Why a New Drug Now Costs $879 Million, and What the Bust Could Fix

$879 million. That's the average cost, in 2018 dollars, of getting one new drug through the US approval process once you count the failures and the capital tied up along the way, according to a two-decade JAMA analysis of drug development economics. The figure is nearly five times what it was in the mid-2000s. Each approved medicine now arrives with a longer, more expensive R&D ledger behind it, and the industry's own data shows the problem has been compounding for 30 years.

Economists call it Eroom's Law, which is Moore's Law spelled backward, and it describes the steady decline in drug approvals per dollar of R&D spending since the 1950s. Every cycle of new technology, every genomics platform, every AI-driven target screen was supposed to break the curve. None of them did. A 2024 Deloitte report on biopharma R&D productivity puts the industry's return on R&D investment at 5.9%, up from a 2022 low of 1.2% but still well below the 11%-plus cost of capital that large drug companies and their investors require.

For more than two decades, the venture model papered over that gap. Cheap money, IPO windows, and a steady supply of crossover funds kept the pipeline of unprofitable science moving. That is no longer the case. Crunchbase deal data shows biotech venture deal counts in 2025 are down roughly two-thirds from 2021. In Seattle, one of the three US hubs that anchor drug development alongside Boston and San Diego, lab vacancy has crossed 50%, according to broker HughesMarino. Empty lab space is the physical footprint of a venture model that stopped writing checks.

Why does getting a new drug approved keep getting more expensive? Part of the answer is the Baumol effect, a labor-cost phenomenon economists William Baumol and William Bowen described in the 1960s. Sectors whose output cannot be automated, like teaching, live performance, and drug discovery, see wages rise with the rest of the economy even when their productivity does not. A modern drug is a 12-year science project staffed by PhD scientists, clinical coordinators, and regulatory specialists, and the cost of that labor has tracked the broader professional economy even as the science has gotten more efficient in narrow ways. The result is a structural cost disease. R&D gets cheaper per data point, but more expensive per approved drug.

The current crash is doing what no prior bust did. The 2008 downturn was too short. The 2015 and 2016 correction was a market story. The post-COVID reset is the first one where the underlying cost problem is the central pressure, not a backdrop. The JAMA figure, the Deloitte IRR series, the empty Seattle labs, and the Crunchbase deal collapse are all symptoms of the same imbalance: the industry has been earning less than its cost of capital for years, and the capital that funded the gap is now gone.

What "build back better" would actually require is a different cost structure, not a smaller version of the current one. That means funding the parts of the pipeline where marginal science still works, including platform technologies, validated targets, and late-stage clinical trials, and starving the parts where it does not. It means treating Seattle, Boston, and San Diego as labor markets whose productivity matters, not just real-estate markets to be backfilled. It means accepting that some of the ventures that raised money in 2020 and 2021 will not exist in 2027, and that the next decade of medicines will come from a smaller, more disciplined set of programs.

The column that named this moment, in GEN News, treats the post-COVID bust as a chance to rebuild the industry around the cost economics that Eroom's Law and the Baumol effect have described for three decades. Whether the surviving hubs use it that way, or just rebuild the same venture-funded model at smaller scale, will be visible in the next twelve months of hiring data and Series B term sheets. The 5.9% IRR is a number. The choice of what to do with the empty labs is the news.