Naver Invests in Wave-Powered AI Data Centers

Dealroom reports that Panthalassa's node platforms combine wave-energy generation, AI processing, and seawater cooling in a single offshore unit. This architecture differs from conventional hyperscale facilities, which generally require access to large power supplies, land, cooling infrastructure, and terrestrial network connections.

According to Dealroom, Panthalassa has conducted sea trials of its Ocean-1 and Ocean-2 platforms and intends to deploy Ocean-3 in 2026. Yonhap likewise reported that the company intends to demonstrate its latest wave-powered platform this year, with commercialization targeted for 2027.

The investment amount and ownership terms were not disclosed.

Offshore data-center designs are being explored as AI workloads increase demand for electricity and cooling capacity. In comparable infrastructure projects, the central engineering questions include power-generation consistency, hardware reliability in corrosive marine conditions, maintenance logistics, satellite-network latency and throughput, and compliance with maritime and environmental requirements.

For ML infrastructure teams, Panthalassa's approach is most relevant to workloads that can tolerate remote execution and network constraints. Training clusters and latency-sensitive inference systems typically depend on high-bandwidth, low-latency interconnects, while batch-oriented processing, asynchronous inference, and specialized edge workloads can have different deployment tradeoffs. Panthalassa's planned demonstration and subsequent commercial deployment would provide operational evidence about how those tradeoffs perform outside a land-based data-center environment.