San Diego’s AI Leap: Expanse Upgrade & Cooling-Ready Labs

San Diego just got a serious AI compute boost—and it won’t stay inside the data hall for long. 

UC San Diego’s Supercomputer Center (SDSC) is expanding Expanse with an $5.3M upgrade that adds liquid-cooled, NVIDIA H100–class GPU nodes, accelerating AI research across the region. 

The new nodes are Dell PowerEdge XE9640 systems—built for direct liquid cooling and high-density AI workloads—precisely the kind of hardware that pushes traditional power and HVAC assumptions off the table. 

Why this matters beyond campus

When public research institutions scale cutting-edge compute, private teams follow—by partnering on grants, colocating talent nearby, or spinning up pilots that need lab + office + edge compute under one roof. SDSC’s Expanse already underpins AI-heavy science, and the H100 expansion raises the ceiling on what local teams can train and test here in San Diego. 

Translation for real estate: more tenants will ask for power-dense, cooling-ready space where AI, robotics, and instrumentation live alongside offices and collaboration areas.

The new R&D space checklist (AI edition)

If your facility targets AI/ML, autonomy, or biotech-with-AI, expect requests like these:

• Power capacity & distribution. Sufficient service, redundant panels, and clean paths for future step-ups and additional circuits.
• Cooling strategies beyond office norms. Suites that can accommodate liquid-ready or high-airflow racks (e.g., chilled-water stubs, riser access, roof/yard equipment zones) to support dense GPUs like the XE9640 class. 
• Room to integrate lab workflows. Flex bays for benches, fabrication/prototyping, or electronics with proper ESD and ventilation.
• Reliability & resiliency. Generator tie-ins, UPS/clean power for short ride-throughs, and monitoring.
• Connectivity. Diverse fiber entrances and low-latency routes to campus, cloud on-ramps, and regional IX points.
• Compliance & growth. Clear paths for permits, mezz expansions, rooftop rights, or modular buildouts as teams scale.

Neighborhood dynamics: why proximity helps

Being near SDSC and UC San Diego shortens cycles between academic partners and industry teams—grant work, joint appointments, recruiting, and equipment sharing all get easier. As Expanse’s capacity increases and NAIRR/NSF programs broaden access, expect more cross-pollination between lab benches and GPU clusters—especially for startups that want quick experiments without committing to their own full data rooms. 

For landlords: how to capture the demand

• Pre-engineer “AI-ready” suites. Provide base calculations for heat loads and water loops, identify condenser/CRAC/CRAH landing zones, and show upgrade paths in marketing materials.
• Spec a small “compute closet.” Even a single high-density rack zone with proper cooling and power separation can win deals.
• Offer phased capacity. Make it obvious how a tenant can step from today’s load to a higher tier without re-permitting the whole floor.

For tenants: questions to ask on tour

• What’s the available kW at suite and building level—and the upgrade path?
• Can the building support liquid-assisted cooling if needed for future GPU racks? (Dell’s XE9640 class thrives on it.) 
• Where can we place roof/yard equipment, and how fast can we add it?
• How is network redundancy handled (diverse entries, carriers, SLAs)?
• What’s the landlord’s playbook for lab + compute TI timelines?

How IPG helps in San Diego

IPG works with owners and tenants to source, evaluate, and retrofit spaces for AI-forward R&D—blending lab footprints with power-dense, cooling-ready infrastructure. From site shortlists and load modeling to TI scopes and vendor coordination, we de-risk decisions so teams can build, train, and ship—faster.

Ready to align your space with San Diego’s next wave of AI? Let’s talk.