Biotech Is Entering Its Netflix Era

AI Just Created a Bigger Problem Than Cancer

The pharmaceutical industry has spent the better part of the last three decades worrying about a shortage of innovation. Investors complained about empty pipelines. Executives spoke with the hushed solemnity of medieval peasants discussing a weak harvest. Governments panicked over antibiotic resistance and neurodegenerative disease. Analysts produced charts showing the ever-rising cost of discovering a single approved drug, as though nature itself had decided to become less cooperative.

Then artificial intelligence arrived and quietly inverted the problem.

The popular narrative is that AI will make better drugs. It probably will. But that observation, while true, misses the more disruptive point entirely. The real consequence of AI is not merely higher quality drug candidates. It is the industrial-scale production of plausible ones. Biology is entering its content-generation era, and much like the internet before it, society appears wildly unprepared for what happens when supply becomes effectively infinite.

For most of modern pharmaceutical history, generating a credible drug candidate was the bottleneck. One required armies of medicinal chemists, years of wet-lab iteration, vast screening libraries and enough patience to survive multiple management restructurings. Discovery was expensive because ideas themselves were scarce, risk was high, and there wasn’t enough capital to chase them. AI changes this equation in the same way machine tools changed manufacturing or cloud computing changed software deployment: by collapsing the marginal cost of experimentation.

A sufficiently capable generative biology system does not merely suggest one molecule. It suggests ten thousand. It does not produce a single protein binder. It produces entire landscapes of them, each with tolerable ADMET properties, respectable novelty scores and docking simulations polished enough to impress a venture capitalist who once read half of The Innovator’s Dilemma on a flight to San Francisco.

This creates an awkward economic reality. The pharmaceutical industry has historically behaved as though the limiting factor was imagination. Increasingly, the limiting factor is something much more mundane: the number of human beings available to run clinical trials on.

There are only so many patients, investigators, regulatory reviewers and phase II oncology slots at major academic hospitals before the waiting list begins to resemble Soviet bread distribution. The industry is about to discover that while software scales elastically, biology does not. A human liver remains stubbornly analogue technology.

This means an extraordinary number of “good” drugs will never be tested - not bad drugs, not fraudulent drugs, but perfectly reasonable, potentially efficacious, commercially viable drugs. They will simply die in spreadsheet hell because the system lacks the throughput to evaluate them all.

The public imagination still treats drug failure as primarily scientific. In reality, an increasing fraction of future failure will be logistical and financial. A molecule may demonstrate strong preclinical evidence, acceptable toxicity and mechanistic elegance, yet still disappear because another candidate with marginally better biomarker stratification received the budget allocation instead. Investors like to believe capital flows toward the best science. In practice, it often flows toward the most narratively convenient science.

This is not unprecedented. Hollywood learned the same lesson years ago. Cameras became cheap. Editing software became free. Distribution became frictionless. The result was not a golden age of universally brilliant cinema, but an incomprehensible tsunami of content in which the scarcest resource ceased to be production capability and became human attention. Pharmaceutical R&D is drifting toward the same equilibrium, except instead of forgotten Netflix thrillers there will be abandoned therapies for inflammatory bowel disease.

The irony is particularly sharp because biotech spent years envying software economics. Now it may inherit software’s worst pathology: overwhelming abundance combined with catastrophic selection difficulty.

The venture industry is especially vulnerable here because it still behaves as though candidate generation itself confers defensibility. It increasingly does not. A startup announcing that its platform identified a novel target is beginning to sound suspiciously like a startup in 2022 announcing it had “an AI chatbot”. Interesting, certainly, but no longer rare enough to matter.

What becomes valuable instead is judgment.

Not scientific judgment alone, though that matters immensely. Rather, economic judgment under radical abundance. Which candidates deserve scarce trial slots? Which indications justify capital concentration? Which biomarkers actually reduce downstream uncertainty rather than merely decorating a pitch deck with heat maps? In a world overflowing with plausible molecules, the decisive capability is refusing most of them.

This represents a profound inversion of pharmaceutical strategy. Historically, large portfolios existed because discovery was uncertain; firms needed many shots on goal to survive. In the future, portfolios may expand because discovery becomes too successful. The challenge shifts from finding needles in haystacks to deciding which of the thousands of glittering metallic objects are worth the magnet time.

The consequences for valuation could become uncomfortable. If AI systems can routinely generate high-quality candidates, then molecules themselves begin to commoditize. Investors may discover, to their horror, that owning one promising preclinical asset is less valuable than owning superior trial infrastructure, patient recruitment networks or regulatory optimization capabilities. The future giants of biotech may look less like traditional pharmaceutical firms and more like sophisticated capital allocators with clinical logistics engines attached.

There is also a darker implication, which the industry prefers not to discuss in polite company. The current regulatory system was designed for a world where drug candidates were scarce. It is catastrophically mismatched to a world where they are abundant. Regulators already struggle with workload. Adding orders of magnitude more credible therapies without changing evaluation frameworks guarantees one of two outcomes: either vast quantities of beneficial drugs never reach patients, or approval standards become increasingly automated and probabilistic.

Neither option inspires universal comfort.

And yet the oversupply is coming regardless, because the incentives are too powerful. Every improvement in generative biology compounds the problem further. Better foundation models create more candidates; better simulation tools reduce filtering costs; cheaper synthesis accelerates validation; robotic labs increase iteration speed. The industry is building a machine whose output grows exponentially while the clinical system consuming that output remains almost comically artisanal.

One suspects future historians may look back on this era and conclude that humanity solved the problem of generating medicines long before it solved the problem of choosing among them.

The pharmaceutical industry still speaks as though discovery is the hard part. Increasingly, discovery is becoming the easy part. The hard part is deciding what not to pursue when thousands of seemingly good ideas compete for finite money, finite patients and finite time.

Scarcity, after all, does not disappear when technology advances. It merely migrates.