UC San Diego will nearly triple shared AI computing capacity with 32 new GPUs

UC San Diego is adding four servers to its shared AI computing platform, with a Winter quarter 2027 start expected to nearly triple its GPU capacity. The Halıcıoğlu School of Data Science and Computing is funding the hardware as demand grows for both scheduled coursework and student projects that extend beyond the classroom.

The university announced the expansion on November 3. Some students have already turned to commercial cloud GPU services to obtain computing power for classes and independent work, according to the university. The investment is intended to make more of that hands-on AI work possible through university-supported infrastructure.

The platform, called PARASOL, succeeds UC San Diego’s Data Science & Machine Learning Platform, or DSMLP. Its full name—Platform for Applied Research, Analytics, Science and Open Learning—reflects an effort to connect class assignments with research rather than maintain a resource used only for courses.

More room during class—and outside it

The four new servers will contain 32 NVIDIA RTX PRO 6000 Blackwell GPUs, the graphics processors supplying the additional AI computing power. The school’s investment includes hardware and systems support. It was enabled by the school’s founding gift from computer scientist and philanthropist Taner Halıcıoğlu.

The expected gain differs by use: overall PARASOL GPU capacity will nearly triple, while peak-period capacity available to Halıcıoğlu School courses will nearly quadruple. That larger instructional increase targets the gap between student demand and the resources available when courses need them.

Outside peak instructional periods, the new servers are intended to support research, capstone projects, graduate work and coursework at other times. The university gives separate annual capacity estimates for these two windows, measured in GPU-hours—the hours of processing available on the graphics processors.

Campuswide service, school-first reservations

More hardware comes with an explicit access priority. A new reservation system gives Halıcıoğlu School students and faculty first priority, while PARASOL continues serving coursework and student research across campus. The expansion therefore preserves a shared university resource without giving every user the same place in the queue.

Campus IT leaders project that demand from the school’s programs alone will rise from about 43,000 advanced-GPU processing hours in 2026 to more than 140,000 in 2030. They also expect rising demand elsewhere on campus as enrollment grows and more courses incorporate advanced AI techniques.

Access beyond an assignment

Halıcıoğlu School students will also be the first to receive automatic access to Open Learning workspaces. Those environments are intended for exploring large language models and other AI tools independently, including personal projects, early-stage research and entrepreneurial ideas, rather than only completing assigned classwork.

For instructors and existing users

• Instructors can request PARASOL access for their courses through the university’s ITS Services & Support Portal.
• Current DSMLP, Datahub and Research Cluster users will keep their existing accounts as the platform evolves.

The deployment will also let university leaders monitor demand and utilization to guide further investment. UC San Diego says future development includes software and services intended to make the technology easier for instructors and students to use, particularly outside science, technology, engineering and mathematics fields.