The fastest-growing data center designs of 2026 share a trait that quietly upends three decades of decommissioning practice: humans are not meant to touch the equipment. Sealed subsea pods off Shanghai, immersion tanks running AI accelerators, and direct-to-chip liquid loops on the latest 100-kilowatt racks all push servers beyond the access-model data center ITAD has relied on for decades. The hardware is denser, wetter, and increasingly sited where energy is cheap rather than where technicians can drive a forklift. For decommissioning teams, refurbishers, and data destruction providers, that shift is already changing how end-of-life equipment arrives and its condition.
The architecture of rewriting ITAD for data centers
A recent Resource Recycling analysis on underwater data centers flagged China’s Lingang offshore facility as one of the first commercial-scale subsea sites, housing around 2,000 servers in sealed pods cooled by surrounding seawater. The reported Power Usage Effectiveness (PUE) sits near 1.15. The more interesting figure for an ITAD provider is closer to zero opportunities for rack-level service across multi-year deployment windows.
Subsea and offshore
Subsea modules are sealed, lowered, and retrieved intact at the end of a service interval. Decommissioning shifts from rack-by-rack inventory to module-scale handling that resembles marine salvage more than a conventional teardown. Equipment may arrive in a sealed, pressure-resistant enclosure that has not been opened since installation, which may contain marine biofouling and unknown internal moisture.
Underground and cold-climate halls
The Nordic region has been running an onshore version of the same idea for years. Lefdal Mine Datacenter on Norway’s west coast operates inside a former olivine mine and uses fjord water for cooling. Sweden’s Pionen facility sits roughly 100 feet beneath Stockholm granite. Denmark built its data center reputation on cool-climate hyperscale campuses rather than underground halls, but the regional pattern is the same: dense, sealed, and built around natural cooling. For ITAD, that geography means long transport legs, restricted access windows, and recovery timelines tied to regional logistics rather than IT refresh cycles.
Liquid and immersion cooling
The trend is not limited to remote sites. Single-phase and two-phase immersion cooling and direct-to-chip liquid loops are now standard architectural options for AI racks pushing 50 to 100 kilowatts per cabinet. Servers leave dielectric fluid before they leave a rack, and the fluid itself is a regulated waste stream. Two-phase systems have drawn scrutiny over global warming potential and worker exposure. ITAD providers handling this hardware are no longer only IT vendors. They are also fluid handlers, with documentation obligations covering the cooling medium and the silicon.
What changes for recovery, refurbishment, and resale
Sealed, no-touch architectures reshape the equipment stream. A site that swaps entire modules at a refresh interval releases equipment in larger and more synchronized waves. Data center ITAD planning that assumed steady drip volumes now needs to model lumpier, project-driven flows.
Condition profiles shift in both directions. Hardware designed for multi-year, maintenance-free operation shows different wear patterns than equipment in a serviced rack. Some components arrive in better shape than expected, having avoided the dust, vibration, and human error of an active hall. Others, exposed to seawater, biofilms, or contaminated fluids, arrive with failure modes that traditional grading processes were not designed to detect. Refurbishment yield assumptions need to be revisited on a pod-by-pod basis.
Value recovery follows the same logic. Modular designs are easier to redeploy intact and harder to break down for component harvest, shifting the resale-versus-recycle decision earlier in the data center ITAD workflow.
Data sanitization gets harder before it gets easier
Every architectural change above touches data destruction. NAID AAA expectations and R2v3 sanitization requirements assume drives can be physically presented for logical wiping, sampled at five percent, or destroyed by shredding when wipes fail. In sealed modules, immersion tanks, and densely interconnected accelerator boards, that workflow no longer happens at the device level. NVMe drives are soldered to compute boards. Storage may be spread across CXL fabrics that do not map cleanly to a single asset tag. Some components arrive still coated in dielectric fluid, which must be removed before any wipe tool can connect.
The practical implication is that the chain of custody now needs to cover the cooling medium, the enclosure, and the asset. Documentation must address modules that are never separated from their cooling loop in the field, and sampling regimes must reflect what can and cannot be reached without destructive disassembly. None of this breaks the standards. It does mean data center ITAD providers have to design intake, sanitization, and destruction workflows around hardware engineered to resist exactly that kind of access.
