Samsung Brings Artificial Intelligence to the Sea
Floating AI Data Centers Are Coming in the Middle of the Ocean
When we talk about the future of artificial intelligence, we usually talk about more powerful models, new AI applications, or systems that automate jobs done by humans. But as artificial intelligence grows, a much more physical problem arises: Where will all these models work?
There are huge data centers behind the technologies we use, from large language models like ChatGPT to AI video producers, from corporate artificial intelligence systems to the AI agents of the future. Thousands of high-performance GPUs and servers are running simultaneously in these facilities. More computing power means more electricity, more cooling capacity and more physical infrastructure.
As artificial intelligence develops, the problem of data centers can no longer be solved by just saying "let's buy more servers". Finding suitable land, establishing a high-capacity connection to the electricity grid, and removing the enormous heat generated become some of the most critical infrastructure problems of the new AI economy.
Samsung's answer to this problem is quite extraordinary:
Removing the data center from land and moving it to the sea.
The Floating Data Center, or FDC concept, developed by Samsung Heavy Industries, aims to place artificial intelligence computing infrastructure on specially designed floating platforms. The first design was developed in the 50 MW class and received Approval in Principle from the American Bureau of Shipping (ABS) and Lloyd's Register in 2026. Samsung later expanded the project and received AiP for the Floating Data Center design in the 200 MW class.
At first glance, it may seem like a science fiction movie.
But the engineering problem behind it is very real.
Why Does Artificial Intelligence Need So Many Data Centers?
The least visible aspect of Generative AI is its physical infrastructure.
A user writes a few words to the artificial intelligence and receives a response within a few seconds. When we look from the front of the screen, the process looks almost abstract. However, a large amount of calculations are performed in the background.
Especially training large AI models and then providing inference services to millions of users requires huge clusters of high-performance processors.
Demand for computing is also growing as AI image and video production becomes more widespread, companies build their own artificial intelligence systems, and models become increasingly multimodal.
Samsung Heavy Industries also explains the Floating Data Center project directly through this change. The company stated that the demand for data center infrastructure is growing rapidly with the spread of AI technologies; He states that traditional land-based data centers face problems such as land, electricity supply and cooling efficiency.
So the new race in the world of artificial intelligence is just:
“Who will develop the best AI model?”
not.
Also:
“Who will build the infrastructure that will run all these models?”
race.
Samsung Heavy Industries is preparing to enter this race through shipbuilding engineering.
What Exactly Is Samsung's Floating AI Data Center?
The system developed by Samsung is not simply a matter of placing a few servers inside an existing cargo ship.
The company is designing the Floating Data Center platform from the beginning for this purpose.
According to Samsung's official statement, the system takes advantage of the company's standardized shipbuilding processes. Since design, production and equipment integration can be carried out in parallel, the aim is to shorten the construction time compared to traditional land data centers.
The idea here is actually quite interesting.
Normally, when you want to establish a large data center, you first need to find suitable land, complete the permit processes, prepare the electrical infrastructure, construct the building, install the cooling systems and then place the server infrastructure.
Samsung, on the other hand, wants to apply the industrial production logic that the shipbuilding industry has used for years to data centers.
In a sense, the data center is turning into a digital infrastructure that can be produced in a factory.
It can then be deployed to the needed area.
If this approach is successful, it can transform the concept of the data center from a structure tied to real estate to a more modular and mobile infrastructure.
How Big is 50 MW?
The capacity of Samsung's first Floating Data Center design is 50 MW.
The MW value here is important for understanding the electricity and IT scale of the data center. An AI data center in the 50 MW class is not a small server room or a system consisting of a few racks; stands for large-scale AI computing infrastructure.
But Samsung did not stop there.
Two important developments were announced in the new agreement announced between ABS and Samsung Heavy Industries on August 4, 2026. ABS will carry out further design evaluation of the previously AiP awarded 50 MW design in line with applicable class and IMO guidelines. In the same statement, it was announced that Samsung's 200 MW Floating Data Center design also received Approval in Principle.
The 200 MW rating is an important sign that Samsung does not view the Floating Data Center as just an experimental niche project.
We are talking here about the idea of an increasingly scalable offshore AI infrastructure.
50 MW today.
Then 200 MW.
If the concept proves commercially successful, deploying multiple platforms in the same regions in the future could theoretically pave the way for much larger offshore AI clusters.
So the “AI campus” of the future may not necessarily have to be in the Nevada desert or a huge industrial area.
It could also be on the sea.
So Why Do We Put the Data Center at Sea?
The first answer to this question is quite simple:
Because some of the biggest problems of data centers occur on land.
Large data centers require significant amounts of land. In addition, they demand very high capacity from the electricity grid. Although there is land in some regions where a data center can be established, providing the necessary electrical connection on time can be a problem in itself.
