Riding the Wave: The Rise of Floating Data Centers

extract

Riding the Wave: The Rise of Floating Data Centers Floating and ocean-based data centers are emerging as innovative solutions to address land, power, and cooling constraints in dense coastal markets, offering efficiency gains through proximity to coastal infrastructure and water-based cooling. The race to build AI-scale data centers is running into familiar obstacles: scarce land, constrained power, and rising cooling demand. One emerging response is to move compute onto the water. Floating and ocean-based data centers shift capacity closer to coastal infrastructure, where power, connectivity and cooling are more readily available. Rather than solving a single bottleneck, these designs aim to address multiple pressures at once – particularly in dense, land-constrained markets like Singapore and other coastal hubs. Why the Water Matters Proponents argue that water-based or nearshore infrastructure can deliver material efficiency gains as high-density AI workloads strain traditional air-cooled systems. Early deployments – ranging from floating facilities reporting low power usage effectiveness (PUE) and subsea experiments like Microsoft’s Project Natick – suggest improved thermal performance and, under the right conditions, reliability. Floating Above Permitting Pressures Mohammad Faisal Ahmad, CEO of BIS Research, told Data Center Knowledge that floating and ocean-based data centers are a response to land and permitting pressures in dense coastal hubs. He cited Singapore as a leading example. In 2023, IMDA and EDB awarded about 80 MW of new capacity through the pilot Data Centre Call for Application as part of a sustainable growth strategy. Singapore’s Green Data Centre Roadmap (2024) sets a near-term goal of at least 300 MW of added capacity, with potentially another 200 MW for green-energy deployments. In December 2025, they launched a second call with at least 200 MW of additional capacity, with the possibility of more through innovative green-energy pathways. “That suggests the issue is not simply demand growth, but how new capacity can be added within land, energy, and sustainability limits,” Ahmad noted. This dovetails with rising power demands. According to the IEA’s 2025 analysis, data centers consumed about 415 TWh of electricity in 2024 – roughly 1.5% of global electricity use – and demand is projected to more than double to around 945 TWh by 2030. “The IEA identifies AI as the most important driver of this increase,” Ahmad said. “In that context, floating data center concepts are attractive not because they create new power on their own, but because they can be located near coastal grid infrastructure, submarine cable routes, and, over time, offshore or other low-carbon energy sources.” Thermal management is the third pressure point. High-density AI and HPC workloads are pushing beyond the limits of conventional air cooling; water-cooled infrastructure can offer substantial advantages under the right conditions. A BIS report highlights water-based cooling, proximity to coastal areas, and flexible deployment as key factors in the appeal of floating and underwater data centers. Nautilus’ Stockton 1 site is one of the most concrete commercial examples: a 6.5 MW floating facility the company describes as having a 1.15 PUE, no cooling towers, and no water consumption. A fourth factor is siting friction. Large land-based campuses increasingly face scrutiny for power demand, water use, and physical footprint, especially in land-constrained or environmentally sensitive regions. Keppel Data Centres' Floating Data Centre project features modular design and seawater cooling. (Image: Keppel Data Centres) “Floating facilities do not entirely resolve those issues, but they can shift the siting discussion to port, river, or nearshore industrial zones where access to cooling water and connectivity may already exist,” Ahmad explained. In practical terms, he said, they are most compelling in land-constrained coastal markets, island economies, port cities, and specialized edge environments where cooling, connectivity, and sustainability pressures overlap – and far less compelling in inland markets with abundant land, where marine engineering doesn’t address an acute bottleneck. Microsoft’s Project Natick tested the reliability and thermal performance of a subsea data center. Here, the capsule is being power-washed after two years on the seafloor near Scotland’s Orkney Islands. (Image: Microsoft) Engineering Reality: Above and Below the Waterline Floating data centers introduce engineering and operational complexity, from marine hardening and maintenance logistics to power delivery and dependence on subsea cables. Sam Kanner, co-founder and CEO of Aikido Technologies, underscored a key unknown: how IT equipment behaves under low-frequency motions on floating platforms. He explained that all IT equipment is rated for a certain amount of acceleration at high frequencies (5-10 Hz+ related to shocks/drops), but there is very little testing at low frequencies. He noted that platform- and wave-dominant periods often fall in the 8-120 second range (0.1-0.01 Hz). “Furthermore, there is sensitive electronic equipment up in the nacelle – at full scale, this is nearly 500 feet above the waterline, where accelerations are much more violent,” Kanner said. “We design the platform such that this equipment stays within its operational envelope.” With isolation pads to dampen high-frequency vibration, the team is confident that the IT equipment closer to the waterline will also stay within its design envelope. Aikido Technologies recently unveiled a floating offshore wind platform that co-locates AI-grade compute with wind generation and battery storage, targeting 10–12 MW of IT load per unit alongside a 15–18 MW+ turbine. The concept could scale from 30 MW to more than 1 GW of IT load across wind farms. “Countries want to be able to control their own datacenters and AI industries,” Kanner said, arguing that floating designs can become a meaningful segment of the market. Aikido Technologies has developed a floating wind turbine platform equipped with three ballast tanks, each capable of housing a 4 MW liquid-cooled data center. (Image: Aikido Technologies) Practical Skepticism and Power Realities Others urge caution. “As a former data center manager, I’d be worried about the unique operational environment,” said Sean Farney, vice president of data center strategy at JLL. “The data center industry is generally new to floating environments, and we’d have to adjust operations and maintenance accordingly, which takes some time.” Farney also flagged the difficulty of delivering permanent, large-scale power (100 MW+) to a vessel that does not generate its own electricity. “From an operational perspective, I don’t see a compelling reason to employ such a different form factor vis-à-vis traditional methods, quite frankly, particularly if I compare it to deploying a modular solution.” Ahmad agreed that floating sites must still satisfy the core electrical standards of any mission-critical facility: dependable primary power, redundancy, backup resilience, and safe distribution architecture – now integrated with marine constraints. Nearshore systems depend on robust shore interconnections and reliable transmission to the platform. Remote offshore concepts face harder questions around on-site generation, energy storage, fuel logistics, and long-term resilience. Barriers remain substantial, Ahmad said: marine engineering costs, environmental review, cable dependence, climate hardening, servicing complexity, and a limited number of economically attractive sites. Meanwhile, land-based data centers continue to evolve with stronger liquid-cooling, improved power planning, and further campus design optimization. Panthalassa tested its Ocean-2 wave energy converter to generate power for sea-based data centers. (Image: Panthalassa) Niche Today, Material Tomorrow Floating and offshore data centers have moved beyond experimentation. “They are credible, site-specific options as land-based sites become harder to secure, power, and permit,” Ahmad said. “The category will matter more over time, as the external pressures driving interest in alternative infrastructure are intensifying, not easing.”