Overview
Largest US Academic Supercomputing Center
TACC has run a continuous line of supercomputers since 2001, serving US researchers in physics, climate, medicine, materials, energy and AI. It marks its 25th anniversary in 2026 under executive director Dan Stanzione, principal investigator for the Leadership-Class Computing Facility (LCCF).
NSF selected TACC to host the LCCF, the national flagship for open-science computing. Horizon, its first system, is installed at Sabey Data Centers' Round Rock campus in Williamson County, while legacy systems stay in Travis County.
Systems
Supercomputer Portfolio
Vista moved TACC from x86 to ARM architecture, and Horizon continues on NVIDIA Grace Blackwell. Researchers work across Intel, AMD, ARM and NVIDIA systems, so codes tuned for different architectures each have a home.
| System | Status | Architecture | Role |
|---|---|---|---|
| Frontera | Operational since 2019 | Intel Xeon CPU, 8,368 nodes, 38.75 PF | NSF leadership system; workloads moving to Horizon |
| Stampede3 | Operational | Intel | General scientific computing |
| Lonestar6 | Operational | AMD | Mid-scale system for Texas and US researchers |
| Vista | Operational since 2024 | NVIDIA Grace Hopper and Grace, ARM, 44.9 PF | AI-focused bridge to Horizon; first ARM system at TACC |
| Horizon | Early operations 2026 | NVIDIA Grace Blackwell on Dell PowerEdge; about 4,000 GPUs | LCCF flagship; NVIDIA Vera CPU partition due winter 2026 to early 2027 |
Horizon
Horizon at Sabey Round Rock
Horizon's GPU racks were operational by July 2026 and the system is entering early operations, with its NVIDIA Vera CPU partition expected in winter 2026 or early 2027. TACC projects science codes running 15 to 20 times faster than on Frontera and AI workloads up to 100 times faster. NSF funding exceeds $400 million.
Sabey built its Round Rock campus on the former Sears Teleserve site, demolished in 2024. Horizon's power and liquid cooling needs exceed what UT campus facilities can supply, so TACC colocates with a commercial operator for the first time.
| Partner | Role |
|---|---|
| Dell Technologies | PowerEdge servers and Integrated Rack Scalable Systems |
| NVIDIA | Grace CPU, Blackwell GPU and Vera CPU silicon |
| VAST Data | 400 PB all-flash storage, 10+ TB/s bandwidth |
| Spectra Logic and Versity | Ranch exabyte-scale archive, in production since 2025 |
| Sabey Data Centers | Physical hosting in Round Rock |
Context
Place in the Austin Compute Corridor
Horizon runs open-science work on the same NVIDIA Grace Blackwell platform that commercial AI operators use, including Tesla's Cortex clusters at Giga Texas. TACC trains the computing workforce that Apple, AMD and Dell draw on, and works alongside the UT TIE NGMM Center through the Cockrell School of Engineering.
The Round Rock site places TACC in the Williamson County data center corridor with Switch, Skybox and the Georgetown, Hutto and Taylor cluster near Samsung Taylor. Through the National AI Research Resource (NAIRR) pilot, TACC gives academic and government researchers AI compute outside commercial training operations. See Horizon at Sabey Round Rock for the site profile.
Risks
Funding and Research Demand
Horizon's NSF award covers construction and initial operation; running it to about 2030 depends on continued NSF support for the LCCF. Each new system needs a new funding cycle and architecture decision, a slower cadence than commercial AI buildouts. Federal R&D funding levels also set researcher demand, which commercial operators do not depend on.
Record
Dataset Record
- Sources
- TACC; NSF LCCF; Dell Technologies; VAST Data; City of Round Rock
- Last verified
- Q4 2026