Sustainability math beyond PUE
The sustainability case for underwater, underground, and liquid-cooled designs leans heavily on operational efficiency. PUE near 1.15, low or zero Water Usage Effectiveness, and renewable power are real numbers, but incomplete. Subsea deployments require pressure-resistant enclosures, specialized cabling, and vessel operations at retrieval. Liquid cooling adds dielectric fluid manufacture, handling, and end-of-life disposition to the lifecycle inventory.
A complete lifecycle view treats end-of-life as an input to the carbon picture, not a footnote. Data center ITAD is where most of that math gets settled. Recovery rates, refurbishment yields, and recycling outcomes determine whether a low-PUE facility delivers a low lifecycle footprint or only an attractive operating one. Operators reporting on Scope 3 emissions and circular economy metrics need ITAD partners who can produce that data, not just a certificate of destruction.
How HOBI approaches data center ITAD in the new landscape
The new architectures do not require a brand-new playbook from scratch. They require an existing one applied with sharper edges. HOBI’s data center services already cover on-site decommissioning, secure transport, multi-stage data sanitization including the HOBI Shield erasure tool, refurbishment, and value recovery. For liquid-cooled and sealed equipment, the front end changes. Site assessments need to capture cooling architecture, fluid type, and module-level access before any boots hit the floor. Transport plans must account for sealed enclosures and residual fluids. Sampling and destruction protocols need to account for hardware that may not be present, such as the servers in last year’s project.
The same logic applies upstream. Buyers specifying immersion-ready AI racks or modular pods today are making data center ITAD decisions in the same purchase order, whether they realize it or not.
The window for catching up
Underwater data centers may stay niche. Immersion cooling and direct-to-chip loops will not. The AI hardware refresh wave is pushing mainstream colocation toward liquid in the next refresh cycle, not the one after. ITAD programs that are not already documenting cooling architectures, training technicians in fluid handling, and building intake around modular hardware will do that work under deadline pressure within twenty-four months.
For enterprise IT, security, and sustainability leaders, the practical step is a current-state review: map which sites are moving to liquid cooling, confirm that ITAD contracts cover fluid as well as silicon, verify chain-of-custody language for sealed modules, and ask whether reporting will support both Scope 3 and circular economy disclosures from a single data set. The data center ITAD market is still catching up to underwater experiments. It cannot afford to be left behind by mainstream liquid cooling.
Frequently Asked Questions
How does liquid cooling change data center ITAD compared with air-cooled hardware?
Liquid cooling adds the coolant itself to the workflow. Servers must be drained and decontaminated before logical sanitization or shredding can begin. Dielectric fluids, particularly two-phase coolants, are often regulated waste. Chain of custody and sampling documentation now needs to cover the fluid, the cooling loop, and the asset, not just the device.
Are sealed modular data centers harder to decommission than traditional racks?
Work shifts from rack-level to module-level. Equipment is retrieved as an intact unit rather than dismantled on site. That can simplify transport but complicates refurbishment, component harvest, and data destruction, because the hardware was engineered to resist being opened during its service life.
Do R2v3 and NAID AAA cover immersion-cooled and sealed equipment?
The standards still apply, but workflows must adapt. R2v3 requires physical destruction when logical sanitization fails and routine independent sampling. NAID AAA governs secure destruction and chain of custody. Both can be met for liquid-cooled and modular hardware, provided procedures are designed around the architecture rather than retrofitted.
What sustainability reporting risks come from poor end-of-life handling of liquid-cooled hardware?
Low PUE means little if dielectric fluid is mishandled, modules go to scrap intact, or refurbishment opportunities are missed. Scope 3 and circular economy reporting depends on documented recovery and recycling outcomes. ITAD partners should produce auditable data on fluid disposition, component recovery, and final destruction, not only a certificate.
When should organizations update their data center ITAD program for new cooling architectures?
Before the next major refresh. AI workloads are pushing density past the limits of air cooling in mainstream colocation, which means liquid-cooled hardware will reach end of life on the standard three-to-five-year cycle. Updating site assessments, contracts, and technician training now avoids scrambling later.