Oracle Zettascale Hub: 131,072 GPU AI Supercluster Texas

Status: Active Deployment / Expansion (March 2026)   |   Location: Abilene (Taylor/Shackelford Counties), Texas

Project Overview

Formally unveiled as the foundational infrastructure layer for the OpenAI Stargate initiative, the Zettascale Hub represents Oracle’s strategic response to the “brute force” scaling phase of artificial intelligence. Located in Abilene, the campus departs from conventional hyperscale layouts by concentrating all critical compute resources within a tightly integrated two-kilometer radius, significantly reducing GPU-to-GPU latency and enabling ultra-high-speed model training at scale. As of March 2026, the facility is deploying the advanced Zettascale10 architecture, leveraging Oracle’s proprietary Acceleron RoCE networking fabric to interconnect hundreds of thousands of NVIDIA Blackwell GPUs and AMD Instinct GPUs. This high-bandwidth, low-latency configuration enables peak performance levels of up to 16 zettaFLOPS (sparse FP4), positioning the hub among the most powerful AI compute environments globally. To address the immense power requirements without overburdening the local grid, Oracle is investing in dedicated onsite substations alongside large-scale battery energy storage systems (BESS), ensuring operational stability and energy resilience. In parallel, the project integrates a closed-loop, non-evaporative cooling system, reducing water consumption to levels comparable to a standard office building an essential adaptation for the semi-arid climate of West Texas. Collectively, the Zettascale Hub exemplifies a new class of AI infrastructure, where extreme compute density, energy autonomy, and resource efficiency converge to support next-generation foundation models.

Quick Facts

City Profile

Companies Involved

Technical Specifications

Milestones

Investment Details