The frontier protein language model lab — simulating 500 million years of evolution
HQ: New York, NY, United States | Founded: 2023 | Employees: ~50 (2026, est.) | Stage: Early Stage VC (Seed/Series A) | Website: https://www.evolutionaryscale.ai
EvolutionaryScale was founded in 2023 by Alexander Rives (CEO), Tom Sercu, and Salvatore Candido — the core research team behind ESM (Evolutionary Scale Modeling), the protein language model family developed at Meta AI that showed language-model-style training could capture the grammar of proteins across billions of sequences. When the team left Meta, they raised the largest seed round in AI-biology history — more than $142M, led by Nat Friedman and Daniel Gross with Lux Capital, Amazon, and NVIDIA — to build the next generation of frontier models for biology.\n\nThe company emerged from stealth in June 2024 with ESM3, a frontier generative model that reasons simultaneously over protein sequence, structure, and function — a genuine advance: ESM3 designed esmGFP, a novel variant of green fluorescent protein, which the team described as simulating 500 million years of evolution with chain-of-thought prompting. The paper and model became one of the most-discussed AI-biology results of 2024. In late 2024 the company released ESM Cambrian, a significantly larger and more capable model family covering proteins and small molecules, and in 2025 it positioned the 'esmOS' vision — an AI operating system for biology with agents that design, simulate, and validate experiments.\n\nEvolutionaryScale is a public benefit company that releases open versions of its models alongside commercial tiers, partnering with AWS (ESM available on Amazon Bedrock and SageMaker) and engaging pharma and industrial biotech partners. With the ESM lineage at its core, the largest seed in the category, and a research velocity that produced two frontier model generations in a year, EvolutionaryScale is the leading independent frontier lab for generative biology — the protein-language counterpart to the frontier LLM labs.
EvolutionaryScale is the purest bet on the 'ChatGPT moment for biology': take the protein language model research pioneered at Meta (ESM), spin it out, and scale it into frontier generative models that scientists can prompt to design new proteins. Its ESM3 model stunned the field by generating esmGFP — a functional fluorescent protein the company says represents 500 million years of simulated evolution — and ESM Cambrian extended the frontier into multimodal biology. The $142M seed (led by Nat Friedman and Daniel Gross, with Amazon and NVIDIA) is the largest ever for an AI-biology startup, and the team's position as the independent frontier lab for protein models makes it a prime platform for drug discovery, materials, and enzyme design. The bull case is that generative biology becomes one of AI's biggest markets; the bear case is that the field is young, revenue is unproven, and rivals (Isomorphic Labs, Xaira, Big Pharma) are better capitalized.
Browse: Home | Blog | About | All Startups A-Z | View full profile on StartupWiki