Samsung Heavy Industries positions the Floating Data Center concept as an alternative to the problems of land availability, electricity supply and cooling efficiency.
On the sea, a different equation emerges.
First, there are no restrictions on physical space as great as on land.
Second, the platform may have its own electricity generation systems.
Third, and perhaps most importantly:
There is one of the world's largest natural cooling systems around the data center.
Sea.
The Ocean Could Turn into a Giant Cooler
The heat generated by servers while operating is one of the most important problems of data centers.
To run a GPU cluster, it is not enough to simply provide electricity to the GPUs. The heat produced must be constantly removed. For this reason, cooling infrastructure in data centers is a serious engineering and energy item.
One of the most logical aspects of the Floating Data Center approach emerges here.
Seawater can be used as a natural heat sink.
In Samsung's concept, it is planned to use sea water for cooling purposes. Thus, a different alternative can be created to the cooling infrastructure used in large land facilities.
This is especially important in the age of artificial intelligence.
Because as the density of AI processors increases, it's just "how much electricity do we use?" not the question,
“How do we remove this heat?”
The question is also growing.
In a sense, instead of taking the data center to the cooler, Samsung is placing the data center in the middle of the cooler.
Why is it important to use the sea instead of fresh water?
One of the growing debates about data centers is water consumption.
Especially in regions where water resources are limited, the cooling needs of large data centers can put pressure on local resources.
One of the potential advantages of the Floating Data Center concept is that it can reduce dependence on fresh water.
But it is necessary to make an important distinction here. When we say "sea water cooling", we should not think that salt water is passed directly through the server racks. Seawater can be used as part of heat exchange systems; The electronic equipment itself needs to be protected from the salty environment.
And this is exactly where one of the most difficult engineering problems begins.
Because the Sea is Not a Friendly Environment for Electronics
The sea has advantages for data centers.
But the sea is also one of the most difficult environments in which electronic systems can operate.
Salt.
Humidity.
Corrosion.
Vibration.
Wave motion.
Continuous slope changes.
All of these can pose serious problems for sensitive AI servers that are expected to operate uninterrupted for years.
Samsung Heavy Industries makes clear in its statement that offshore environments can subject AI servers to vibration, tilt, high salinity air, and rapid humidity changes, which can affect server stability and lifespan.
Therefore, the "we put a GPU inside the ship and ran it" approach is not enough.
The real question is:
Can you reliably run thousands of high-performance AI processors in a moving, humid and salty environment for years?
Samsung has chosen an important partner to find the answer to this.
Samsung and Supermicro Test AI Servers at Sea
Samsung Heavy Industries signed a Joint Development Project agreement with US-based AI server specialist Supermicro in June 2026.
The division of labor is quite meaningful.
Samsung knows marine engineering.
Supermicro provides high-performance AI server infrastructure.
According to Samsung's official statement, the joint work includes verifying the operating conditions of AI servers in river and marine environments. Samsung is working on offshore positioning control and salinity and humidity isolation technologies, while Supermicro AI is verifying the operating conditions of its servers on the water.
These tests are perhaps the most critical phase of the project.
Because you can solve the cooling problem.
You can produce electricity.
You can build the ship.
But if the servers do not work reliably in sea conditions, the whole concept becomes meaningless.
For this reason, what will determine the future of the Floating Data Center will be long-term operational tests rather than impressive concept visuals.
So Where Will It Find Electricity?
Operating 50 MW or, in the future, 200 MW class AI infrastructure is a huge energy problem.
Samsung's concept does not depend on a single source here either.
When the platform is deployed near the coast or in areas with suitable infrastructure, it can be connected to external electrical systems via submarine cables. In the other scenario, electricity can be produced on the platform.
In the technical information announced about the project, solid oxide fuel cell, or SOFC systems that use LNG, stand out.
The strategic advantage of this model is that the data center is not dependent on a major onshore electricity connection in any case.
Samsung's official statement also states that dependence on land-based electricity can be reduced by establishing FDC's own production systems.
This can create significant flexibility in where data centers can be located.
But the use of LNG also prevents the project from being interpreted as an “all-green AI data center”
More efficient cooling with seawater could provide a significant environmental advantage; However, the carbon impact of the energy source must be evaluated separately.
For this reason, it would not be correct to evaluate the environmental performance of the Floating Data Center solely on cooling.
Why Did Samsung Get Into This Business?
At first glance, it may not be surprising to see the words Samsung and data center side by side.
But the company developing the project here is Samsung Heavy Industries, not Samsung Electronics.
In other words, the company of the Samsung group specializing in ship and offshore platform engineering.
Actually, this is exactly why the project makes sense.
