Material innovation

Chemistry and Materials Discovery are Among the Hardest Problems in Applied Science. Large Quantitative Models Are Built to Solve Them.

SandboxAQ's Large Quantitative Models (LQMs) combine physics-based simulation with machine learning to accelerate chemical and materials discovery through a proprietary three-layer infrastructure.

Who We Are

SandboxAQ is a science-first technology company spun out of Alphabet in 2022. Our chemistry and materials team brings together computational chemists, materials scientists, physicists, AI engineers, and software developers from leading institutions, all working on a single, shared platform built from first principles.

Our academic advisors include world-class researchers from Stanford, Columbia, the University of Wisconsin, the University of Michigan, the University at Buffalo, NYU, and the University of Florida, spanning materials science, quantum chemistry, mechanical engineering, and applied physics.

Recent Breakthrough Research

Introducing AQVolt

Advancing battery innovation from simulation to discovery with AI-driven materials modeling, enabling safer, higher energy density next-generation cell chemistries beyond lithium-ion.

Introducing the AQCat25-EV2 Models

The first AI model family for heterogeneous catalysis achieving quantum-level accuracy for all industrially relevant atom types– at a fraction of the compute cost and time– enabling enabling all magnetic calculations such as next-generation cell chemistries beyond lithium-ion.

Record-Breaking PFAS Simulation: Quantum Chemistry at Unprecedented Scale

Partnering with AWS, Intel, and Accenture, SandboxAQ performed the first near-exact simulation of toxic PFAS bond-breaking chemistry — across more than one million vCPUs — enabling next-generation green chemistry applications.

How We Work With You

SandboxAQ offers three ways to access and deploy our LQMs

LQM-Enabled LLMs

Connect SandboxAQ's LQMs to your existing chat LLM or cloud environment via our Model Context Protocol (MCP). Run scientific queries through the interface your team already uses—no infrastructure overhaul required.

Enterprise Licensing

Deploy SandboxAQ LQMs within your own operational environment. Fine-tune models on your proprietary data and access dozens of specialized modules — from binding affinity prediction to generative molecule design — on a scalable inference model.

Frontier Partnerships

For organizations pursuing category-defining outcomes, we offer long-term, shared-risk engagements with milestone-based alignment and ongoing commercial milestones. We co-create novel IP alongside your team and we have skin in the game.

Results

95% Reduction in Battery Testing Time

Working with NOVONIX, SandboxAQ's LQMs predicted lithium-ion battery end-of-life with 35x greater accuracy using 50x less data than traditional AI models — compressing months of testing to days and lowering R&D costs while supporting next-generation energy storage commercialization.

First-Ever Near-Exact Simulation of PFAS Bond-Breaking Chemistry

Partnering with AWS, Intel, and Accenture, SandboxAQ turned the cloud into a massively distributed supercomputer — running over 1.1 million vCPUs to simulate toxic PFAS molecules at a level of accuracy previously impossible. The breakthrough opens new pathways for PFAS remediation, green chemistry, and sustainable materials design.

Largest Organometallic Catalyst Ever Computed at Near-Exact Accuracy

In collaboration with DIC and AWS, SandboxAQ simulated the largest organometallic catalyst ever computed at near-exact quantum accuracy — making high-fidelity quantum chemistry practical at industrially relevant molecular scale for the first time in the history of chemistry.

80x Speed Improvement in Complex Chemistry Simulations

Combining SandboxAQ's LQMs with NVIDIA's CUDA-accelerated Density Matrix Renormalization Group (DMRG) algorithm achieved computing speeds more than 80x faster than traditional 128-core CPUs — enabling unprecedented calculations for transition metal metalloenzymes critical to industrial catalysis, energy, and medicine.