Recursion announced the completion of BioHive-2, Recursion?s new NVIDIA DGX SuperPOD AI supercomputer, powered by 63 DGX H100 systems with a total of 504 NVIDIA H100 Tensor Core GPUs interconnected by NVIDIA Quantum-2 InfiniBand networking. This NVIDIA-powered AI supercomputer results in four times faster speeds than Recursion?s original supercomputer, BioHive-1, in benchmark performance tests. Based on available data, BioHive-2 is a fast supercomputer wholly owned and operated by any pharmaceutical company worldwide.

Recursion has demonstrated the importance of scaled computation by developing new foundation models like Phenom-1, a deep-learning model designed to extract biologically meaningful features from images of cells. As Recursion increased the size of the training data and the number of model parameters, the model?s performance increased, demonstrating a need for ample compute to train larger models. The experimentation and training to produce Phenom-1 required several months of computational time on BioHive-1. With BioHive-2, multiple ambitious AI projects of similar or greater size could be executed in parallel in shorter timeframes, enabling both the Recursion and Valence Labs teams to push on the frontier of AI in drug discovery and unlock the latent value of Recursion?s data.

A smaller model similar to Phenom-1, called Phenom-Beta, has been released for external use on the NVIDIA BioNeMo platform. The Phenom series is just one of several different models Recursion has developed to accelerate the drug discovery process using biological, chemical, and real-world patient data. Phenom-Beta is the first third-party model to be made available on the BioNeMo platform.