Most AI-first techbios are modeling clinical costs that are roughly half of what they will actually face.

TL;DR

Most AI-first techbios are modeling clinical costs at roughly half of what they will actually face. The gap between a founding-era financial model and the real cost of reaching an IND has roughly doubled within the lifespan of one mRNA company. AI investment is concentrated upstream in target ID and lead generation, while the cost wall sits downstream in CMC, manufacturing readiness, and regulatory throughput — a different problem, largely unaddressed. Jacob Becraft of Strand Therapeutics has taken this operational critique public, calling for IND reform and questioning duplicative regulatory gatekeeping that slows first-in-human trials.

Most AI-first techbios are modeling clinical costs that are roughly half of what they will actually face.

When Jacob Becraft spun Strand Therapeutics out of Massachusetts Institute of Technology in 2017, the financial model projected phase 1 trials at $50K to $100K per patient and a total IND cost of around $9M.

When Strand actually submitted that IND about two years ago, the per-patient cost was over $400K. Reaching the clinical starting line, including manufacturing and preclinical work, had cost more than $20M.

That arc, inside one company's own founding-to-IND lifespan, is the data point most AI-first techbios, in their transition to a product-first biotech, are not modeling.

Becraft put those numbers in front of the House Select Committee on China in March and expanded the argument in Fierce Biotech last week. His broader case: the defining competitive metric in biotech has shifted from who discovers a therapy first to who can turn that discovery into first-in-human data fastest. Early human data derisks the program. Investors step in. Partnerships form. Manufacturing scales. Sites gain trial-running experience. Lose that flywheel to other jurisdictions and the entire downstream ecosystem follows.

His specific ask is narrower than the headline. He wants CMC requirements on IND filings reduced for novel modalities. And he questions whether IRB and FDA both need to gate first-in-human trials when Australia and China get away with one. An operational critique from someone running a programmable mRNA platform, not a policy slogan.

This is where the AI-pharma capital flows get uncomfortable. Boehringer, AstraZeneca, Sanofi, and Roche have all stood up dedicated AI capability centers in the last 18 months. AI-first techbios are scaling discovery throughput on the same logic. Nearly all of that investment sits upstream of the IND, in target ID and lead generation. The cost wall Becraft is describing sits downstream, in CMC packages, manufacturing readiness, and regulatory throughput. Different problem, different teams, largely unaddressed by the current AI buildout.

The capital required to put an AI-derived molecule in front of a patient has roughly doubled within the lifespan of one mRNA company. Discovery cost reductions are real and welcome. The binding constraint sits elsewhere.

Met Jake in Riyadh last year and was struck by how operationally specific his framing was even then. This piece reads like that conversation extended into a public position.

https://www.fiercebiotech.com/biotech/concerted-industry-government-effort-strand-ceo-jake-becraft-fda-ind-reform-and-china

Key takeaways

• Most AI-first techbios are modeling clinical costs at roughly half of what they will actually face.
• Strand Therapeutics' own founding-to-IND arc shows per-patient costs rising from a projected $50K–$100K to over $400K, and total IND costs from ~$9M to more than $20M.
• The defining competitive metric in biotech has shifted from who discovers a therapy first to who can turn that discovery into first-in-human data fastest.
• Losing the first-in-human flywheel to other jurisdictions means the entire downstream ecosystem — investors, partnerships, manufacturing, trial-site experience — follows.
• Nearly all current AI investment sits upstream of the IND; the cost wall sits downstream in CMC, manufacturing readiness, and regulatory throughput — a different problem, largely unaddressed.
• Becraft's specific ask is narrower than the headline: reduce CMC requirements on IND filings for novel modalities and question whether both IRB and FDA need to gate first-in-human trials.
• Discovery cost reductions from AI are real and welcome, but the binding constraint sits elsewhere.