NVIDIA introduced a suite of AI-accelerated software tools that compress materials discovery timelines from weeks to days, allowing researchers to simulate millions of molecular structures simultaneously and process petabytes of scientific data in real time. The new tools, unveiled at the ISC High Performance conference in Hamburg, represent a shift in how scientists approach chemistry and materials research by combining GPU acceleration with specialized algorithms designed for the lab bench and the telescope.

What Are These New Materials Discovery Tools?

NVIDIA released three interconnected software systems designed to speed up different stages of scientific research. ALCHEMI is a collection of domain-specific microservices and a toolkit for accelerating chemical and materials discovery, with applications across battery materials, catalysts, OLED displays, and other industrial applications. DAQIRI, short for Data Acquisition for Integrated Real-time Instruments, is a high-performance networking library that streams data from fast detectors and sensors into NVIDIA software without dropping information. cuPhoton is a reference code for extracting insights from multidimensional data collected from telescopes, X-rays, and laser experiments.

The performance gains are substantial. In early access testing, cuPhoton accelerated the loading and reading of astronomical images collected by the Rubin Observatory's Legacy Survey of Space and Time by 14,900 times, and enabled up to 8,400 times faster signal processing using 32 NVIDIA Grace Blackwell superchips. These speedups mean that observatories can extract insights from the LSST camera, the largest digital camera ever built, far more quickly than traditional methods allowed.

How Are Companies Using These Tools to Speed Up Discovery?

Lila Sciences, a company building an autonomous lab platform for life sciences and materials research, collaborated with NVIDIA to demonstrate real-world applications of ALCHEMI. The company accelerated high-throughput materials screening by 50 times using the ALCHEMI batched geometry relaxation microservice, which identifies stable molecular candidates that have higher chances of being successfully synthesized. For shortlisted candidates, Lila Sciences then accelerated the calculation of magnetic properties by 30 percent using the ALCHEMI VASP microservice in early access.

The speedups compound across the research pipeline. ALCHEMI's specialized computational kernels for TensorNet, a machine learning model for atomic interactions, gave Lila Sciences a 6 times speedup in training and inference while reducing memory usage by 3 times. This meant simulations that previously took weeks could be completed in just days, fundamentally changing how researchers approach materials discovery.

"The work showcases using a powerful computing stack assembled to accelerate discovery at a scale no individual scientist could achieve alone," said Andy Beam, cofounder and chief technology officer of Lila Sciences.

Andy Beam, Cofounder and Chief Technology Officer at Lila Sciences

How to Leverage GPU Acceleration for Materials Research

• Batched Geometry Relaxation: Use ALCHEMI's BGR microservice to find the most stable structures of millions of molecules at once, rather than simulating them sequentially, reducing screening time from weeks to days.
• Real-Time Data Streaming: Deploy DAQIRI to handle continuous data streams from laboratory instruments and detectors without losing information, enabling AI analysis of signals that would normally be rejected due to storage constraints.
• Molecular Dynamics Simulation: Apply ALCHEMI's batched molecular dynamics microservice to simulate how millions of materials move and behave over time, accelerating the discovery of candidates with desired properties.
• Custom Workflow Development: Use the ALCHEMI Toolkit to train AI surrogate models called machine learning interatomic potentials and build custom, high-performance atomistic simulation workflows tailored to specific research questions.

Why Does Real-Time Data Processing Matter for Science?

Traditional scientific instruments often produce data faster than systems can save it, forcing researchers to discard over 99 percent of collected information due to storage constraints. DAQIRI solves this bottleneck by handling data streams as they arrive, enabling researchers to run artificial intelligence analysis in real time on collision data from experiments like CERN's ATLAS Experiment. A research project called A-GHOST, developed by scientists from CERN, the University of Chicago, and University College London, uses DAQIRI to catch potentially interesting signals that would otherwise be lost forever.

For astronomy, the implications are equally significant. Researchers at Princeton University and Harvard University collaborated with NVIDIA to develop cuPhoton and will use it to process and analyze massive datasets collected from observatories and dark energy surveys. The ability to load, process, analyze, and visualize petabytes of data in real time transforms how scientists extract insights from the universe's largest digital camera.

What Hardware Powers These Accelerated Workflows?

NVIDIA also announced the Vera Rubin platform, a rack-scale supercomputer designed to unite high-precision simulation, AI, and data analytics for scientific discovery. The Vera Rubin platform combines NVIDIA Rubin GPUs and NVIDIA Vera CPUs connected via high-speed NVLink-C2C and other interconnects in a direct liquid-cooled architecture. With more than 7 exaflops of AI computing power, 5 petaflops of native double-precision floating-point performance, and up to 144 GPUs per rack, a single Vera Rubin system delivers performance comparable to systems on the TOP500 list of the world's most powerful supercomputers.

Leading research institutions are adopting Vera Rubin to build next-generation supercomputers. The Leibniz Supercomputing Centre will deploy Blue Lion, powered by Vera Rubin and scheduled to come online in 2027, delivering approximately 30 times the computing power of the center's current system. The National Energy Research Scientific Computing Center is building Doudna, a Dell Technologies system powered by Vera Rubin, for large-scale HPC workloads, AI training and inference, and data-intensive workflows across molecular dynamics, high-energy physics, fusion energy, materials science, drug discovery, and astronomy. Los Alamos National Laboratory has selected Vera Rubin and Vera CPU for its Mission, Vision, and Veritas systems, with Vision specifically designed to advance open science research including foundation models, agentic AI, and complex simulations spanning materials science, nuclear energy, and fusion energy.

Global system manufacturers including Bull, Dell Technologies, GIGABYTE, HPE, and Supermicro are bringing Vera Rubin NVL4 systems to market through direct liquid-cooled AI and HPC racks, with availability expected in the fourth quarter of 2026. The ALCHEMI NIM microservice for VASP, a widely used Vienna Ab initio Simulation Package, is expected to be available later in the summer, achieving a 3 times speedup for geometry optimization, the process of finding the most stable arrangement of atoms in a material.

These tools represent a fundamental shift in how materials science operates. Rather than running one experiment at a time, researchers can now evaluate multiple materials simultaneously in GPU memory, generalizing approaches across materials discovery, energy applications, and electromagnetics research. The combination of specialized software, accelerated hardware, and real-time data processing capabilities creates a new paradigm for scientific discovery at scale.