn’t happen when you’re curing disease, but it makes it hard for biotech companies to keep growing.

Lada’s proposal to target age-related diseases touches on this point: the biggest successes in drug development (anti-TNF drugs, GLP1Ras) have targeted very large patient groups. But therapeutic success still shrinks the market for your product, and your patent-protected profits will expire after a decade. To achieve the dynamics that robustly support a trillion-dollar biotech based on solving a single problem, we must go further than targeting a very large market: a product that increases its own demand could help create a flywheel. Within healthcare, we have a terrible example of this playing out: addictive substances including oxycontin. But an effective therapy that delays aging would produce the same dynamic, while making lives better: It starts with a massive TAM, and the product working keeps the users alive and hungry for more.

What can a poor trillion-dollar biotech seeker do?

Perverse flywheel and TAM dynamics in biotech are the result of how we as a society invest in medicine. There are no trillion-dollar biotechs in part because we don’t really want them. We pay for treatments rather than prevention because it solves an urgent problem and we can quickly tell whether we succeeded. Biotech investors are intelligent but operate with constraints: the cost of capital, the rate of progress, the patience and the literal lifespan of everyone involved. They are making rational decisions within those constraints and decide (implicitly or explicitly) to aim for something other than a trillion-dollar company. For example, a good chance of a 3x return when their company with a single promising heart failure drug sells to Novartis in Phase 2 [2].

There’s a lot of focus on technologies that could make it easier, and thus faster/cheaper, to do biotech. This could shift these ecosystem incentives and make investors more likely to double down on platforms. But most of the cost and time happens in clinical trials, where a new CRISPR tool doesn’t make a difference. And without a functional and competitive flywheel, there is no reason why value would accumulate in a single company versus distributed across many.

Successfully setting up a way to sell preventative medicine to a massive market could produce a trillion-dollar biotech, but you’d need to figure out a moat other than patent protection. Especially if you aim to sell directly to consumers, thus competing with people who don’t limit themselves to selling things that actually work.

I suspect that the first trillion-dollar biotech will be one that solves connecting a core technology to repeatable value capture in order to create a flywheel effect. Achieving this will probably mean building towards it from day 1. ‘TechBio’ companies are often pitched as ‘flywheel first’. High-throughput experiments generate data for AI models that will discover new therapies. But a new preclinical candidate is often not enough to capture value. So the companies face a challenge at the clinical stage, where the capital required to take many shots is immense. A flywheel that relies on convincing investors to fund dozens of clinical trials before you show a profit will be hard to get going. If you can link value capture directly to the output of your platform, with a platform that does not start from scratch when you move to a different disease, you might have a shot.

Given how biotech works, what combination of tricks would suffice?

• Targeting a very large market D will pay off if successfuly, but limited duration of protected profits will prevent a proper flywheel from forming.

• A technology that creates a flywheel at just stage B (for example new way to target genes, like Alnylam) could outperform within status quo, but will be no more likely than average to generate a second success.

  • If it opens up a meaningful number of existing insights (e.g. genetic targets), it could still have a very good run!

• There are two designs that seem feasible to create a value-capturing flywheel:

  • A technology platform that can solve both A and B in a self-improving way, combined with either a way to monetize prior to C (challenging given current value curve) or access to funds/profit to take multiple shots at C.

  • A (correct) insight A that fits a market D that does not shrink with success, combined with a self-improving platform for B and access to funds/profit to take multiple shots at C.

---

[1] Notice that most of the breakthroughs from LLMs and other recent AI are in areas where it’s easy to tell whether the output is good: winning at games, plain language responses, code that runs.

[2] We do sometimes swing for trillion dollar bets. One company that seemed to have a chance a few years ago is Moderna (~$200B at peak). Notably, it had both a platform that could rapidly be redeployed across diseases, had quick clinical readouts (for the areas they eventually broke through with), and in the case of vaccines with boosters avoided the disappearing TAM issue. On the other hand, the most valuable therapeutic companies currently (whether century-old pharmas or the winners of the first wave of ‘founder led biotech’ in 80s-90s) don’t really fit the criteria I describe. Could be I’m wrong, or could be that they’ve grown over time but are not on a trillion-dollar trajectory.

[3] If I’m wrong, it’ll be Eli Lilly solving another huge market like neurodegeneration, after doubling down on innovation using Tirzepatide revenues.