anycloud · Jun 2026
A case study described Ångström AI’s CSP-MACE-Å work, the use of more than 100,000 GPU hours, and claims that the model outperformed Meta’s UMA-OMC on evaluation suites.
Gen AI molecular simulations that reproduce wetlab results
Company research
Public company, workplace, funding, and market signals
Updated Jul 30, 2026
Overview
Ångström AI is a Cambridge-founded, physics-first AI company building molecular simulation software for pharma and biotech. Its platform uses generative AI plus machine-learning interatomic potentials to accelerate or replace wet-lab and DFT-heavy workflows for properties like solubility, lipophilicity, hydration, binding, and crystal structure stability.
Primary product
Physics-informed generative AI molecular simulation platform for drug discovery
Founded
2024
Headquarters
San Francisco, California, United States
Team size
1-10 employees
Industry
Technology, Information and Internet
Sub-industry
Scientific software / molecular simulation
Offices
San Francisco, California, United States
167-169 Great Portland Street, 5th Floor, London W1W 5PF, United Kingdom
Business model
B2B enterprise software
Research partnerships
Pilot projects
Licensing
Funding
Stage
Seed
Total raised
$500K
Latest round
Seed VC · Sep 2024
Latest amount
$500K
Sep 2024 · Y Combinator
People & investors
Leadership
Cofounder, CEO
Cofounder, Head of AI
Gábor Csányi
Cofounder, Head of Science
Co-founder
Bálint Máté
Member of Technical Staff
Graeme Day
Advisor
Work & culture
Research-led and scientifically rigorous, with a customer-centric posture toward pharma/biotech collaboration. Public materials emphasize practical application, academic depth, and fast iteration on simulation research rather than a consumer-product culture.
Product & market
Pricing
Not publicly disclosed; public materials suggest demo-, pilot-, and partnership-led enterprise sales.
Differentiators
- —Physics-constrained generative AI instead of black-box prediction
- —MACE-based quantum-mechanically accurate simulations
- —Claims of experimental-error or subchemical accuracy on select problems
- —Orders-of-magnitude speedups versus wet lab or DFT workflows
- —Focus on pharma-relevant outputs like solubility, hydration, binding, and crystal stability
Technology
MACE (multi-atomic cluster expansion)
Machine learning interatomic potentials (MLPs)
Diffusion models
Generative AI
Alchemical free-energy methods
SE(3)-equivariant coupling flows
FAB / Boltzmann distribution sampling
Physics-constrained molecular simulation
Customers
Pharma companies
Biotech companies
Drug discovery teams
Computational chemistry teams
AstraZeneca (research collaboration)
Competitors
Schrödinger
D. E. Shaw Research
SandboxAQ
QSimulate
Fathom Therapeutics
Scale & contact
Estimated revenue
≈$900K ARR (estimated, 2024)
Estimated monthly visits
2.3K
Traffic estimate as of Jun 2026
Latest news
LinkedIn · May 2026
Laurence Midgley announced a collaboration with AstraZeneca on CSP-MACE-Å, describing the model as a DFT-replacement for crystal structure prediction and claiming a 10,000x speedup with comparable accuracy.
AI Insider · Jul 2024
A launch-profile article summarized the company’s mission, team background, MACE technology, and diffusion-model acceleration approach for drug discovery simulations.
Y Combinator
YC’s launch page introduced Ångström AI as a generative-AI molecular simulation company for pharma and biotech, highlighting the team’s Cambridge background and its experimental-accuracy claims.
Jobs and careers at Ångström AI
0 jobs at Ångström AI
No open positions
Check back later for new openings