Samsung Wants to Put AI Data Centers on the Water — And the Reason Goes Far Beyond Saving Land
The next generation of artificial intelligence infrastructure may not look like a traditional data center at all.
Instead of another enormous warehouse filled with servers, transformers, cooling equipment and miles of electrical infrastructure, part of the future computing network could sit on a floating platform.
That is the idea behind a new project from Samsung Heavy Industries (SHI), which is developing floating data centers designed to operate on water.
The concept sounds futuristic, but it is already moving beyond the idea stage.
Samsung Heavy Industries has obtained approvals for a 50-megawatt floating data center design from classification societies and has been building partnerships with companies involved in data centers, shipping, engineering and regulation. In 2026, SHI also signed an engineering agreement with U.S.-based data center developer M3 for floating facilities intended for the U.S. market.
The company has said it is pursuing multiple projects with the goal of commercializing its floating data center platform by the second quarter of 2028.
The idea is being driven by a simple but increasingly difficult problem:
AI needs an extraordinary amount of infrastructure.
Training and running advanced AI models requires powerful computing hardware. That hardware consumes electricity, produces heat and requires sophisticated cooling systems.
At the same time, finding suitable locations for new data centers is becoming increasingly complicated.
Developers need land.
They need access to enormous quantities of electricity.
They need reliable network connections.
And they need ways to remove the heat generated by thousands of high-performance processors.
Samsung's answer is to move at least part of the equation onto the water.
Why AI Is Changing the Data Center Problem
Data centers have existed for decades, but the rapid development of generative AI is changing the scale of the challenge.
Traditional data centers were already significant consumers of electricity. Modern AI facilities can be considerably more demanding because they are designed around dense clusters of specialized accelerators and GPUs.
The International Energy Agency estimates that global data center electricity consumption reached about 415 terawatt-hours in 2024 and projects that it could rise to around 945 TWh by 2030 in its base case. That would mean data centers consuming close to 3% of global electricity demand by the end of the decade.
AI is a major reason for the increase.
The IEA says electricity consumption by accelerated servers, which are mainly driven by AI adoption, is projected to grow by about 30% annually through 2030 in its base case.
That growth creates a chain reaction.
More AI applications require more computing.
More computing requires more servers.
More servers require more electricity.
More electricity requires more generation and grid infrastructure.
And more powerful servers generate more heat that has to be removed.
The result is that building an AI data center is no longer simply a matter of finding a large empty building.
It can become an infrastructure project involving electricity generation, transmission, cooling, construction, networking and environmental considerations.
That is where the floating data center concept begins to make sense.
The Basic Idea Is Surprisingly Simple
A floating data center is essentially what its name suggests.
Instead of constructing a permanent data center entirely on land, the computing infrastructure is installed on a floating structure such as a barge or ship-like platform.
The platform can contain server rooms, electrical systems, cooling equipment and other infrastructure required to operate the facility.
Samsung Heavy Industries says its FDC concept is designed to address several limitations of traditional land-based facilities, including land availability, electricity supply and cooling.
The approach also allows Samsung to apply technologies it has developed through decades of offshore engineering and shipbuilding.
That is an important part of the strategy.
Samsung Heavy Industries is not starting from zero.
The company already knows how to design and manufacture large offshore structures that must operate in demanding marine environments.
The challenge is to adapt those capabilities to a completely different payload:
AI computing infrastructure.
The Ocean Provides Something Data Centers Constantly Need
There is an obvious reason why water is attractive for cooling.
Water is extremely effective at carrying heat away.
A modern data center can use a range of cooling technologies, including air cooling, direct-to-chip liquid cooling and immersion cooling.
As computing power increases, liquid-based cooling becomes increasingly attractive because liquids can transfer heat much more efficiently than air.
Samsung C&T, another Samsung group company, has previously examined underwater and seawater-based data center cooling and has also developed immersion-cooling technologies for data centers. Samsung C&T has reported that its immersion cooling approach can substantially reduce electricity consumption compared with conventional air cooling.
The floating data center concept takes advantage of the surrounding water environment.
Rather than depending entirely on conventional land-based cooling infrastructure, the facility can use water as part of its thermal-management strategy.
However, the precise Samsung design is important.
Recent reporting on the first U.S. projects describes a non-evaporative cooling system that does not consume drinking water and does not discharge wastewater into surrounding waters.
So the concept should not simply be described as "putting servers in seawater."
The servers themselves are not sitting directly in the ocean.
Instead, the floating facility contains the computing and cooling systems, while the surrounding water becomes part of the thermal environment that helps the system reject heat.
That distinction matters when discussing the environmental claims.
Why Freshwater Has Become an Important Issue
Cooling is one of the biggest environmental considerations associated with data centers.
Depending on the design, location and cooling technology, a data center may require significant quantities of water.