Wedoany.com Report-Oct. 30, SandboxAQ, a Palo Alto-based technology company, announced the launch of AQCat25-EV2, a new quantitative AI model designed to accelerate catalyst discovery across industries including energy, chemical production, agriculture, automotive, and consumer goods. Built upon the AQCat25 dataset, this innovation represents a major step forward in computational catalyst development, enabling high-speed, high-accuracy research in one of the most critical areas of modern industry.

Catalysts play a vital role in the global economy, influencing the production of more than 80% of all commercial goods. However, traditional catalyst design methods are limited by slow laboratory throughput, typically analyzing fewer than 100 catalysts per week, which restricts innovation to incremental improvements. The introduction of quantitative AI models such as AQCat25-EV2 offers a way to significantly expand throughput and accelerate progress by several orders of magnitude.

Until now, AI-driven catalyst models were limited in their ability to accurately represent all industrially relevant elements. AQCat25-EV2 overcomes this limitation by incorporating the quantum effect of spin polarization, extending its predictive capabilities to cover all elements used in industrial catalysis. This makes it the first heterogeneous catalyst model that can be applied consistently and robustly across the full range of materials used in production.

The model predicts energetics with accuracy comparable to physics-based quantum-mechanical simulations but operates up to 20,000 times faster, making large-scale, high-accuracy virtual screening feasible for the first time. This efficiency removes one of the most significant barriers that has long constrained materials and catalyst innovation.

Trained using 13.5 million quantum chemistry calculations across 47,000 intermediate-catalyst systems, AQCat25-EV2 is also the first large-scale catalytic AI model to include magnetic spin polarization data—crucial for improving prediction accuracy, particularly for abundant metals like cobalt, nickel, and iron.

Dr. Bob Maughon, former Chief Technology Officer at SABIC, commented: “AQCat25-EV2 is among the first models that will allow screening in silico on a wide set of chemistries with unprecedented precision and speed, opening the door to novel catalysts and applications. For critical, unsolved industry problems, from CO2 reduction to advanced battery materials, this technology will be an indispensable tool for accelerating discovery and securing better chemical solutions.”

Aayush Singh, Head of Catalytic Sciences at SandboxAQ, added: “A wide range of industries face critical unsolved problems in catalysis today, all of which are expected to directly benefit from AQCat25-EV2. These include plastic recycling, CO2 reduction, hydrogen fuel production, methane-to-methanol conversion, and syngas-to-ethanol transformation. For these industries, we’re fundamentally de-risking the R&D process across the entire spectrum of materials science.”

Developed using NVIDIA DGX™ Cloud with over 500,000 GPU-hours on NVIDIA H100 Tensor Core GPUs, AQCat25-EV2 leverages cutting-edge computational infrastructure. SandboxAQ will adopt the NVIDIA ALCHEMI platform to expand global accessibility and performance. The model is now available on Hugging Face, accompanied by a technical paper detailing its architecture and validation results.

SandboxAQ, a B2B company specializing in AI and quantum solutions, continues to push the boundaries of scientific computing with innovations designed to enhance industrial research and accelerate technological progress.