SandboxAQ AQChemSim

paid

SandboxAQ AQChemSim accelerates materials and chemistry discovery using Large Quantitative Models that combine physics-based simulation with machine learning for batteries, catalysts, and sustainable materials.

Data & Analytics

AI Models & Infrastructure

Research & Education

About

SandboxAQ AQChemSim is an advanced computational chemistry and materials discovery platform built on proprietary Large Quantitative Models (LQMs) — a three-layer infrastructure combining physics-based simulation, quantum chemistry, and machine learning. Developed by a science-first team spun out of Alphabet in 2022, AQChemSim enables enterprises to explore vastly larger candidate spaces, compress experimental testing timelines, and de-risk high-stakes R&D programs. The platform covers critical application domains including next-generation battery chemistries (AQVolt), heterogeneous catalysis (AQCat25-EV2), PFAS bond-breaking simulation, and organometallic catalyst design. Its models achieve near-exact quantum accuracy while running at a fraction of the compute time and cost of conventional approaches. Organizations can access AQChemSim through three deployment modes: LQM-enabled LLM integration via Model Context Protocol (MCP) for seamless chat-based scientific queries; Enterprise Licensing for on-premise deployment with fine-tuning on proprietary data; and Frontier Partnerships for co-development of novel IP with milestone-based alignment. Proven results include a 95% reduction in battery testing time with NOVONIX and the first-ever near-exact simulation of PFAS molecules across 1.1 million vCPUs in partnership with AWS, Intel, and Accenture. AQChemSim is purpose-built for computational chemists, materials scientists, and R&D engineers at industrial and pharmaceutical organizations seeking a decisive scientific edge.

Key Features

Large Quantitative Models (LQMs): Proprietary three-layer infrastructure combining physics-based simulation and machine learning to achieve near-exact quantum chemistry accuracy at a fraction of conventional compute cost.
Domain-Specific AI Models: Specialized modules including AQVolt for battery innovation and AQCat25-EV2 for heterogeneous catalysis, covering all industrially relevant atom types including magnetic calculations.
Flexible Deployment Options: Access via MCP-connected LLM chat interfaces, enterprise on-premise licensing with proprietary data fine-tuning, or long-term frontier co-development partnerships.
Massively Scalable Simulation: Capable of running distributed quantum chemistry simulations across over 1.1 million vCPUs, enabling unprecedented-scale molecular modeling previously impossible with standard infrastructure.
End-to-End R&D Acceleration: From candidate screening to binding affinity prediction and generative molecule design, the platform compresses months of physical testing into days while lowering R&D costs.

Use Cases

Accelerating battery electrolyte and cathode material screening to identify next-generation cell chemistries beyond lithium-ion with 95% faster testing cycles.
Simulating heterogeneous catalyst behavior at quantum accuracy to optimize industrial chemical processes including ammonia synthesis, Fischer-Tropsch, and oxidation reactions.
Modeling PFAS 'forever chemical' bond-breaking pathways to support green chemistry remediation strategies and regulatory compliance R&D.
Predicting binding affinities and molecular properties for pharmaceutical materials discovery, reducing costly wet-lab iterations.
Integrating LQM-powered scientific reasoning into existing enterprise LLM chat environments via MCP for on-demand molecular and materials queries.

Pros

Proven Industrial Results: Demonstrated 95% reduction in battery testing time with NOVONIX and world-first PFAS simulations, validating real-world R&D impact beyond academic benchmarks.
Quantum-Level Accuracy at Scale: AQCat25-EV2 achieves quantum-chemistry accuracy for all industrially relevant atom types at a fraction of the compute cost, making high-fidelity simulation economically viable.
Flexible Access Models: Three deployment tiers — MCP integration, enterprise licensing, and frontier partnerships — accommodate teams from early evaluation to deep co-development.
Elite Scientific Pedigree: Team and advisors drawn from Stanford, Columbia, MIT-adjacent institutions, and Alphabet; ensures models are grounded in rigorous computational chemistry and materials science.

Cons

Enterprise-Only Pricing: No self-serve free tier or transparent public pricing; access requires direct engagement with the SandboxAQ sales team, creating a barrier for smaller teams or individual researchers.
Narrow Domain Focus: Purpose-built for chemistry and materials science; organizations outside batteries, catalysts, or molecular simulation will find limited applicability.
Integration Complexity: Enterprise licensing and frontier partnership tiers may require significant IT and scientific infrastructure investment for on-premise deployment and model fine-tuning.

Frequently Asked Questions

LQMs are SandboxAQ's proprietary AI models that combine physics-based simulation with machine learning across a three-layer infrastructure. Unlike pure data-driven models, LQMs are grounded in quantum chemistry and materials physics, enabling near-exact accuracy for molecular and materials predictions.

SandboxAQ offers three access modes: (1) LQM-enabled LLM integration via Model Context Protocol for chat-based scientific queries in existing environments; (2) Enterprise Licensing for on-premise deployment and fine-tuning on proprietary data; (3) Frontier Partnerships for long-term co-development with milestone-based commercial alignment.

AQChemSim is primarily suited for energy storage (batteries), industrial catalysis, PFAS and green chemistry, pharmaceutical materials, and sustainable materials R&D. Any organization performing high-throughput screening or computational materials characterization stands to benefit.

Key results include a 95% reduction in battery testing time with NOVONIX (35x more accurate predictions using 50x less data), the first near-exact PFAS bond-breaking simulation across 1.1M vCPUs with AWS/Intel/Accenture, and the largest organometallic catalyst ever computed at near-exact quantum accuracy with DIC and AWS.

AQChemSim uses proprietary Large Quantitative Models developed by SandboxAQ's in-house team with academic advisors from leading universities. While some breakthrough research is published, the core commercial platform is proprietary and accessed via licensing or partnership agreements.