{"resourceId":"ucsd-parasol-gpu-capacity-expansion-20261103","versions":[{"version":"external-97ab48e426c0c76ebe0f92936f37651d7b20da5060b04b1db63fd22dbb71a231","resource":{"id":"ucsd-parasol-gpu-capacity-expansion-20261103","title":"UC San Diego’s Halıcıoğlu School Expands AI Computing Infrastructure to Address Critical Shortages (PARASOL)","organization":"UC San Diego Halıcıoğlu School of Data Science and Computing / IT Services","sector":"Public research university (shared campus AI/GPU computing for coursework and research)","geography":"United States (California)","publishedAt":"3 November 2026 (UC San Diego Today)","publicationDate":"2026-11-03","eventDate":null,"sourceName":"UC San Diego Today","sourceLabel":"Tiffany Fox, \"UC San Diego’s Halıcıoğlu School Expands AI Computing Infrastructure to Address Critical Shortages,\" UC San Diego Today, 3 Nov 2026","sourceUrl":"https://today.ucsd.edu/story/uc-san-diegos-halcolu-school-expands-ai-computing-infrastructure-to-address-critical-shortages","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["infrastructure","accessibility-workforce","operating-model"],"finding":"UC San Diego announces PARASOL (Platform for Applied Research, Analytics, Science and Open Learning), the next generation of its Data Science & Machine Learning Platform (DSMLP). School-funded GPU servers with 32 NVIDIA RTX PRO 6000 Blackwell GPUs will nearly triple overall GPU capacity and nearly quadruple peak-period capacity for Halıcıoğlu School courses beginning Winter 2027. The university projects Halıcıoğlu School demand to rise from about 43,000 advanced GPU hours in 2026 to more than 140,000 by 2030, and states the new servers will provide about 66,800 peak instructional GPU-hours and more than 213,000 off-peak GPU-hours yearly for research, capstones and other coursework. A reservation system gives School students and faculty first priority while continuing campuswide service; some students had been buying commercial cloud GPUs due to shortages.","sledRelevance":"Directly relevant to public research universities facing instructional and research GPU shortages: shows a shared-platform response with reservation priorities, philanthropic funding and explicit GPU-hour projections rather than only a hardware purchase announcement.","evidence":"Institutional news release with named capacity figures and Winter 2027 timeline; self-reported projections and planned hardware. No independent utilization audit or cost comparison versus cloud. Classed as vendor-claim (university self-report).","architectureImplications":"Shared campus AI platform spanning coursework and research, with reservation/priority scheduling and evolution from an existing DSMLP stack; capacity added as discrete GPU servers rather than a full supercomputer buy.","governanceImplications":"Priority rules favoring one school while serving campuswide demand require transparent allocation policy; not detailed beyond first-priority reservation language.","securityPrivacyImplications":"Not discussed in the inspected article; shared student/research platforms typically need tenancy and data-handling controls.","caveats":"Capacity and demand figures are institutional projections; Winter 2027 start means benefits are prospective; research vs instructional split of off-peak hours not independently verified.","streamIds":["research"],"roles":{"sales":"Capacity-planning framing: UC San Diego is nearly tripling shared GPU capacity on PARASOL after students resorted to commercial cloud GPUs, with ~66,800 peak and 213,000+ off-peak GPU-hours projected yearly. That supports bounded conversations with research computing and data-science deans about shared platforms versus shadow cloud spend. Interpretation: figures are self-reported plans, so they cannot support guaranteed utilization or cost-savings claims.","engineering":"Fit is campuses with mixed instructional/research GPU demand. Architecture is a shared platform (DSMLP→PARASOL) plus reservation priorities and additional Blackwell GPU servers. Prerequisites include scheduler/reservation software, identity integration and chargeback or quota policy. Constraints: School-first priority may starve other units. Proof of value: measure queue wait times, cloud spillover spend and research GPU-hour delivery after cutover.","delivery":"Delivery work is a shared-AI-platform capacity expansion: hardware install, reservation policy, and communication to stop unnecessary cloud spillover. Ownership sits with IT Services and the data-science school. Governance checkpoints include transparent priority rules and research/instructional quotas. Acceptance criteria are proposals such as published GPU-hour delivery versus plan; risks include priority capture and underestimating power/cooling."},"roleDetails":{"sales":{"customerProblem":"Students and researchers hit shared GPU shortages and buy commercial cloud capacity, raising cost and uneven access.","stakeholders":["CIO / IT Services","Data science dean","Research computing director","Faculty senate computing committee"],"discoveryQuestions":["What is current queue wait and cloud spillover spend?","How are instructional vs research GPU hours allocated?","Who sets reservation priority?","What is the Winter-term peak demand forecast?"],"valueHypothesis":"Expanding a shared campus platform with clear reservations may absorb demand that today leaks to commercial cloud GPUs.","potentialEngagement":"A GPU demand and allocation assessment culminating in a shared-platform expansion plan.","unsupportedClaims":["Guaranteed cost savings versus cloud","Tripled capacity already online today","Equal access for all units under School-first priority"]},"engineering":{"fit":"Strong for universities with an existing shared ML platform needing capacity; weaker where only a few research groups need dedicated nodes.","architectureIntegration":"Shared GPU servers, reservation/scheduler, identity and quota layers spanning coursework and research workspaces.","prerequisites":["Facility power/cooling","Scheduler and reservation tooling","Support staffing"],"constraints":["Prospective Winter 2027 delivery","School-first priority politics","Self-reported hour projections"],"security":"Enforce tenancy isolation between student coursework and sensitive research projects; log privileged access.","proofOfValue":["Track wait times before/after","Measure commercial cloud spillover","Report delivered vs projected GPU-hours"]},"delivery":{"work":"Install GPU servers; Configure reservations/quotas; Communicate access changes and Open Learning workspaces","dependencies":["Facilities","IT Services","School administration"],"ownership":"IT Services lead architect with School funding owner.","skillsAdoption":"Instructors and TAs learn reservation workflows; researchers learn off-peak research queues.","governanceCheckpoints":["Publish priority and quota policy","Review spillover cloud exceptions","Quarterly capacity vs demand review"],"acceptanceCriteria":["Proposed: Winter 2027 peak capacity online","Proposed: measured reduction in unjustified cloud GPU purchases","Proposed: research off-peak hours delivered per plan"],"risks":["Priority capture by one school","Power/cooling delays","Over-promised hour projections"]}},"retrievedAt":"2026-11-05T22:05:00-06:00","enrichedAt":"2026-11-05T22:08:41-06:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Aims to keep hands-on AI training on campus rather than only for students who can buy cloud GPUs; equity effects not measured.","procurementImplications":"Philanthropic/school-funded GPU servers into a shared platform; procurement should preserve multi-tenant scheduling and avoid lock-in that blocks campuswide access.","operatingModelImplications":"Reinforces a shared-service operating model for instructional and research AI compute with explicit priority rules.","sourceVerification":{"openedUrl":"https://today.ucsd.edu/story/uc-san-diegos-halcolu-school-expands-ai-computing-infrastructure-to-address-critical-shortages","referenceExcerpt":"new School-funded GPU servers will nearly triple PARASOL’s overall GPU capacity and nearly quadruple the peak-period capacity","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}