NetworkOcean Operated Floating GPU Data Center
The startup launched a solar-powered floating server system in San Francisco Bay to bypass land infrastructure limits.
Updated on Oct. 2, 2026 in Data Centers
Live Poll
Should data centers transition to sea-based solar arrays to accelerate compute capacity and sustainability?
Founded in 2024, NetworkOcean has successfully operated an Nvidia H100 GPU on a floating data center in San Francisco Bay. The system uses a 20kW solar array to power compute operations, aiming to avoid traditional land-based grid requirements.
Why it matters
The venture seeks to circumvent common land infrastructure delays and grid connection hurdles by leveraging modular, water-based deployments. This approach offers a potential alternative to the rising power demands of high-performance computing.
The pilot deployment utilized 12 floating solar panels to generate 20kW of power for a container-based server. The company completed the transition from initial design to water deployment in just 53 days.
The players
NetworkOcean
This startup is a spin-out from Y Combinator that focuses on alternative data center infrastructure.
Sam Mendel
He is a co-founder of NetworkOcean.
Eric Kim
He is a co-founder of NetworkOcean.
Y Combinator
This is a well-known startup accelerator that has funded many prominent technology companies.
Nvidia
This technology company is a global leader in the design of high-performance graphics processing units.
The details
NetworkOcean previously shifted its strategy from failed underwater data center attempts to floating, solar-powered systems. Founders Sam Mendel and Eric Kim successfully deployed the current unit after the company previously struggled to secure permits for other Bay Area sites.
Timeline
NetworkOcean was founded in 2024.
The floating data center trial occurred in 2026.
A 1MW sea-based array deployment is planned for 2027.
The company targets a capacity of 1GW for compute by 2030.
The Tech Race
This project represents a push to sustain the rapid growth of high-performance computing platforms like the Nvidia H100 GPU. By decoupling infrastructure from terrestrial real estate, the company aims to bypass the physical constraints facing land-based server farms.
As high-performance computing demand grows, innovations like floating data centers could reduce the strain on local power grids and land use. Users of AI services may eventually see lower costs if infrastructure developers successfully bypass current terrestrial bottlenecks.
The takeaway
The move toward marine-based computing suggests that startup infrastructure is increasingly prioritizing mobility to avoid grid limitations. Readers should monitor whether these pilot systems can scale reliably to match the industrial capacity of traditional land-based data centers.
Further reading
Learn more about emerging infrastructure trends on the Data Centers section.
Source note: This article includes information reported by Datacenterdynamics.
Live Poll
Should data centers transition to sea-based solar arrays to accelerate compute capacity and sustainability?
