Companies

NetworkOcean

networkocean.io

NetworkOcean builds ocean data centers that deliver cheaper, more sustainable GPU compute through seawater cooling.

HQSan Francisco, California, United States
Employees1-50
Jobs checked 17h ago
Cloud InfrastructureAI InfrastructureInfrastructure

About

NetworkOcean builds and operates floating and underwater data centers, selling GPU compute, colocation, and liquid-cooled capacity to AI companies and other compute users. Its ocean-based infrastructure uses seawater cooling to reduce power consumption, eliminate freshwater usage, and lower costs versus conventional land-based facilities.

Market

NetworkOcean competes in AI GPU infrastructure and sustainable data-center markets, with a specialized focus on ocean-based, subsea, and floating facilities. It differentiates through modular underwater capsules and floating barges that use seawater cooling to reduce power consumption by roughly 30–40% and eliminate freshwater use, while supporting high-density GPU racks. Its model combines operation of ocean data-center infrastructure with direct GPU capacity reservations, positioning it against both subsea-data-center developers and conventional cooling, data-center, and GPU infrastructure providers.

Target Customers

AI developers, research teams, startups, and enterprise or cloud operators that need GPU-intensive compute and can reserve capacity from one week to large multi-GPU deployments. The likely buyers are technical founders, AI infrastructure leaders, and cloud-capacity procurement teams seeking lower-cost, more sustainable H100 capacity.

At a Glance

Problem

NetworkOcean addresses the escalating cost, power, land, and water burden of conventional data centers, particularly for AI workloads that require dense GPU capacity. Its launch materials project more than $1 trillion in data-center construction spending and over 1 trillion gallons of annual water consumption by 2030; they also estimate that building a data center costs $10–20 million per megawatt, with roughly two-thirds attributable to land, buildings, and cooling. The immediate pain is that available power and suitable real estate are increasingly constrained, while evaporative cooling consumes substantial water.

The killer use case is high-density AI compute: providing GPU capacity without forcing customers to bear the full capital cost and environmental footprint of an on-land facility. NetworkOcean’s target customers are AI companies and cloud-infrastructure users that need affordable, scalable GPU hosting, colocation, or new data-center capacity near coastal markets.

Product / Service

NetworkOcean builds and operates ocean data centers in two formats: floating data-center barges and underwater capsules. Seawater provides passive or “free” cooling, which the company says can reduce power usage by up to 40% and eliminate water consumption. The company also claims lower construction costs, faster deployment, potential co-location with offshore power, and low latency to coastal cities; its service is designed to handle high-density GPU racks rather than merely sell cooling equipment.

The delivery model combines infrastructure development with compute and colocation services. Customers can discuss new build-outs, reserve liquid-cooled space, colocate equipment, or reserve GPUs; NetworkOcean’s launch materials advertised 2,048 Nvidia H100s at no more than $2.10 per GPU-hour. Its YC profile reported that a roughly 1 MW capsule was being tested underwater in San Francisco Bay, while later reporting described a planned 500 kW test capsule, indicating that the technology was moving from demonstration toward commercial deployment.

Market

NetworkOcean competes in data-center infrastructure, green data centers, GPU colocation, and AI-cloud infrastructure. Its differentiated category is ocean-based data-center capacity, where direct precedents include Microsoft’s Project Natick and commercial underwater deployments by China’s Highlander and HiCloud. In adjacent cooling and data-center infrastructure, identified competitors include Iceotope, Schneider Electric, and Nautilus Data Technologies; conventional GPU colocation and cloud providers remain broader substitutes.

The company was founded in 2023, joined Y Combinator’s Summer 2024 batch, and is listed as active with a small San Francisco team. Public evidence shows early commercial intent and traction rather than a fully scaled network: NetworkOcean continues to offer reservations for 2,048 H100 GPUs and targets AI and cloud-infrastructure customers, while GetLatka estimates—but does not independently verify—$330,000 of 2025 revenue and reports that customer-count information is unavailable. Crunchbase classifies it as a private pre-seed company, so it should be viewed as an early-stage infrastructure business with a test deployment and reservations, not yet a proven large-scale data-center operator.

Founders & Leadership

Sam MendelFounder
CEO, Co-Founder
Eric KimFounder
CTO, Co-Founder

Funding History

2024-09
Pre-Seed (listed as Seed by Tracxn and PitchBook)$500K

Y Combinator

Recent News

2026-03-13
Underwater data center powered by tidal energy proposed off the coast of Maine

Data Center Dynamics reported that NetworkOcean wants to test underwater data centers in San Francisco Bay, but has reportedly encountered permitting issues and appears to be experimenting with floating servers.

2025-12-03
Beneath the waves: China’s bold bet on underwater data centers

W.media named NetworkOcean among several planned American underwater data-center projects, while noting that no concrete results had yet emerged from those installations.

2025-12-02
Data Center Report 2026: Key Innovations & Insights

StartUs Insights identified NetworkOcean as a startup developing ocean-based infrastructure with floating barges and underwater capsules. The report highlighted seawater cooling, reduced power usage, elimination of freshwater consumption, and support for high-density GPU racks.

Active Roles

0

No active roles right now.

Get notified when they post

Business Model

NetworkOcean sells access to GPU compute and data-center capacity, including reserved H100 GPUs, colocation, liquid-cooled space, and new build-outs. Its launch offering advertised a minimum one-week reservation at up to $2.10 per GPU per hour, with lower operating costs enabled by ocean-based cooling.

Products

Modular underwater data center capsules, including a 1 MW capsule undergoing underwater testingFloating data center bargesHigh-density GPU compute capacity, including reservations for up to 2,048 NVIDIA H100 GPUsCustom ocean data center builds and deployments

Tech Stack

Seawater coolingUnderwater data center capsulesFloating data center bargesHigh-density GPU racksNVIDIA H100 GPUs

Competitors

Subsea Cloud
Nautilus Data Technologies
Panthalassa
Iceotope
Schneider Electric