Why Hyperscale Operators Are Looking Underwater
As AI and 5G workloads surge, conventional facilities are running into limits on energy, cooling and land. Underwater data centers offer an alternative: sealed server modules placed on the seabed, where stable temperatures and abundant seawater provide a natural heat sink. A large-scale deployment off a major coastal metropolis has immersed more than 2,000 servers inside pressure-resistant pods situated 35 meters below the surface. This offshore server infrastructure is designed for GPU-intensive compute, catering to training and inference tasks that demand high power density and constant cooling. By shifting away from sprawling land campuses and their energy-hungry chillers, operators aim to boost data center power efficiency while freeing up real estate onshore. The experiment signals an emerging infrastructure trend in which critical cloud workloads move closer to coastal population hubs, but physically disappear beneath the waves.
Seawater Cooling Systems Turn the Ocean into a Giant Heat Sink
Cooling remains one of the biggest energy drains in large facilities, with traditional chillers often consuming around 40% of total power. In submerged designs, seawater cooling systems exploit the ocean’s immense thermal capacity to draw heat directly from densely packed racks. The 24MW facility in the Lingang Special Area routes heat from its GPU clusters into surrounding seawater, eliminating chillers, cooling towers and freshwater use entirely. Operators report a power usage effectiveness below 1.15, a sharp improvement over the 1.5 and higher figures typical of air-cooled halls. This uplift in data center power efficiency translates into an overall power consumption reduction of 22.8% versus comparable land-based builds. Instead of fighting external temperatures with mechanical refrigeration, the ocean absorbs and dissipates heat passively, turning the environment itself into the primary cooling asset for the servers below.
Closing the Loop with Offshore Wind and Submerged Servers
Locating underwater data centers near offshore wind farms creates a tight coupling between generation and consumption. In the Lingang deployment, project sponsors state that roughly 95% of the facility’s power comes from co-located offshore wind turbines. This configuration reduces dependence on grid electricity and stabilizes power delivery for hyperscale workloads that cannot tolerate frequent interruptions. By keeping energy production and compute infrastructure in the same marine zone, transmission losses fall and planning for capacity becomes more predictable. The result is a quasi-closed-loop green energy system where clean power feeds high-density racks that, in turn, shed heat directly into the sea. For operators under pressure to decarbonize, underwater data centers linked to offshore wind offer a way to scale AI and 5G services without proportionally scaling fossil-based power consumption or land-based infrastructure footprints.
From Experimental Pods to Viable Offshore Server Infrastructure
Submerged computing is not entirely new. Microsoft’s Project Natick placed 855 servers on the seabed off Scotland in 2018, proving that sealed underwater pods can run reliably. The controlled marine environment cut failure rates to just 0.7%, about one-eighth of comparable land facilities, by eliminating dust, temperature swings and accidental human damage. Yet the initiative was shuttered in 2024, with cost and deployment complexity outweighing its technical success at that time. The new 2,000-plus server installation marks a shift from prototype to commercial-scale offshore server infrastructure. Operators are betting that maturing manufacturing, better remote monitoring and rising energy prices now tip the economics in favor of long-lived, unmanned modules. The focus has moved from whether underwater hardware can work to how quickly it can be rolled out and integrated into mainstream cloud footprints.
The Unanswered Questions: Maintenance, Ecology and Regulation
Running a data center on the seabed introduces operational and environmental unknowns. Maintenance is inherently high stakes: when hardware fails, technicians cannot simply walk into a server room. Entire pods must be hauled to the surface, so designs rely on redundancy, remote telemetry and an assumption of multi-year operation without physical intervention. At the same time, the ecological footprint of dissipating heat into marine ecosystems remains under-explored, from potential impacts on local biodiversity to longer-term warming of enclosed coastal zones. Regulatory frameworks for placing digital infrastructure offshore are still evolving, spanning maritime safety, environmental protection and energy market rules. As underwater data centers spread, policymakers and operators will need to establish standards for monitoring, lifecycle management and end-of-life retrieval, ensuring that efficiency gains do not come at the expense of fragile ocean environments.