UT eclipses 5,000 GPUs to extend open-source AI leadership and U.S. research capacity
Nov 17, 2025 - Round Rock, Texas
The University of Texas at Austin has crossed 5,000 advanced NVIDIA GPUs across its academic and research facilities, adding more than 4,000 new NVIDIA Blackwell architecture chips integrated on Dell PowerEdge servers. The systems will anchor Horizon, the National Science Foundation's Leadership-Class Computing Facility (LCCF) supercomputer hosted at UT's Texas Advanced Computing Center (TACC) and housed at Sabey Data Centers in Round Rock.
Horizon will come online next year and is projected to deliver about 10x the compute of Frontera, the current top academic system also operated by TACC. The expansion positions UT to train and release open-source large language models and to run large-scale simulations supporting public-interest research.
Why this matters for open-source AI
Open-source computing lets researchers inspect, adapt, and improve models for specific scientific goals. That's essential for transparent methods, reproducibility, and responsible deployment in areas such as health, drug discovery, materials, climate, and national security.
"As a leading public flagship university, our greatest responsibility is to benefit society and support our state and nation through our teaching and research activities," said UT President Jim Davis. "UT's combination of computing capacity, facilities and breadth of research expertise is unrivaled, and it positions us at the forefront of life-changing discoveries that also bolster our country's economic and national security."
Hardware at a glance
- More than 4,000 new NVIDIA Blackwell architecture GPUs added, contributing to a total that now exceeds 5,000 GPUs across UT.
- NVIDIA GB200 systems and NVIDIA Vera CPU servers integrated into Horizon.
- Dell PowerEdge platforms and high-performance networking designed for multi-node training and simulation at scale.
- Hosted at Sabey Data Centers (Round Rock) and operated by TACC.
Access and allocation
More than 1,000 of the advanced GPUs will be dedicated to UT's Center for Generative AI, reserved for projects that need hundreds of accelerators working in parallel on massive datasets. This nearly doubles the center's capacity and gives UT researchers uncommon frequency of access to large-scale accelerated compute.
"This is a next-level system for large-scale generative AI and computing," said Adam Klivans, UT computer science professor and director of the UT-led NSF Institute for Foundations of Machine Learning. "The new clusters will increase our researchers' access to accelerated computing and position UT as the academic leader in building large AI models."
Track record and expected impact
"Our work with Dell Technologies and NVIDIA gives Horizon breakthrough capabilities, particularly in the use of AI for scientific innovation," said Dan Stanzione, director of TACC and associate vice president for research at UT. "Our previous large-scale systems have enabled the work of more than 100,000 students and researchers, leading to countless discoveries ranging from subtle changes in building codes to increased resilience in hurricanes, to Nobel Prize-winning research that advances our understanding of the universe. We expect Horizon will continue that tradition."
With support from the State of Texas - including a recent $20 million appropriation - and NSF LCCF funding, UT is positioned across key national R&D priorities cited by the White House Office of Management and Budget: AI, quantum information science, semiconductors and microelectronics, future computing technologies, and advanced manufacturing.
What researchers can do now
- Plan for scale: profile workloads for multi-node training and large memory footprints; validate data pipelines and checkpointing strategies.
- Prioritize openness: prepare documentation, evaluation protocols, and release plans for datasets, training recipes, and model artifacts.
- Engage early: coordinate with project PIs and center leads to secure allocation windows for pretraining, fine-tuning, and large simulations.
Timeline and partners
Horizon is under construction now in partnership with NSF, Dell Technologies, NVIDIA, and Sabey Data Centers. The system is scheduled to go online next year at TACC, expanding access for researchers across the U.S. and advancing open, high-impact science.
Learn more
Horizon's scale, open ethos, and academic mission make it a practical win for teams that need transparent models, rigorous evaluation, and reliable access to large clusters. It's a clear step forward for public-interest AI and computational science.
Your membership also unlocks:
