Utah AI Supercomputer
An AI supercomputing platform for Utah
The Utah AI Supercomputer is a statewide AI resource created to provide Utah's institutions of higher education, state organizations, and commercial sector with access to advanced computing resources, training, and support. It will accelerate and support research, education, innovation, and AI workforce development across the state.
The Center for High Performance Computing (CHPC) at the University of Utah will manage the AI supercomputing platform for Utah. The supercomputer is built for national impact and will support regional AI innovation across sectors and position Utah as a leader in AI-enabled research and innovation. The platform represents a significant increase in high end GPU computing capacity at the University of Utah, fueled by $50 million in investment across public and private sectors over five years.
Enablement is a key aspect of the Utah AI Supercomputer program, ensuring that users can successfully access, learn, and apply the resource. This includes training, onboarding, outreach, and utilization support, which will be led by the CHPC in coordination with user representatives from partner organizations.
Technologies and partnerships
Systems in the Utah AI Supercomputer, which will support research and innovation across Utah
The platform has five key public–private partners: Hewlett Packard Enterprise, NVIDIA, Huntsman Family Foundation, State of Utah, and University of Utah. It is a first-of-its-kind HPE–NVIDIA–university partnership that will advance Utah as a national leader in AI research and innovation, allowing users across the state to train, apply, and scale large AI models across multiple servers with many GPUs.
The system will provide researchers and innovators in Utah with access to 33 HPE Cray XD670 nodes. Each node has eight H200 SXM5 GPUs, and each GPU has 141 GB of high-speed VRAM to support demanding AI workloads. Additionally, each node has two Intel Xeon-P 8570 56-core CPUs and 2 TB of DDR5-5600 RAM. In aggregate, the platform will have 264 GPUs, 3,696 CPU cores, 66 TB of RAM, and 1 PB of temporary scratch storage in addition to storage local to each node for high-speed input and output. Uniquely among CHPC systems, it also provides optimal multi-GPU connectivity, with 900 gigabytes per second of bidirectional GPU-to-GPU bandwidth for AI workflows that span GPUs. Additionally, each GPU has a 400 gigabit-per-second network interface to an Infiniband fabric for high-speed network traffic between nodes, supporting cutting-edge, large-scale AI research and innovation. The high-speed network connectivity allows AI projects to efficiently scale beyond a single node. The platform is also connected to the Utah Education and Telehealth Network (UETN) and the University of Utah's Downtown Data Center at 100 and 400 gigabits per second, respectively, to facilitate data movement. The platform is based on HPE GreenLake, which provides flexibility in provisioning resources to meet the needs of researchers and innovators across the state.
Initial benchmarks of the system indicate that it will be around position 150 in the TOP500 most powerful supercomputers list, putting an academic supercomputer based in Utah on the list after more than 20 years of absence. We are planning to place the system in the TOP500 list for its June 2026 refresh.
Availability and access
Early access to the statewide AI supercomputer is expected around mid-summer 2026, after the installation is completed. We are currently establishing a governance structure and creating onboarding and enablement strategies. More information on these, and the access, will become available as we get closer to the deployment time.
| The AI supercomputing platform is not suitable for sensitive information. If you have questions or concerns about the applicability or availability of the system for your projects, please contact the CHPC at helpdesk@chpc.utah.edu. |