The facility runs actual data center equipment while demonstrating three tiers of cooling architecture side-by-side. Contractors attend sessions showing how precision air handlers integrate with rear-door heat exchangers and liquid cooling distribution units serving GPU clusters. DDC Solutions built the space as both a sales tool and training ground — techs walk through operating aisles, see real-time thermal maps, and watch refrigerant-based CDUs cycle through load variations as server racks pull 40kW to 80kW per cabinet.

AI workloads generate heat densities traditional CRAC units cannot handle alone. A single NVIDIA H100 GPU rack can hit 70kW, requiring liquid cooling loops that deliver chilled water or dielectric fluid directly to cold plates mounted on processors. The San Diego lab runs hybrid configurations: perimeter air handlers maintaining ambient conditions at 75°F, in-row coolers targeting hot aisles, and rack-mounted CDUs feeding chip-level cold plates. Operators demonstrate failover scenarios, showing what happens when a pump trips or a secondary loop loses pressure. These are the failure modes contractors will troubleshoot when hyperscalers and colocation providers call at 2 a.m.

DDC Solutions offers multi-day training courses at the site covering hybrid cooling design, glycol loop commissioning, and predictive maintenance for liquid-cooled infrastructure. Sessions include hands-on work with Daikin's Rebel rooftop units adapted for high-sensible-load data halls, plus programming exercises for BACnet controllers managing cooling tower sequencing and chiller plant optimization. Contractors leave with load calculation worksheets specific to AI server deployments and refrigerant piping diagrams for two-phase immersion cooling systems.

This week, walk your commercial team through the economics: a 2MW AI data center requires roughly $4 million in cooling infrastructure versus $1.2 million for equivalent office space. Owners need contractors who understand liquid cooling liability, leak detection placement, and how to size backup systems when uptime SLAs demand five-nines reliability. If your market includes colocation builds, cloud on-ramps, or corporate AI labs, schedule time in San Diego before these systems show up on your next bid sheet. The training pipeline for liquid-cooled data centers is still thin — early movers will own the local service contracts.