Samsung Heavy Industries has been designing massive ships, LNG carriers and offshore systems for years. Floating Data Center offers the opportunity to combine the company's existing marine engineering expertise with one of the fastest growing infrastructure problems of the AI era.
The company's CTO, Young-kyu Ahn, also defines FDC as a new business model in which shipbuilding technology expands into the digital infrastructure sector.
This sentence actually summarizes the importance of the project quite well.
Samsung no longer wants to just produce ships.
It wants to produce the physical platform on which the digital infrastructure will run.
Samsung is not doing this project alone
The partner network behind the Floating Data Center also shows how seriously the project is taken.
Samsung Heavy Industries has established collaborations with ABB on the energy and automation side and Mousterian Corporation, which develops near-water/floating data center infrastructure, within the scope of FDC development efforts in the US market.
In June, a new agreement was made with Greece-based Capital and Lloyd's Register.
In this business division, Samsung undertakes technology development and construction, Capital takes part in project development and investment, and Lloyd's Register takes part in regulations and rules. In addition, we are collaborating with LR Advisory on the North American data center market, infrastructure and commercial feasibility studies.
Supermicro comes into play on the AI server side.
So Samsung isn't just designing a ship.
The energy company is trying to create a Floating Data Center ecosystem consisting of server manufacturers, investors, ship operators, certification bodies and infrastructure developers.
This is critical for the project to move from concept drawing to real commercial product.
Samsung's New 200 MW Design Ups the Game
In the June news you sent, a maximum of 50 MW platform is discussed.
But an important development took place in August.
ABS and Samsung Heavy Industries announced an agreement to move the 50 MW FDC design to further design approval on August 4, 2026.
At the same time, Approval in Principle was given for the 200 MW Floating Data Center.
This development changes our perspective on the project.
The 50 MW system can be considered a demonstration or the first commercial product.
200 MW is a much more serious-scale AI infrastructure.
Furthermore, the modular scalability of the shipbuilding-based approach developed by Samsung means that capacity can be increased according to different needs in the future.
If demand for AI computing continues to grow, offshore data centers could evolve into digital infrastructure enclaves consisting of multiple platforms working in close proximity, not just single ships.
Today we see offshore oil platforms.
We may see offshore compute platforms in the future.
So How Will Internet Connection Be Provided?
This is one of the natural questions that come to users' minds when considering the idea of a floating data center.
After all, putting thousands of GPUs in the middle of the sea is not enough. The data produced and processed must communicate with the rest of the world at extremely high speed.
Therefore, positioning FDCs close to fiber infrastructure and submarine communication cables is one of the logical scenarios.
A significant portion of the world's intercontinental internet traffic is already carried via submarine fiber optic cables. Therefore, the sea is not a completely foreign environment to digital infrastructure.
But this is where the location of the data center becomes critical.
While millisecond end-user latency may be less important in some high-computation tasks, such as AI training, latency is much more critical in real-time services.
Therefore, the location of FDCs in the future cannot be determined solely by the logic of “we found an empty place in the sea”.
Energy, fiber connection, port infrastructure, sea conditions, security and distance to the user have to be considered together.
Can a Floating Data Center Really Move?
Technically, the ability to transport a floating platform to another region is an important advantage.
However, we should not think of this as "an AI ship operating in Istanbul today and Singapore tomorrow."
Large data centers have energy, fiber, operation, maintenance and security connections. Dismantling them and reinstalling them in another location is not simple.
Therefore, mobility here means the redeployment of infrastructure rather than daily mobility.
This is still a significant difference compared to the traditional data center.
When a building is built in the wrong place, you cannot move it to another country.
A floating platform theoretically has this option.
This may have strategic value, especially in the future when energy prices, regulations, or demand for AI infrastructure vary regionally.
How Will the Security of the Data Center in the Middle of the Sea Be Ensured?
One of the least talked about but most important issues of the FDC idea is security.
Physical access in onshore data centers is tightly controlled.
Different risks arise at sea.
Extreme weather events, storms, ship traffic, sabotage, fire, energy failures and physical security are among them.
Cyber security is also added to this.
Therefore, the offshore data center may have to be protected like a ship, an energy facility, and critical digital infrastructure.
For this reason, it is important that maritime certification bodies such as ABS and Lloyd's Register are involved in the project. Floating Data Center needs to comply not only with IT standards but also with marine and offshore security standards.
What If There's a Storm?
It is impossible not to ask this question.
Imagine a platform with billions of dollars of AI hardware in the middle of the sea.
The storm is starting.
The waves are rising.
The movement of the ship is increasing.
Servers must operate without interruption.
For this very reason, conditions such as vibration and inclination, that is, vibration and tilt, are examined in the tests performed with Supermicro.
The engineering problem here is not just that the ship won't sink.
