SandboxAQ, a company using quantum-based algorithms and simulations to train AI models, is expanding its existing 17-month-old partnership with Nvidia, a move which comes just days after Nvidia and Google were among the participants in SandboxAQ’s $450 million Series E funding round.
The next phase of the collaboration involves the partners leveraging the Nvidia DGX Cloud AI platform on Google Cloud “to build a state-of-the-art Large Quantitative Model (LQM) platform” to help fuel AI-driven scientific discovery across the fields of biopharma, chemicals, advanced materials, financial services, cybersecurity, navigation, and medical imaging.
“This is all new and includes us leveraging next-gen compute over time,” said a SandboxAQ spokesperson via e-mail regarding the use of the DGX Cloud platform. “Sandbox is building on the latest and greatest as part of the collaboration.” The company did not respond to a question about whether or not systems using Nvidia’s newest Blackwell GPUs would eventually play a role in the collaboration.
SandboxAQ is a member of the NVIDIA Inception program for startups, and the expanded collaboration also represents the next step for SandoxAQ’s LQMs. “This marks a meaningful level of alignment between SandboxAQ and Nvidia that establishes LQMs as a key part of the next era of AI–AI for the physical world–and helps position SandboxAQ as a leader in terms of compute and development,” the spokesperson stated.
LQMs are not to be confused with LLMs. Nadia Harhen, general manager of AI simulation at Palo Alto, California-based SandbxAQ, told Fierce Electronics last year, “LQMs are trained on large volumes of proprietary, high accuracy physics-based data to predict real-world properties of drugs and materials and to generate novel chemical compounds for different applications. LQM’s are founded on a paradigm of equation-based computing used to generate accurate training data for physics-aware ML models, which are then used to speed up simulation and prediction of physical quantities, creating a virtuous cycle of accelerated data generation as a result.”
Last August, SandboxAQ and Nvidia said they had used LQMs and Nvidia’s CUDA-accelerated Density Matrix Renormalization Group (DMRG) quantum simulation algorithm on Nvidia H100 GPUs to achieve “an 80x acceleration in quantum chemistry calculations… enabling accurate simulation of enzyme active sites and complex catalysts previously impossible due to computational limitations.”
Now, matching LQMs and DGX Cloud, the partners will be able to demonstrate how quantum-based simulations can speed up drug discovery cycles by up to 4x. Replacing “slow, resource-intensive design-make-test cycles with high-performance, equation-based simulations” will reduce discovery timelines from months to weeks,” SandboxAQ said in a statement. “Enhanced modeling capabilities support simultaneous optimization across multiple parameters, enabling faster validation of promising candidates and accelerating breakthroughs with greater confidence.”
SandboxAQ is generating high-fidelity scientific datasets, curated by DGX Cloud, that combine chemical and biological simulations. “These methods leverage equation-based LQM models to reveal interactions between small molecules and complex biological targets that were previously difficult to detect including conformer libraries for generative chemistry and synthetic affinity data for training predictive models,” according to the statement. “By powering causal knowledge graphs and more accurate molecular design, these datasets reduce false positives and improve success rates across the R&D pipeline.”
Also central to the collaboration is SandboxAQ’s new Agentic AI Chemist, which the statement said “combines and orchestrates multiple LQMs to transform the scale of the research and development process. It autonomously explores millions of potential chemical pathways, far beyond what a human chemist could evaluate, enabling the discovery of novel molecules and the optimization of compounds for clinical and scale up success.”
The partners also recently published a joint research paper, "Orbital Optimization of Large Active Spaces via AI-Accelerators," that details how their researchers successfully performed orbital optimization on a system with 82 electrons in 82 orbitals, more than doubling the size of simulations compared to previous works. The companies said this breakthrough allows GPU-accelerated quantum chemistry calculations to push the previous boundaries of molecular simulations, with potential implications in catalysis, material science and high-dimensional parameter optimization.
“Our expanded work with Nvidia accelerates our customers’ ability to innovate and lead in their fields,” said Jack Hidary, CEO of SandboxAQ. “By developing our platform on Nvidia DGX Cloud and continuing our research collaboration, SandboxAQ will deliver a level of performance and insight that gives our customers a clear edge in accelerating innovation.”
“SandboxAQ is pushing the boundaries of AI-native science," said Alexis Bjorlin, Vice President of Nvidia DGX Cloud. "Nvidia DGX Cloud provides an AI development platform with essential scale and optimized application performance, empowering SandboxAQ to deliver cutting-edge capabilities and drive real-world impact for organizations tackling society's most critical challenges.”
Meanwhile, on the funding front, the recent investment round pushed SandboxAQ’s total funding over the last three years to more than $950 million. Nvidia and Google were joined in the round by fellow new investors, including hedge fund billionaire Ray Dalio, investment firm Horizon Kinetics, and BNP Paribas. They join a long list of luminaries that already have invested in the company, including Breyer Capital, former Google CEO Eric Schmidt, financial firms Alger, Paladin Capital, S32, and TIME Ventures, and funds and accounts advised by T. Rowe Price Associates, Inc. SandboxAQ was started in 2016 by Hidary as a project inside Google parent firm Alphabet before being spun out of the company in March 2022.