Cyprus has significant disadvantages when it comes to attracting and utilizing large artificial intelligence data centers, said Marios Dikaakos, a professor of computer science at the University of Cyprus, in an interview with CNA.

These disadvantages, he said, include, among others, the very high cost of electricity, high temperatures, and a lack of water needed to cool the infrastructure.

In his interview with CNA, Mr. Dikaakos explains what AI data centers are, where most of them are located, and analyzes their environmental impact.

Today, he noted, approximately 75% of the world’s computing power for Artificial Intelligence is located in the U.S., 12–15% in China, and just 5% in the European Union.

He emphasized that the overconcentration of AI systems in the U.S. and China creates significant technological, economic, and geopolitical imbalances, as access to computing power is becoming a critical factor in competitiveness and national sovereignty.

Noting that the expansion of artificial intelligence poses enormous challenges for the environment and the management of natural resources, Mr. Dikaakos said that there are data centers under construction that will consume, on average, more electricity than the entire Republic of Cyprus.

He also focused on the water footprint of cloud computing, since water is essential for cooling cloud computing infrastructure.

By 2027, he emphasized, global demand for AI is estimated to require 4.2–6.6 billion cubic meters of water annually, an amount greater than the total annual water withdrawal of 4–6 countries the size of Denmark.

What Are Artificial Intelligence Data Centers (AIDCs)?

Mr. Dikaiakos said that AI Data Centers are industrial facilities designed to house and ensure the operation of clusters of hundreds or thousands of computing nodes.

These facilities, he added, in addition to computers and networking and data storage infrastructure, include complex electromechanical systems for power transformation and internal distribution, cooling systems to dissipate the heat generated, monitoring, control, and physical security systems, etc., as well as backup units to ensure uninterrupted operation.

Mr. Dikaakos told KYPE that Artificial Intelligence Data Centers (AIDCs) primarily house specialized computing equipment for the “training” and use of generative artificial intelligence models, such as GPT.

“Their specialized equipment consists of computing clusters with graphics processing units (GPUs—Graphical Processing Units), which perform extremely fast mathematical calculations in linear algebra on large volumes of data. Compared to conventional processors (CPUs), “GPUs can run artificial intelligence applications 50–300 times faster, though they require 2–4 times more power,” he emphasized.

Consequently, the professor continued, over the past two years, the rapid development of Generative AI has led to “an unprecedented expansion of HPC infrastructure, which is necessary for training large language models on massive volumes of digital data and for providing AI services (such as ChatGPT and Claude) to millions of users.”

The Colossus LLAN will consume more electricity than the entire country of Cyprus

He cited as an example the Colossus AI data center currently under development by xAI, which will house approximately 500,000 GPUs and require a total electrical power of around 1 GW to operate, with an estimated investment cost of $40–60 billion.

“Such a supercomputing center would consume, on average, more electricity than the entire Republic of Cyprus,” the professor emphasized in his interview with CNA.

He said that according to recent reports, there are currently supercomputers in operation worldwide with approximately 20 million specialized computing nodes equipped with GPUs, with that number expected to double roughly every nine months, reaching 200 million by the end of 2028.

The U.S. has the most AI data centers—ten times more than China

Regarding the locations of most data centers, Mr. Dikaakos said that the United States dominates the global AI infrastructure, hosting, according to recent studies, approximately 5,500 data centers—about ten times more than China.

“It is estimated that today, approximately 75% of the world’s computing power for artificial intelligence is located in the U.S., 12–15% in China, and just 5% in the European Union,” the professor noted.

He also said that Gulf countries (Saudi Arabia, the United Arab Emirates, and Qatar) had announced ambitious investment plans to establish large data centers on their territories, a development that, due to the war in the Persian Gulf, appears to have been affected, as Mr. Dikaakos pointed out.

He also spoke about geopolitical imbalances caused by overconcentration in the U.S. and China.

“The overconcentration of AI systems in the U.S. and China creates severe technological, economic, and geopolitical imbalances, as access to computing power is becoming a critical factor in competitiveness and national sovereignty,” he emphasized.

Cyprus Faces Challenges in Attracting AI Metals

When asked about the situation in Cyprus, the professor noted that the HPC sector has not developed significantly, since, as he explained, “the country faces significant disadvantages when it comes to attracting and utilizing large-scale renewable energy projects.”

These disadvantages, he said, are “the very high cost of electricity and the well-known problems with electricity generation and distribution, high temperatures and a lack of water needed to cool the infrastructure, the relatively high cost of land, the relative lag in the production and management of digital data and services, the absence of strong technical expertise and a critical mass of scientists and engineers specializing in the design, development, and operation of data centers, the great distances from major urban centers abroad, from which the demand for cloud computing services primarily originates, and, more generally, the comparatively low levels of investment in research and technological development related to these areas.”

