Happy New Year! My 5 bets for AI in Healthcare + TechBio in 2026

TL;DR

I'm putting forward five concrete bets for where AI in Healthcare and TechBio is headed in 2026. Ambient scribes get commoditized, the first "AI-discovered drug" headline arrives, and self-driving labs move from demo to production — but stay narrow. Clinical trial AI becomes a default input, and the TechBio Series B shakeout accelerates as 2024's bridge rounds turn into quiet shutdowns. I'm asking: if one of these is wrong, which one, and what would you need to see to change your mind?

Happy New Year! My 5 bets for AI in Healthcare + TechBio in 2026

1) Ambient scribes become table stakes

Note-taking gets bundled into EHR/RCM or absorbed into workflow suites. Europe and ROW finally see traction.

2) We get the first real "AI-discovered drug" headline

Takeda's zasocitinib (TAK-279), originally associated with Nimbus/Schrödinger's platform, has positive Phase 3 readouts and is positioned for a 2026 FDA filing. It won't end the fundamental debate: clinical validation is still the bottleneck.

3) Self-driving labs move from demo to production, but stay narrow

As foundation models commoditize the in-silico layer, differentiation shifts to whoever owns the closed-loop data flywheel. Pure "better algorithm" stops being a story.

4) Trial-risk prediction becomes the default, not the differentiator

AI-enabled clinical trial modelling and optimization becomes a standard input to protocol, sites, and go/no-go decisions.

5) The TechBio Series B shakeout accelerates

2024's "bridge to nowhere" rounds become 2026's quiet shutdowns or fire-sale acquisitions. The survivors are the ones with true pharma revenue, not just pharma "partnerships."

If one of these is wrong, which one, and what would you need to see to change your mind?

Key takeaways

• I expect ambient scribes to become table stakes in 2026, bundled into EHR/RCM or workflow suites, with Europe and ROW finally seeing traction.
• I'm betting we get the first real "AI-discovered drug" headline — but clinical validation remains the bottleneck, so the fundamental debate won't end.
• Self-driving labs will move from demo to production, though they'll stay narrow; differentiation shifts to whoever owns the closed-loop data flywheel, not whoever has the better algorithm.
• AI-enabled clinical trial modelling and optimization becomes a standard input to protocol, sites, and go/no-go decisions — the default, not the differentiator.
• The TechBio Series B shakeout accelerates: 2024's "bridge to nowhere" rounds become 2026's quiet shutdowns or fire-sale acquisitions, with survivors defined by true pharma revenue, not just pharma "partnerships."