[Science Scope] The Hidden Side of AI: Consuming Both Electricity and Water

The Secret of Data Centers
To the Arctic, Under the Sea, in Abandoned Mines, and Beside Nuclear Plants
The AI Race Is About Power, Cooling, and Location—Not Algorithms
Even 'Space Data Centers' Are Emerging

On screen, generative artificial intelligence (AI) is perceived as a digital technology that answers questions and creates images. For the public, AI still feels like an invisible, intangible technology. However, the story changes when you enter the world of infrastructure. The data centers that power AI are highly physical industrial facilities. They consume massive amounts of electricity and generate significant heat. This is why the technological race sparked by generative AI is expanding beyond semiconductors and algorithms, evolving into a competition for electricity, cooling, and site acquisition.

A data center that withstood two years under the sea. Images show the 'Project Natick' data center, which Microsoft installed on the seabed of the Orkney Islands in Scotland, being raised after two years (left) and undergoing internal inspection (right). It drew attention as the world's first underwater data center experiment utilizing cold seawater as a natural coolant. Provided by Microsoft (Jonathan Banks).

Why Are Data Centers Heading to Colder Regions?

The International Energy Agency (IEA) has projected that global data center electricity consumption will reach approximately 945 terawatt-hours (TWh) by 2030. This is close to Japan's total annual electricity usage. The IEA has analyzed that the spread of generative AI will be a key driver behind the rising power demand of data centers. The recent trend of data centers around the world moving to the Arctic, submerging underwater, or burrowing deep into abandoned mines vividly illustrates this phenomenon.

The first places that AI data centers look for are 'cold regions.' This is because they can reduce both the cost and electricity required for cooling servers. In Lulea, northern Sweden, a large-scale data center operated by Meta is located. This area, adjacent to the Arctic Circle, utilizes cold outdoor air all year round for server cooling. Meta chose this location as its first major data center site outside the United States due to both the low temperatures and the abundant supply of renewable energy.

A panoramic view of Meta's data center located in Luleå, northern Sweden. This region, close to the Arctic Circle, is considered a prime data center location for global big tech companies as it leverages low temperatures year-round to reduce power consumption for server cooling. Provided by Meta

The Google data center in Hamina, Finland, is also unique. Google uses the old paper mill’s seawater tunnel to cool servers with cold water from the Baltic Sea. This is an example of an industrial-age facility being transformed into a key piece of infrastructure for the AI era by combining it with the natural environment.

In the past, the most important conditions for a data center site were network accessibility and proximity to users, since reducing network latency was paramount. However, with the advent of the generative AI era, the formula for site selection is changing.

No Sangmin, head of Naver Cloud Data Center, explained, "For training infrastructure that repeatedly trains large-scale models, large-scale electricity supply, cooling efficiency, and site expandability are more important than real-time response speed. These training clusters are likely to follow power and cooling conditions."

On the other hand, inference infrastructure, which responds to user requests in real-time, is directly affected by latency in terms of service quality. Services such as chatbots or search engines are still likely to prefer locations with network accessibility and proximity to metropolitan areas. In other words, not all AI data centers seek the same type of site.

Underwater and in Abandoned Mines... Experiments Beyond Imagination

Even bolder experiments are underway. Microsoft has launched 'Project Natick,' which involved placing a data center underwater. In 2018, a container-type underwater data center was installed 30 meters below the North Sea off Scotland, and two years later, it was recovered and its performance analyzed.

Underwater data centers can use cold seawater as a natural cooling source to reduce cooling costs and can be easily connected to offshore wind or tidal power generation.

Conceptual diagram of the Shanghai underwater data center (left) and the actual platform built at sea (right). Electricity is generated on the sea surface in conjunction with offshore wind power, and the data center module installed about 35 meters below the water surface uses cold seawater as a natural cooling source to cool the servers, drawing attention as a new form. Provided by Shanghai Underwater Data Center Project and Hiwin Technology

Although challenges such as corrosion, maintenance, and undersea cable construction remain, this experiment is considered to have broken the stereotype that data centers must be located on land.

China has gone even further. Recently, it began operating a commercial underwater data center near Shanghai in conjunction with an offshore wind farm. About 2,000 servers are installed 35 meters below sea level, using seawater as a cooling source. This new form of infrastructure generates electricity above the sea and cools heat with seawater below.

Dark, abandoned mines are also being transformed into data centers. The Lefdal Mine Datacenter in western Norway is a remodeled former mine. It is resistant to external shocks and climate changes and can utilize cold seawater from the nearby fjord as cooling water. The ability to secure abundant hydropower is another advantage.

Spaces that once extracted coal and minerals have now been reborn as digital infrastructure for data storage and AI computation.