Furthermore, Mr. Dikaakos added, “it is clear that the existing national infrastructure at universities and research centers is insufficient for training young scientists in emerging high-performance computing technologies with high computational requirements.”

This problem, he continued, “is certainly not unique to Cyprus, but is an international one, and is exacerbated by the skyrocketing cost of computers due to the frenzy surrounding the development of high-performance computing in the U.S. and China.”

“Cyprus’s small size is a limiting factor that is difficult to overcome,” he said.

However, Mr. Dikaakos noted that according to Stanford University’s Global AI Vibrancy Index for 2025, which also includes computing infrastructure for AI, “we see that small countries such as Luxembourg, Singapore, Estonia, the United Arab Emirates, and Ireland have managed to rank very high on the Global AI Vibrancy index, even taking their small populations into account.”

The Largest Investment Wave in History: Infrastructure for Start-ups

When asked how the AI and machine learning sector is expected to evolve in the coming years, Mr. Dikaakos noted that, according to The Economist and other authoritative sources, it is estimated that “the development of infrastructure for Artificial Intelligence will constitute the largest investment wave in history.”

In 2026 alone, he added, approximately $900 billion is expected to be spent on processors, GPUs, data centers, and energy infrastructure, of which more than 400 billion will come from borrowing.

“To financially justify the current wave of investment in AI infrastructure, it is estimated that the sector will need to generate annual revenue from AI applications in the range of $2.5 trillion, “an amount greater than the total revenue currently generated by the entire global technology sector from all its activities,” the professor emphasized.

Furthermore, he said, according to a recent article on the Nikkei Asia website, the five largest U.S. technology giants, which manufacture and operate data centers, have accumulated $1.65 trillion in “hidden” debt, which is not reflected on their balance sheets.

“When we see the exponential growth of a new technology with investments of this magnitude, it is very difficult to make reliable predictions,” Mr. Dikaakos pointed out.

First of all, he added, “it cannot be ruled out that new scientific research findings could completely overturn the AI development model of the past 2–3 years, which requires ever-larger infrastructure and more data.”

It is also possible, he said, “that at some point there will be a correction in the current investment frenzy, with unforeseeable consequences for the sector and the global economy.”

Finally, he noted that “it is safe to assume that the successful integration of AI into traditional production processes will not be so rapid and will not yield the economic results that would generate returns sufficient to cover theexisting investments, loans, obligations, and future expenses for maintaining and updating the KDTN infrastructure.”

Enormous Challenges for the Environment Surrounding Data Centers

Mr. Dikaakos told KYPE that if we focus on the environmental impacts of the development of cloud data centers, we must take into account that the energy consumption of cloud computing companies is growing at an exponential rate.

According to recent reports, he noted, Google’s electricity consumption increased by 7 TWh from 2023 to 2024 and by 12 TWh from 2024 to 2025, which indicates that not only is consumption increasing, but so is the rate of increase itself.

Equally important, he emphasized, is the cloud’s water footprint, since water is essential for cooling cloud computing infrastructure.

The professor told CNA that, according to researchers at the University of California, training the GPT-3 model in Microsoft data centers required the evaporation of approximately 700,000 liters of clean freshwater, while by 2027, global demand for AI is estimated to require 4.2–6.6 billion cubic meters of water annually, an amount greater than the total annual water withdrawal of 4–6 countries the size of Denmark.

In another study, he said, the Brookings Institute notes that a typical data center consumes about 300,000 gallons of water per day (approximately 1.1 million liters), while a large hyperscaler can reach 5 million gallons per day (approximately 19 million liters), an amount equivalent to the needs of a city of 50,000 residents, and that water consumption for cooling is likely to increase by up to 870% in the coming years.

He noted, however, that there are differing estimates from representatives of renewable energy companies, who argue that with the implementation of water recycling, the water footprint of AI systems will be virtually negligible.

“In any case, it is clear that the expansion of Artificial Intelligence poses enormous challenges for the environment and the management of natural resources,” emphasized Mr. Dikaakos.

The question is how to mitigate the negative consequences

Noting that the development of AI is already bringing about major changes in all sectors, Mr. Dikaakos stated that the utilization of AI will inevitably affect jobs and activities across a wide range of sectors.

He also noted a “paradigm shift” in the conduct of scientific research, “which is expected to yield impressive results in many fields, particularly in countries that have trained scientific personnel and the capacity to collect and utilize digital data.”

The professor emphasized that “alongside the use of generative AI and tools such as ChatGPT, critical threats to human autonomy, the intellectual development of young people, culture, and democracy are emerging”

Consequently, Mr. Dikaakos concluded, “the question is not only how we will harness Artificial Intelligence, but also how we will mitigate its negative consequences, by refocusing our attention on the fundamental values of education and science—namely, the development of human autonomy, critical thinking, and the intrinsic value of intellectual inquiry.”.

Source: KYPE