Isomorphic about to raise $2bn - and why I'm not sure that's a good thing

TL;DR

Isomorphic Labs is in advanced talks to raise more than $2B, with Thrive Capital leading and Alphabet participating. I see real upside — capital at platform scale closes the gap between AI demos and actual drug programs — but the deeper question is about timelines and patience. Every frontier AI lab is now running some version of this bet, and Google/Isomorphic is the furthest along the full drug-development timeline, making this round a useful test case. The binding constraint in AI pharma has already moved downstream from discovery, and almost nobody is funding that side of the equation at $2B scale.

Isomorphic Labs is in advanced talks to raise more than $2B, with Thrive Capital leading and Alphabet participating. Bloomberg broke it last Friday.

I am of two minds about this round.

The upside is real. Any week where billions flow into AI-first techbio is good for the field. Compute, wet lab, and clinical execution are all expensive. Capital at platform scale closes part of the gap between AI demos and actual drug programs.

The deeper question is about timelines

In the rest of tech, frontier AI labs ship something new every week. Models, modalities, product surfaces, pricing. Feedback loops run in days.

Drug development runs on a different clock. Five to ten years from preclinical candidate to a credible Phase 2 readout is still a very good outcome. AI can compress discovery and preclinical work. Phase 1, Phase 2, and Phase 3 remain bound to biological time.

Isomorphic has already pushed the expected start of its first human trials from end of 2025 to end of 2026. The clinical clock is doing what it has always done.

This round is one data point in a larger industry pattern

Most frontier AI companies are now running some version of this bet. DeepMind built Isomorphic. Anthropic acquired Coefficient Bio and launched Claude for Life Sciences. OpenAI is partnered with Moderna and Color Health, and sits close to Retro Biosciences through Altman's personal investment. Microsoft AI for Science sits alongside Microsoft's long-running Novartis relationship. Nvidia and Lilly are co-investing up to $1B in an AI co-innovation lab.

Google/Isomorphic is the cleanest live case because it is furthest along the full drug-development timeline.

That makes this round a useful test of the question every one of these bets will eventually face.

Will tech capital and tech boards stay patient through a decade of clinical iteration when their other businesses compound weekly?

Some will. Hassabis has steered Google toward biology timelines for years, and $2B of runway suggests Alphabet is bought in. Not every lab has a CEO with that conviction.

There is a second-order issue worth flagging

Many AI-first techbio models still seem to underweight the cost, friction, and managerial complexity of clinical execution. The binding constraint in AI pharma has already moved downstream from discovery.

If every frontier AI lab succeeds at producing more candidates, that downstream constraint tightens further. Trial sites are not elastic. Skilled investigators are not elastic. FDA reviewer hours are not elastic. Patient recruitment for novel mechanisms in narrow indications is already one of the hardest parts of any program.

If candidate output doubles by 2028, the bottleneck stops being model quality and starts being the clinical pipeline itself.

Almost nobody is funding that side of the equation at $2B scale.

The smarter $2B in AI pharma right now might be the one that builds capacity for the candidates discovery has already produced.

Key takeaways

• Any week where billions flow into AI-first techbio is good for the field — capital at platform scale closes part of the gap between AI demos and actual drug programs.
• Drug development runs on a different clock than tech; five to ten years from preclinical candidate to a credible Phase 2 readout is still a very good outcome, and AI cannot compress Phase 1, 2, or 3.
• Isomorphic has already pushed the expected start of its first human trials from end of 2025 to end of 2026 — the clinical clock is doing what it has always done.
• The real test is whether tech capital and tech boards will stay patient through a decade of clinical iteration when their other businesses compound weekly.
• The binding constraint in AI pharma has already moved downstream from discovery, and if candidate output doubles by 2028, the bottleneck stops being model quality and starts being the clinical pipeline itself.
• Trial sites, skilled investigators, FDA reviewer hours, and patient recruitment for novel mechanisms are not elastic — almost nobody is funding that side of the equation at $2B scale.
• The smarter $2B in AI pharma right now might be the one that builds capacity for the candidates discovery has already produced.