Inside the Lefdal Mine Datacenter in Norway. This facility converted abandoned mining tunnels into a data center, utilizing the mountain rock and the cold seawater from the nearby fjord as cooling sources. It is considered a prime example of a former mineral extraction site reborn as data infrastructure for the AI era. Provided by Lefdal Mine Datacenter

Professor Park Jongbae of Konkuk University's School of Electrical and Electronic Engineering said, "AI data centers are becoming a kind of 'industrial nomad' in search of electricity. The reason why Arctic regions, coasts, and abandoned mines are attracting attention is that they simultaneously satisfy both the need for electricity supply and cooling cost reduction."

Professor Kim Sungjin of KAIST's Department of Mechanical Engineering also said, "Cooling accounts for around 30% of the total power consumption of a data center. If we can reduce the energy required for cooling, the options for data center site selection will become much broader." This means AI data centers are an industry that seeks out electricity before network connectivity.

"Not Only Consuming Electricity" ... Water-Thirsty AI

One lesser-known fact is that AI uses a considerable amount of water. Evaporative coolers and cooling towers operate to dissipate the heat generated by servers, and this process consumes a large amount of water.

According to the Brookings Institution in the United States, a typical data center uses about 300,000 gallons (about 1.13 million liters) of water per day on average, while hyperscale data centers can consume up to 5 million gallons (about 18.93 million liters) per day. This is enough to fill about 7.5 Olympic-size swimming pools.

The problem is that as the heat output of AI-dedicated graphics processing units (GPUs) rises sharply, it is becoming difficult to cope with conventional air-cooling methods alone.

Professor Kim explained, "Recently, AI semiconductors generate several hundred watts of heat per chip, and next-generation products are expected to exceed 1,000 watts. There are limits to cooling efficiency with air alone, making a transition to liquid cooling unavoidable."

An official from the SK Telecom AI Data Center business division said, "AI data centers must balance not only ultra-high-performance computing, but also extreme heat management and massive electricity supply. Since GPU server racks consume ultra-high-density power of tens of kilowatts or more, adopting next-generation cooling technology is essential."

Accordingly, the technology attracting industry attention is 'Direct Liquid Cooling (DLC).' This method directly supplies cooling water close to the internal chipsets of servers to remove heat, and it has recently emerged as a core technology in AI data center design.

Director No added, "With the increasing power density of the latest GPU servers, DLC is no longer optional but a must-have infrastructure technology. This is not simply about replacing equipment, but about transforming the entire data center design—including power supply structures, cooling water piping, server room layout, leak detection, and water quality management systems."

Professor Kim also stated, "Recently, 'dry cooling' technology, which recirculates cooling water or discharges heat into the external air, is also gaining attention. Improving cooling efficiency is important not only for reducing power usage but also for decreasing water consumption."

Big Tech Moves Next to Nuclear Power Plants

Given these circumstances, global big tech companies are naturally turning their attention to nuclear power plants.

Three Mile Island Nuclear Power Plant in Pennsylvania, USA, being reactivated to supply power for Microsoft's AI data center. Photo by Getty Images

AI data centers must operate 24 hours a day without interruption. While solar and wind power are environmentally friendly, their output depends on the weather. In contrast, nuclear power plants can provide large-scale electricity reliably, regardless of climate conditions.

This explains why Microsoft, Amazon, and Google are signing power purchase agreements with existing nuclear plants and investing in small modular reactor (SMR) companies. The International Atomic Energy Agency (IAEA) has also analyzed that the growing electricity demand from the AI industry and data centers is stimulating discussions about the next generation of nuclear power plants.

Professor Park predicted, "In the future, data centers will likely move away from simply purchasing electricity from energy companies and will instead build their own power generation facilities alongside data centers. We may see an era where data centers themselves become small-scale power companies."

Challenges for Korea's AI Strategy

This global infrastructure race also poses challenges for Korea.

Although demand for AI data center construction is increasing rapidly in Korea, the capacity of the metropolitan area’s power grid and the availability of suitable land are deteriorating. If data centers continue to concentrate in metropolitan areas, the burden on the power grid and transmission network will inevitably grow. As a result, the national AI strategy is shifting beyond merely securing GPUs to also addressing data center siting, electricity supply, and cooling water solutions.

Conceptual image of a space data center. Data centers are already heading to the Arctic Circle, underwater, and abandoned mines, while some companies are even considering establishing data storage and computing facilities in low Earth orbit. Photo by Conceptual Image of Space Data Center

Recently, the concept of 'in-orbit data centers,' which involves establishing data storage and computing facilities in low Earth orbit, has begun to be discussed in Europe and the United States. Although still in its early stages, this is a symbolic example showing that the competition for data center sites is expanding beyond the Arctic, the sea, and abandoned mines, and now even into space.

Even though the AI we encounter on our smartphones and PC screens may seem like nothing more than a few lines of text, behind the scenes there is a massive infrastructure race involving electricity, water, land, and power plants. In the AI era, competitiveness is no longer determined by algorithms alone. Who can secure power, water, cooling, and sites more reliably is emerging as the core variable in the next-generation AI race.

This content was produced with the assistance of AI translation services.