The Leadership-Class Computing Facility's first new site selection in a generation lands at an HBCU. The workforce implications matter as much as the FLOPS.
Morehouse College, a historically Black institution in Atlanta with roughly 2,100 undergrads, has been selected to host Horizon, a leadership-class supercomputer funded by a $457 million National Science Foundation grant through the Leadership-Class Computing Facility (LCCF) program. The project is led by the Texas Advanced Computing Center at UT Austin, with Morehouse serving as a primary partner for deployment, programmatic support, and workforce development.
This is the most consequential site selection in academic supercomputing in years, and it has almost nothing to do with diversity optics.
Let's put this in context. The NSF's Leadership-Class Computing Facility program is the successor to the funding mechanisms that built NCSA, TACC, and the other pillars of the U.S. academic computing infrastructure. These aren't grants that get sprinkled around. LCCF represents the national strategy for where leadership-class computing resources live and who has access to them for the next decade-plus.
When TACC won the original LCCF award, it was understood that the facility would be distributed, not a single site but a network. Morehouse's selection as the site for Horizon means an HBCU is now part of the core architecture of the national computing strategy. Not an affiliate. Not a minor partner getting access credits. A primary facility host.
The initial $5 million covers site construction and the beginning of workforce development programs. Additional funding will follow for ongoing operations. The supercomputer itself will be among the most powerful academic systems in the Southeast.
Dr. Kinnis Gosha, the grant's principal investigator and chair of Computer Science at Morehouse, framed it directly: "This contribution cements Morehouse's place as the undisputed HBCU leader in artificial intelligence. As a national resource provider, we will empower other HBCUs and non-research-intensive institutions to contribute to growing their research capacity and enhancing student learning."
That "national resource provider" designation is doing a lot of work in that sentence. Here's what it means in practice:
Morehouse's Center for Broadening Participation in Computing will serve as the national epicenter for LCCF programmatic support. The college will lead summer enrichment programs for middle and high school boys, a postbaccalaureate program in artificial intelligence, three weeklong faculty accelerators annually in Punta Cana (Dominican Republic) focused on research methodology, teaching, and grant proposal writing, and the annual InSPIRE workshop (Integrating Supercomputing-Powered Instruction, Research and Entrepreneurship) in Austin.
That's a cradle-to-career pipeline. Middle schoolers getting their first exposure to computational thinking. Undergrads working on a leadership-class system. Postbacs gaining AI specialization. Faculty at under-resourced institutions learning to write competitive grants and integrate HPC into their curricula.
The hardware will eventually be decommissioned. The pipeline won't.
The U.S. academic computing community has a workforce problem that everyone acknowledges and nobody has adequately addressed. The people who design, build, operate, and use supercomputers are not representative of the country's population. According to national labor statistics cited in Morehouse's announcement, 62% of tech jobs are held by white Americans. In HPC specifically, the numbers are worse.
This isn't just an equity issue. It's a capacity constraint. The United States is in a global competition for AI and computing leadership. Every other major nation-state is investing heavily in compute infrastructure. The talent bottleneck is as real as the silicon bottleneck, and you don't solve it by putting recruitment flyers at the same 20 R1 universities that have always had access to leadership-class systems.
Morehouse President Dr. F. DuBois Bowman put it this way: "By hosting one of the Southeast's most powerful academic supercomputers, we are providing HBCUs with unprecedented computational power to explore bold ideas, accelerate discovery, and unleash new frontiers of creativity and innovation. This investment positions our students and faculty to help shape the future of science, technology, and global problem-solving."
The framing matters. This isn't "we're giving an HBCU access to a supercomputer." It's "an HBCU is hosting a national facility and leading its workforce mission." The difference is institutional agency.
TACC's role as the lead institution is significant for operational reasons. Dan Stanzione and his team at TACC have deployed and operated some of the largest academic systems in the world: Frontera, Stampede, Lonestar. They know how to build and run leadership-class facilities. That operational expertise, combined with Morehouse's institutional reach into communities that HPC has historically failed to engage, creates a partnership model that could become a template.
The distributed LCCF structure means Horizon won't just serve Morehouse researchers. It's a national resource. Researchers across the country will allocate time on it. But the physical presence of a leadership-class system on a Morehouse campus changes the gravitational field for students and faculty there. Proximity matters. When the machine is down the hall, not on the other side of the country, the barrier to engagement drops dramatically.
Several questions remain as the project moves from announcement to implementation:
System specifications haven't been publicly disclosed yet. Given the LCCF's scale ambitions and the timeline (site construction starting now), this will likely be a next-generation system, potentially GPU-accelerated. The vendor selection will be closely watched.
On the operational timeline: the $5 million initial funding covers construction and early workforce development, but the ramp to full operational status will take years. The question is whether the pipeline programs can start generating impact before the system is fully online.
If the Morehouse-TACC model works, and there are strong reasons to believe it will, does NSF replicate it? Are there other institutions positioned to host distributed LCCF nodes?
And ultimately, the system will be judged by the science it enables. Climate modeling, machine learning, biomedical research: these are areas where computational capacity is directly rate-limiting.
There's a temptation to frame this as a feel-good story. Resist it. The NSF didn't award $457 million because it felt good. It awarded $457 million because the national computing infrastructure needs both more capacity and broader participation to remain competitive. Morehouse was selected because it can deliver both.
The supercomputing community's center of gravity has historically been concentrated in a handful of institutions: ORNL, ANL, TACC, NCSA, the usual suspects. Horizon at Morehouse doesn't displace those pillars. It extends the foundation. And in a decade where AI compute demand is growing faster than any single institution can absorb, that extension isn't optional.
It's infrastructure.
