EvolutionaryScale is putting new funding behind a large bet on AI for biology. The startup said on Tuesday that it raised $142 million in a seed round and released ESM3, a model built to generate novel proteins for scientific research.
The company describes ESM3 as a “frontier model” for biology. Its stated goal is to help create proteins that could be useful in drug discovery and materials science.
What EvolutionaryScale Announced
The seed round was led by ex-GitHub CEO Nat Friedman, Daniel Gross and Lux Capital. Amazon and NVentures, Nvidia’s corporate venture arm, also participated.
Alongside the financing, EvolutionaryScale introduced ESM3. The company says the model can work across protein sequence, structure and function, giving it the ability to generate new proteins.
EvolutionaryScale’s co-founder and chief scientist, Alexander Rives, framed the release as part of a broader shift in how biology could be engineered. “ESM3 takes a step toward a future of biology where AI is a tool to engineer from first principles, the way we engineer structures, machines, and microchips and write computer programs,” Rives said in a statement.
The company is making the full 98-billion-parameter model available for non-commercial use through its cloud Forge developer platform. It is also releasing a smaller version of the model for offline use.
Why Protein Design Is Difficult
Proteins matter because studying them can reveal mechanisms behind disease, including possible ways to slow or reverse it. Creating proteins can also open the door to new classes of drugs, tools and therapeutics.
The challenge is that designing proteins in the lab is expensive in both computational and human resource terms. A researcher must first imagine a structure that could plausibly perform a task inside the body or inside a product.
Then comes another hard step: finding a protein sequence likely to fold into that structure. A protein sequence is the sequence of amino acids that make up the protein.
Folding is central to the problem. Proteins must correctly fold into three-dimensional shapes in order to perform their intended function.
ESM3 was trained on a dataset of 2.78 billion proteins. Rives says it can “reason over” protein sequence, structure and function, which is the basis for the company’s claim that it can generate new proteins.
What ESM3 Has Already Produced
EvolutionaryScale says it used ESM3 to generate a new variant of green fluorescent protein, or GFP. GFP is responsible for the glowing of jellyfish and luminescent colors in coral.
The company has a preprint paper on its website detailing that work. Rives also described the release as an invitation to researchers: “We’ve been working on this for a long time, and we’re excited to share it with the scientific community and see what they do with it,” he said.
The availability plan reflects two audiences. Non-commercial users can access the full model through Forge, while others may work with smaller offline versions or future cloud integrations.
How The Business Could Work
EvolutionaryScale is not presenting ESM3 only as a research release. The company, which employs roughly 20 people, told TechCrunch that it expects to make money through partnerships, usage fees and revenue sharing.
One possible path is working with pharmaceutical companies to integrate ESM3 into their workflows. Another is revenue sharing with researchers if discoveries made with ESM3 are commercialized.
The model is also headed toward major cloud and enterprise AI channels. EvolutionaryScale says it will soon bring ESM3 and its derivatives to select AWS customers through SageMaker, Bedrock and HealthOmics.
ESM3 will also be available to select customers using Nvidia’s NIM microservices, supported with an Nvidia enterprise software license. EvolutionaryScale says both AWS and Nvidia customers will be able to fine-tune ESM3 using their own data.
The Long Bet Behind The Model
EvolutionaryScale’s work began before the company existed. Rives, Tom Sercu and Sal Candido started developing generative AI models for protein exploration while at Meta’s AI research lab, FAIR, in 2019.
After their team was disbanded, Rives, Sercu and Candido left Meta and continued the work. The new company is now trying to scale that research into a general-purpose AI model for biotech applications.
That ambition may take time. In a pitch deck that Forbes managed to obtain last August, EvolutionaryScale repeatedly emphasized that it could take a decade for generative AI models to help design therapies.
The company also faces competition. The field includes DeepMind’s spin-off, Isomorphic Labs, which already has contracts with big pharma companies, as well as Insitro, publicly traded Recursion and Inceptive.
EvolutionaryScale’s central argument is that biology can benefit from the same forces driving broader AI progress: larger models, larger datasets and more computational power. An EvolutionaryScale spokesperson said, “The incredible pace of new AI advances is being driven by increasingly large models, increasingly large datasets and increasing computational power. The same holds true in biology. In research over the last five years, the ESM team has explored scaling in biology. We find that as language models scale, they develop an understanding of the underlying principles of biology, and discover biological structure and function.”
For now, the company has funding, a released model and planned distribution through major platforms. The larger question is how quickly protein-generating AI can move from promising research tool to practical engine for new therapeutics, materials and biological discovery.