Can floating data centres meet AI's huge energy demand?
Oregon-based startup Panthalassa is developing autonomous floating data centers powered by wave energy to meet the growing demands of AI, having recently secured $140 million in funding. These offshore platforms aim to bypass traditional grids and offer carbon-free computing, but experts raise concerns about the harsh marine environment, maintenance challenges, and potential latency issues. The concept, while innovative, faces significant hurdles in proving its economic and operational viability against land-based data centers.
Panthalassa, an Oregon-based start-up, recently secured $140 million in funding to develop autonomous floating data centers designed to operate in international waters. This innovative approach seeks to address the escalating energy demands of artificial intelligence, projected to reach 945 terawatt-hours annually by 2030, exceeding Japan's entire electricity consumption. The company proposes that these platforms, which generate their own electricity from wave power, can circumvent overloaded electrical grids and provide carbon-free computing without traditional emissions or engines.
However, experts like Jonathan Koomey, formerly of Lawrence Berkeley National Laboratory, caution about the extreme challenges posed by the marine environment, citing salt and waves as major threats to machinery. Jacqueline Davis of the Uptime Institute also highlights the difficulties of managing critical infrastructure remotely, noting that automation in data centers is typically limited to monitoring, with human intervention still crucial for many issues. The floating data centers, 85 meters tall and made of plate steel, cool servers by dissipating heat into the surrounding cold ocean water, though their impact on marine ecosystems remains underexplored.
Another significant hurdle is latency, as data will be transmitted via Starlink satellites, offering limited bandwidth compared to fiber-optic cables. This makes the technology more suitable for long-running AI workloads like model training or scientific simulations rather than consumer applications requiring rapid responses. While Panthalassa’s concept is unique, other companies like Aikido Technologies and Mitsui O.S.K. are also exploring offshore computing. Critics argue that despite potential benefits, ocean-based systems must still prove their economic competitiveness against conventional land-based data centers, which benefit from economies of scale.