SambaNova Systems has secured deployments at three major U.S. Department of Energy (DOE) labs—Argonne National Laboratory, Oak Ridge National Laboratory (ORNL), and the Texas Advanced Computing Center (TACC)—to accelerate artificial intelligence (AI) adoption in scientific research. These facilities are integrating SambaNova Suite, powered by the DataScale SN40L systems and the Composition of Experts (CoE) AI model framework, to boost their AI infrastructure and enable fast, energy-efficient inferencing.
Argonne is expanding its AI resources by deploying SambaNova systems as part of the Argonne Leadership Computing Facility’s AI Testbed. This installation includes 16 Reconfigurable DataFlow Units (RDUs) to support large-scale foundation models such as AuroraGPT, which is being developed for autonomous scientific exploration. ORNL, known for hosting the Frontier supercomputer, plans to use SambaNova Suite for parallel inferencing across multiple scientific models to improve prediction accuracy while reducing energy costs. TACC, home to the Frontera supercomputer, will deploy the platform for inferencing AI models trained on its existing systems, seamlessly integrating them into scientific workflows.
These partnerships highlight the role of advanced AI in reshaping scientific discovery across fields like climate modeling, drug discovery, and materials science. “Inferencing large language models and foundation models are crucial to our efforts to apply AI to complex scientific problems,” said Rick Stevens, Associate Laboratory Director at Argonne. “With SambaNova’s platform, we are creating tools that enable a broader scientific community to harness AI’s potential.”
• Key Points:
• Argonne: Deploying SambaNova Suite with 16 RDUs to advance AuroraGPT and expand the ALCF AI Testbed.
• ORNL: Using SambaNova’s fast inference capabilities to enhance scientific predictions across domains.
• TACC: Leveraging SambaNova’s CoE framework for integrating AI inference into diverse scientific workflows.
• Efficiency: Focus on reducing energy consumption and scaling AI capabilities for large workloads.
• Applications: AI-driven advancements in areas like climate science, brain mapping, and autonomous scientific exploration.
- SambaNova Systems, established in 2017 in Palo Alto, California, specializes in developing enterprise-scale AI technologies. Founded by industry experts from Sun/Oracle and Stanford University, including CEO Rodrigo Liang and Chief Technologist Kunle Olukotun, the company’s mission is to enable organizations to harness the power of AI through integrated hardware and software solutions. SambaNova has secured over $1.1 billion in funding from investors such as SoftBank Vision Fund 2, BlackRock, and Intel Capital. In 2024, the company made significant strides by introducing Samba-1, a one trillion-parameter generative AI model for enterprise use, and launching the SN40L AI chip, designed to handle models up to five trillion parameters. Additionally, SambaNova announced partnerships with organizations like Analog Devices and welcomed semiconductor industry veteran Lip-Bu Tan as Executive Chairman.