Highly sensitive electronic systems inside server racks need to be kept in acceptable operating conditions.
The movement that a ship can tolerate is not the same as the movement that an AI server can tolerate.
This is exactly the reason for the combination of Samsung's marine engineering expertise and Supermicro's server engineering.
What are the Disadvantages of Floating Data Centers?
No matter how impressive the idea is, there are serious problems that still need to be solved.
The first problem is corrosion and moisture.
The second problem is maintenance. Accessing faulty equipment in an onshore data center does not cost the same as accessing the system on an offshore platform.
Third is bad weather conditions.
The fourth is energy infrastructure.
Fifth is fiber connection and latency.
Sixth is the possible environmental impacts on the marine ecosystem.
Seventh is security.
And perhaps most importantly, economic feasibility.
Just because a technology is engineeringly feasible does not mean it makes commercial sense.
Samsung's FDC concept has passed important certification stages today, but it is not yet at the commercial maturity level where we can say "Samsung AI ships are running around the world."
Therefore, instead of describing the project as a finished revolution, it is more accurate to evaluate it as a new infrastructure model that has been developed seriously and has high potential.
When Will Samsung's Floating Data Center Become Available?
Here, too, it is necessary to maintain the difference between the concept and the real product.
Samsung has made significant progress in terms of design approvals, partnerships, and operational verification efforts with Supermicro. However, large-scale commercial use has not yet begun.
In the industry news in the summer of 2026, targets around 2028 are mentioned for the first platform; However, this date should not be considered as a definitive guarantee of entry into service.
So we're not talking about a commercial fleet of Samsung AI data centers operating in the middle of the sea right now.
What we are seeing now is the preparation of the infrastructure for this.
And perhaps this is the more interesting part.
Because Samsung is not competing in an existing market yet.
Trying to create a new market.
The Artificial Intelligence Race Is No Longer Just a Race for Software Companies
There is a bigger story that Samsung's Floating Data Center project tells us.
In the early stages of the AI revolution, we had our eye on OpenAI, Google, Anthropic and other model developers.
Then NVIDIA and AI chips started to be talked about.
Now the third layer of the race is becoming increasingly visible:
infrastructure.
Who will produce the GPU?
Who will provide electricity?
Who will cool the servers?
Who will build the data center?
Who will provide the fiber connection?
And where will we put this infrastructure?
This is exactly where Samsung Heavy Industries comes into the game.
A ship manufacturer is not developing an artificial intelligence model.
It redesigns the physical world on which AI will operate.
This shows us how broad an industrial transformation the artificial intelligence economy can create in the coming years.
Can an “AI City” Really Be Established in the Middle of the Ocean?
It is too assertive to say yes to this today.
But it's not all science fiction anymore.
Samsung received certification for the 50 MW design.
AiP received for 200 MW design.
Supermicro is testing AI server operations in the marine environment.
ABB is included in the ecosystem on the electrical systems side.
Capital is on the investment and project development side.
ABS and Lloyd's Register work in certification and regulation processes.
The pieces are slowly coming together.
If the first commercial FDC proves economically successful, scaling up this model could yield even more interesting results.
A platform.
Second platform next to it.
Then the third.
Common energy infrastructure.
Submarine fiber connections.
Tens of thousands of AI processors.
At this point, we start talking about a compute cluster built on the sea, not a single data center.
Perhaps some AI models of the future will actually be trained on the ocean.
The Future of Artificial Intelligence is Much More Physical Than We Thought
We often see artificial intelligence as an abstract technology.
We write prompt.
The answer is coming.
An image is being formed.
A video is being produced.
But behind the screen lies a very physical world of steel, copper, silicon, electricity, water, fiber optic cables, power plants and massive data centers.
As AI grows, this world needs to grow as well.
Samsung Heavy Industries' Floating Data Center project is therefore not just an interesting ship project in our opinion.
It is one of the signs that the age of artificial intelligence has moved beyond the software world and started to transform energy, construction, shipping and global infrastructure.
The question Samsung is trying to solve is simple but huge:
If finding enough space, energy and cooling for AI on land is increasingly difficult, why not use the sea?
We don't know yet if this is the right answer.
There are serious problems to be solved, such as the impact of the marine environment on electronics, maintenance costs, connection infrastructure, security, energy model and economic feasibility.
However, Samsung's progress from the 50 MW concept to the 200 MW design, testing real operating conditions with Supermicro, and establishing an ecosystem of international energy, investment and certification companies shows that the project has gone further than a simple concept study.
The next big battle of the artificial intelligence race may not only be fought to produce smarter models.
It can also be given to find the electricity, cooling and physical space to run those models.
And this time, the stage of the race may not be a technology campus, but the middle of the ocean.
Nur Oğuz