Computational drug design has had a credibility problem for the better part of a decade.

Every AI biotech pitches the same thing: bigger models, more data, faster screening, higher hit rates. The drugs that actually emerge are usually conventional small molecules that could’ve been found a few different ways.

Fathom Therapeutics, formerly known as Atommap, just raised $47 million in an oversubscribed Series A with a different angle.

Their thesis?

Most computational drug design treats proteins like rigid statues. Fathom’s platform, called Microcosmos, models proteins as moving objects, atom by atom, using quantum chemistry combined with AI. Then it simulates how small molecules behave inside that motion in a cellular context.

The bet is that protein dynamics, not protein structure, is what determines whether a molecule actually binds and modulates the target.

That distinction matters for the targets nobody can drug…the so-called “undruggable proteins”. You know, the ones that don’t have nice pockets in their crystal structures because they don’t have stable structures at all.

Fathom is building Microcosmos specifically for those – molecular glues, targeted protein degraders, allosteric binders. The kind of stuff that requires you to catch the protein in a particular conformation, not just dock to a static surface.

The lead investor is the headline.

Sutter Hill Ventures doesn’t write a lot of biotech checks. They’re more famous for tech investments going back to Snowflake and Pure Storage. When Sutter Hill leads a biotech round, it’s usually a platform play with software-economics potential, not a one-shot drug.

Other participants: Chemistry, Alexandria Venture Investments, and Empire State Development’s NY Ventures. The NY money is interesting – the company is building lab capacity in New York alongside its computational team.

The competitive question is whether Microcosmos actually outperforms what Nurix, Amphista, Arvinas, and the dozens of other degrader shops are already shipping.

Fathom claims its early platform outputs include potent, selective degrader candidates generated within weeks of starting design. But, no targets, no chemical names, no quantitative selectivity data.

That’s seems to be the standard Series A platform pitch from an AI-heavy company these days, “trust us, the data is there…or will be soon”.

What’s different here is that Sutter Hill led. Fathom doesn’t need to convince a generalist biotech VC that quantum chemistry is real. They convinced one of the more sophisticated tech-software investors in the Valley that the platform has software-style scaling economics.

The proof will be the first Fathom-designed asset to enter the clinic. The company hasn’t disclosed an IND timeline yet. But $47 million buys a lot of compute and a lot of medicinal chemists, and protein dynamics is genuinely a real frontier.

Worth tracking.