What is AI protein design?

AI protein design uses machine learning models -- particularly protein language models and generative models -- to design new proteins with desired functions, structures, or binding properties. Instead of relying on natural evolution or random mutagenesis, AI systems like ESM-3 (EvolutionaryScale), RFdiffusion (Baker Lab / UW), and AlphaFold (DeepMind) learn the grammar of protein sequences and structures from billions of natural proteins, then generate novel proteins that do not exist in nature. This enables design of therapeutic antibodies, industrial enzymes, biosensors, and biomaterials orders of magnitude faster than traditional methods.

Can AI design new proteins?

Yes. AI can now design functional proteins that have never existed in nature. In 2024-2025, several breakthroughs demonstrated this: EvolutionaryScale's ESM-3 generated functional green fluorescent proteins with no natural homologs. David Baker's lab at UW (RFdiffusion) designed novel protein binders that were experimentally validated. Absci used generative AI to design de novo therapeutic antibodies that passed wet-lab validation. Generate Biomedicines designed novel protein therapeutics from their generative platform. The field has moved from structure prediction (AlphaFold solved this) to generative design -- creating proteins with specified functions from scratch.

ESM-3 is a multimodal protein language model developed by EvolutionaryScale (founded by former Meta AI researchers). It was trained on billions of protein sequences and structures and can simultaneously reason about sequence, structure, and function. ESM-3 made headlines in 2024 when it generated a functional green fluorescent protein (GFP) with extremely low sequence identity to any known natural protein -- demonstrating genuine de novo protein creation. EvolutionaryScale raised $142M in what was the largest seed round in synbio history. The company has since released ESM-4 with 15B parameters and is commercializing its models for drug design and industrial applications.

Which companies use AI for drug design?

Major companies using AI for protein-based drug design include: EvolutionaryScale (protein language models ESM-3/ESM-4), Absci Corporation (NASDAQ: ABSI -- generative AI antibody design with wet-lab validation via SoluPro), Generate Biomedicines (Flagship Pioneering -- generative protein therapeutics), Recursion Pharmaceuticals (NASDAQ: RXRX -- AI-driven drug discovery at scale), Cradle (generative AI for protein engineering), Arzeda (computational enzyme design), and Insilico Medicine (AI for target discovery and drug design). Academic groups include David Baker's lab at UW (RFdiffusion, ProteinMPNN), DeepMind (AlphaFold), and Meta FAIR (ESMFold). The field is growing rapidly with over $3B invested in AI+bio startups in 2024-2025.

AI Protein Design: Every Company Using AI to Engineer Proteins (2026)

The convergence of AI and protein engineering is the fastest-growing vertical in synthetic biology. From foundation models like ESM-3 to generative antibody design, these 8 companies are using machine learning to design proteins that nature never created. This page covers every major AI protein design company, their models, funding, and key results.

Companies

Publicly Traded

Sector Funding

$3B+

Last Updated

Mar 2026

AI Protein Design Companies8 companies

Company	Category	Model / Approach	Stage	Funding	Key Result
EvolutionaryScale	Foundation Models	ESM-3 / ESM-4 (15B params)	Commercial API	$142M+ (largest synbio seed)	Generated novel functional GFP with no natural homologs
Absci Corporation	AI Drug Discovery	Proprietary generative antibody models	Revenue (Public)	$500M+ (NASDAQ: ABSI)	De novo antibody design validated in wet lab; pharma partnerships
Generate Biomedicines	Generative Therapeutics	Chroma (protein diffusion model)	Clinical Pipeline	$700M+	Designing novel protein therapeutics from scratch across modalities
Cradle	Protein Optimization	Proprietary protein optimization models	Commercial Platform	$103M+	Reduces protein engineering cycles from months to weeks for biotech teams
Arzeda	Enzyme Design	Physics-based + ML enzyme design	Revenue Stage	$70M+	Designed novel enzymes for food ingredients and specialty chemicals
Recursion Pharmaceuticals	AI Drug Discovery	Phenomics + chemical + genomics ML platform	Revenue (Public)	$600M+ (NASDAQ: RXRX)	Largest proprietary biological dataset; multiple clinical candidates
Insilico Medicine	AI Drug Discovery	Chemistry42 + Biology42 + PandaOmics	Phase 2 Clinical	$400M+	First fully AI-designed drug (INS018_055) to reach Phase 2 trials
Profluent	AI Gene Editing	OpenCRISPR-1 (AI-designed gene editor)	Research / Pre-Clinical	$35M+	First AI-generated gene editor (OpenCRISPR-1) -- functional in human cells

THE BOTTOM LINE

AI protein design has moved from academic curiosity to commercial reality. The field crossed a critical threshold in 2024-2025 when multiple companies demonstrated functional de novo proteins validated in wet-lab experiments -- not just computational predictions.

The competitive landscape is stratified: foundation model companies (EvolutionaryScale) are building horizontal platforms analogous to OpenAI for proteins. Vertically integrated companies (Absci, Generate Bio) combine AI design with their own wet-lab validation. Optimization platforms (Cradle) serve existing biotech teams. And established drug companies (Recursion) are scaling AI across the full discovery pipeline.

The key question for the field is whether protein language models will commoditize as quickly as text LLMs, or whether proprietary training data and wet-lab integration create durable moats. The answer matters enormously -- it determines whether the value accrues to model builders or to companies that integrate AI with physical biology infrastructure.

Frequently Asked Questions// AI protein design

RELATED INTELLIGENCE