Samsung and a Dallas-based infrastructure company are betting that the future of AI computing floats on water. The two organizations unveiled a floating data center concept designed to host AI systems like OpenAI's ChatGPT on offshore barges that connect directly to coastal power sources, potentially cutting deployment timelines from years down to quarters.

Why Are Tech Companies Moving AI Data Centers to the Ocean?

The challenge facing AI companies is straightforward but urgent: data centers consume enormous amounts of electricity, and traditional power grids cannot keep pace with demand. According to the U.S. Department of Energy, data centers are expected to consume between 6.7 and 12 percent of all U.S. electricity by 2028. Rather than wait for utilities to build new grid connections, Samsung and Mousterian Corp., the Dallas-based partner, propose anchoring fully equipped computing vessels near existing power plants, particularly those with thermal or nuclear generation capacity.

The floating model addresses a critical bottleneck in AI infrastructure: the time required to secure both power and cooling infrastructure. Traditional land-based data centers can take years to plan, permit, and connect to the grid. Offshore barges, by contrast, can dock near established energy sources and begin operations far more quickly.

"Speed to power is the new moat. We've thoughtfully partnered with some of the leading global conglomerates, allowing us to deliver over 1,500MW of capacity over the next 3 years," said Min Suh, CEO of Mousterian Corp.

Min Suh, CEO at Mousterian Corp.

The partnership between Samsung and OpenAI, formalized through a letter of intent in October 2025, signals confidence in the concept. Each floating vessel would house thousands of servers optimized for AI training and inference tasks, with liquid cooling systems built directly into the barges.

What Technical and Operational Challenges Could Slow This Down?

Despite the appeal of faster deployment, floating data centers introduce complexities that land-based facilities do not face. Saltwater environments corrode equipment, storms threaten physical infrastructure, and emergency response times are longer when facilities sit offshore. Cybersecurity risks also expand when computing infrastructure is isolated on water, and maintaining fiber optic connections to shore requires specialized engineering.

The ambitious timeline of delivering 1.5 gigawatts of capacity within 36 months depends on several unproven variables. Shipbuilding capacity, regulatory approvals, and the availability of suitable coastal locations near baseload power plants all present potential delays. Some industry analysts express skepticism about whether the pace can be sustained in practice, noting that the model may serve as a niche option rather than fundamentally reshape how most AI computing infrastructure is built.

How Could Floating Data Centers Change AI Infrastructure Deployment?

• Faster Power Access: Barges dock near existing thermal or nuclear plants, eliminating years of grid connection delays and allowing compute capacity to come online in quarters rather than years.
• Reduced Permitting Burden: Offshore deployment treats the shoreline as a flexible infrastructure zone, potentially requiring fewer land-use approvals than traditional data center construction.
• Scalable Cooling Solutions: Fully liquid-cooled data halls built into vessels can scale according to demand without requiring separate cooling infrastructure buildouts.
• Proximity to Baseload Energy: Positioning barges near nuclear or thermal plants ensures access to reliable, continuous power sources that can support the constant electricity demands of AI workloads.

The floating data center concept arrives at a moment when nuclear power is gaining renewed attention as a solution to AI's energy crisis. The University of Utah announced plans to produce electricity from its TRIGA research reactor for the first time, with the power feeding a small AI data center in partnership with Elemental Nuclear Energy. While that project generates only 2 to 3 kilowatts, it demonstrates growing interest in pairing compact nuclear systems with AI infrastructure.

The broader energy infrastructure challenge remains daunting. Utilities and grid operators face pressure to accommodate surging electricity demand driven by data centers, electric vehicles, and industrial electrification. ABB, a major provider of electrical equipment and grid modernization solutions, emphasized that successful infrastructure upgrades require balancing safety, affordability, sustainability, and reliability while modernizing existing systems rather than replacing them entirely.

Samsung's floating data center model represents one potential answer to the infrastructure bottleneck, but its success will ultimately depend on execution. The true test will be how many barges actually come online as planned and whether the promised timelines hold up against real-world shipbuilding, permitting, and operational challenges.