TACC Opens Horizon Allocation Window Through June 15, 2026

The Texas Advanced Computing Center has officially opened its first allocation window for the Horizon leadership-class supercomputer, marking a critical transition point for U.S. high-performance computing infrastructure. Submissions are now being accepted through June 15, 2026, with rolling reviews commencing May 15th. System access is anticipated to begin in the summer of 2026, though exact dates remain fluid pending final commissioning.

This allocation cycle represents the Early Operations Period for Horizon, which is funded through the National Science Foundation as part of the Leadership-Class Computing Facility (LCCF) project. The system delivers an order of magnitude performance improvement over the previous Frontera system, a specification that matters significantly for researchers whose workloads have outgrown existing capacity.

Horizon's architecture combines 4,750 Dell/NVIDIA Vera CPU nodes with 2,000 Dell/NVIDIA Grace-Blackwell GPU nodes. For researchers who have spent years waiting for GPU resources that actually match their simulation requirements, this configuration addresses a persistent bottleneck (a problem that has plagued users for years, frankly). The mix enables both traditional CPU-intensive workloads and modern AI/ML applications requiring tensor acceleration.

According to the official LCCF Allocations documentation, the current solicitation focuses on Leadership Resource Allocation (LRAC) tracks. These large allocations range from 125,000 to 500,000 Service Units (SUs) for Horizon, with Vista allocations capped at 50,000 SUs. Each allocation runs for six months, and successful applicants must demonstrate both readiness to use the cycles and current peer-reviewed research funding to support their activities.

Only Principal Investigators may submit allocation requests, though the definition of eligible PI extends beyond traditional faculty roles. NSF Graduate Student Fellows and Frontera Fellows qualify as exceptions to the graduate student restriction. Federal agency research staff, K-12 educators, independent museums, and U.S. commercial organizations—particularly small businesses with strong scientific capabilities—may also apply under specific conditions.

The submission process requires researchers to navigate the Resource Allocation System, a portal that demands careful preparation. Proposals must include documentation of eligibility, detailed research descriptions, sources of research support, and justification for the requested SU amount. The friction here is real: researchers who have submitted allocation requests before know that the difference between approval and rejection often comes down to how well they articulate their resource needs.

Horizon's allocation philosophy emphasizes compelling science or engineering challenges that genuinely require large-scale computing resources. Proposals from or including junior researchers are explicitly encouraged, reflecting a stated goal of building a community capable of using computing at the highest scales. The Horizon team will provide consulting support to each project team granted access through the solicitation.

Collaborative projects involving non-U.S. researchers are permitted provided they include substantive intellectual participation by U.S. researchers. This requirement maintains the NSF's focus on domestic research capacity while acknowledging that modern science operates across borders. The expectation is that any research performed on Horizon will result in publication in a broadly available journal or conference.

Independent reporting from HPCwire corroborates the timeline and allocation structure outlined in TACC's official documentation. The coverage confirms that this represents the first Horizon opportunity for demonstrating large-scale science, with additional allocation tracks designed to accommodate wider research needs planned for future cycles.

For context, the transition from Frontera to Horizon represents more than incremental improvement. Researchers who have worked with Frontera know the physical reality of queuing for weeks, watching jobs sit in limbo while waiting for resources to become available. Horizon's architecture aims to reduce that friction, though the allocation process itself introduces a different kind of waiting game.

The six-month allocation duration creates a specific operational rhythm. Teams must plan their workloads to fit within this window, which means coordinating data transfers, testing code on smaller systems first, and ensuring all collaborators understand the access constraints. The clock starts ticking once access begins, and there's no extension for projects that fall behind schedule.

Smaller research allocations and startup requests will become available in future cycles, but the current solicitation focuses exclusively on large-scale LRAC requests. This narrow focus means that researchers with modest resource needs may need to wait for subsequent allocation windows or seek alternative computing resources.

Whether the allocation review process can handle the expected volume of submissions remains uncertain. Rolling reviews beginning May 15th suggest an attempt to distribute the workload, but the final decision timeline has not been publicly specified. Researchers submitting proposals should prepare for potential delays in notification.

The June 15 deadline is firm. Submissions received after this date will not be considered for the current cycle, and there's no indication that TACC will extend the window. This creates a hard constraint for teams still finalizing their research proposals or securing institutional approvals.

Whether the promised performance gains translate into meaningful research acceleration depends on factors beyond raw compute power. Data transfer speeds, storage capacity, and the ability of research teams to effectively exploit the system's capabilities will all determine whether Horizon delivers on its potential. The allocation process is just the first hurdle.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn