Data centers have transformed from inconspicuous technical infrastructure into one of the most strategic components of the global economy in just a few years. Artificial intelligence requires not only chips and servers, but also vast amounts of electricity, capital, land, water, and transmission infrastructure. The United States, China, the European Union, and Russia are entering this competition with very different advantages. The US dominates technologically and financially, China combines massive energy production with centralized planning, the EU has a strong low-carbon energy base, but relies on American cloud services, and Russia possesses exceptional energy resources that it has not yet been able to fully convert into computing power.
A data center is no longer just a building. It's an inverted power plant
When comparing different countries, the first step is to resolve the problem of statistics. Data centers are sometimes counted by the number of buildings, sometimes by the number of server racks, power consumption in megawatts, or even directly by electricity consumed. Some statistics include corporate data centers, others only commercial colocation centers, and some even include infrastructure for cryptocurrency mining.
Therefore, the most useful common metric remains electricity consumption.
The International Energy Agency estimated global data center electricity consumption in 2024 at approximately 415 TWh, which is about 1.5 percent of world electricity. The updated forecast estimates approximately 485 TWh in 2025 and around 950 TWh in 2030. Thus, consumption could almost double in just five years.
However, it's not just about the amount of energy. The power is concentrated in an extremely small area. A new AI campus requiring several hundred megawatts can resemble a large metallurgical or chemical plant from the perspective of the electricity grid.
And it is here that fundamental differences begin to emerge between different regions.
USA: The world number one, which is starting to run into its own power grid
The United States currently has the most advantageous combination of almost everything a modern AI data center needs.
It is home to Nvidia, AMD, Broadcom, and many other key semiconductor companies. American companies include Amazon Web Services, Microsoft Azure, Google Cloud, Meta, and OpenAI, and the US capital market can finance projects worth tens of billions of dollars.
According to the IEA, in 2024, the United States accounted for approximately 45 percent of global data center electricity consumption, or about 180 TWh.
A more recent study by the American Lawrence Berkeley National Laboratory from June 2026 shows how quickly the situation is changing. Its reference scenario predicts a consumption of around 649 TWh for US data centers in 2030, with a possible range of 521 to 843 TWh. Data centers could thus represent approximately 9.5 to 15.3 percent of all US electricity consumption, with a median estimate of 11.8 percent.
The size of the American market is illustrated by Northern Virginia. CBRE recorded approximately 4.18 GW of data center capacity in this single region alone in the first quarter of 2026. Available capacity, however, has fallen to a mere 0.3 percent.
The energy advantage of the US is its exceptionally diverse mix. According to the IEA, more than 40 percent of the electricity physically powering American data centers currently comes from natural gas, approximately 24 percent from renewable sources, around 20 percent from nuclear power, and about 15 percent from coal.
This allows for a quick combination of gas, solar, and wind energy, batteries, and existing nuclear resources. Technology companies are also supporting the restart of decommissioned nuclear reactors and the development of small modular reactors.
The weakness of the USA is no longer money or chips. Increasingly, it's the electrical grid.
Building a data center can take two or three years, while new power lines or large power plants take significantly longer. CBRE therefore considers the availability of electricity to be one of the main factors that determine where it is still possible to build in the USA. Projects are shifting from traditional centers towards Texas, Pennsylvania, West Virginia, and other regions with more affordable land and energy.
Therefore, America is currently leading, but is beginning to realize that even practically unlimited capital cannot immediately build an electrical infrastructure.
China: Shifting Computations Instead of Electricity Transportation
The Chinese model is fundamentally different.
Beijing views computing infrastructure in the same way as railways, energy, or telecommunications – as strategic state infrastructure.
According to the IEA, China accounted for approximately a quarter of global data center electricity consumption in 2024, which is roughly 100 TWh. By 2030, consumption is expected to increase by another approximately 175 TWh. The USA and China together are expected to account for nearly 80 percent of global growth in data center consumption by the end of the decade.
At the same time, Chinese official figures show the enormous physical scale of the infrastructure. The National Development and Reform Commission stated that as of mid-2025, there were approximately 10.85 million standard racks in operating data centers, with a total intelligent computing power of 788 EFLOPS. Three major telecommunications operators operated 938,000 racks for external customers at the end of 2025. These two figures cannot be directly compared because they use different definitions, but they clearly show the scale of the Chinese ecosystem.
However, the most interesting project is "East Data, West Computing."
A large part of China's digital industry is located on the resource-rich eastern coast, while vast wind, solar, hydro, and coal resources are located in the west of the country. Therefore, China is trying to move some computationally intensive tasks to Inner Mongolia, Ningxia, Gansu, Guizhou, and other energy-rich regions, and send the results back east via optical networks.
For training an AI model, it is not crucial whether the server is 20 or 2,000 kilometers from the user. For applications that require immediate response, distance can be a problem. The Chinese NDRC itself noted in 2026 persistent problems with latency, standardization of systems, and isolated "islands" of computing power.
China also has an environmental problem. Approximately 70 percent of the electricity physically consumed by its data centers currently comes from coal, because a large part of the current centers are located in eastern provinces. Renewable sources provide just under 20 percent, and the core provides approximately one-tenth.
The shift to the west is intended to change this ratio. According to government plans, new data centers in national computing hubs should achieve more than an 80 percent share of green electricity. For example, a 500 MW photovoltaic project connected directly to the Zhongwei computing base was launched in Ningxia in May 2026.
The biggest weakness of China remains advanced chips.
China is rapidly developing its own processors, including Huawei's Ascend, but its access to the most advanced American accelerators and equipment for manufacturing cutting-edge semiconductors remains limited. While the United States allowed limited shipments of some Nvidia H200 chips to Chinese customers in 2026, the most advanced systems remain restricted, and China is simultaneously supporting domestic alternatives.
China's advantage lies in its ability to build massive and coordinated infrastructure; its disadvantage is that a building with a gigawatt of electricity is not very useful if it cannot be filled with the best available accelerators.
European Union: Energy and infrastructure exist, but technological ownership is the problem
Europe is in a somewhat peculiar situation. There are many data centers here, but a large part of the digital infrastructure is not owned by European technology companies.
The European Commission estimated the EU's computing capacity in 2025 at approximately 12 GW, which accounted for about 20 percent of global data center capacity. The United States had approximately 42 percent.
CBRE expects that the broader European data center market – its methodology goes beyond just the EU itself – will reach approximately 13 GW by the end of 2026, representing a year-over-year growth rate of around 20 percent. Orders for colocation capacity specifically intended for AI more than quadrupled during the first half of 2026.
The European Commission, together with the IEA, estimates that data center consumption in the EU in 2024 is approximately 70 TWh and expects it to grow to 115 TWh by 2030.
In terms of energy, Europe has one major advantage: a large share of emission-free sources. The IEA expects that renewable sources and nuclear power will together provide approximately 85 percent of the electricity needed by European data centers by 2030.
Scandinavia offers cheaper hydropower and wind energy, along with a cool climate; France has extensive nuclear power; and Spain and Portugal have very good conditions for solar energy.
However, computing power is not distributed based on where electricity is cheapest. Traditional centers such as Frankfurt, London, Amsterdam, and Paris suffer from a lack of available grid connections, similar to Northern Virginia.
An even bigger problem is cloud ownership.
According to documents from the European Commission, Amazon AWS, Microsoft Azure, and Google Cloud control approximately 70 percent of the European cloud market. The share of European providers has fallen from 29 percent in 2017 to approximately 15 percent, and even the largest individual European players have only around two percent of the market.
Therefore, Europe can physically host a data center in Germany, Ireland, or France, but its technological ecosystem, GPUs, and cloud platform may be American.
Brussels is now trying to address this weakness. The EU is preparing at least 19 AI Factories, several significantly larger AI Gigafactories, and through InvestAI, it aims to mobilize approximately 20 billion euros for the gigafactories. At the same time, the European Commission has set a goal of tripling data center capacity in the EU within five to seven years.
The disadvantage for Europe is not a lack of technical expertise or electricity. It is market fragmentation, slower permitting processes, limited financial strength of European technology companies, and dependence on American GPUs and cloud services.
Russia: Enormous energy potential, but electricity alone will not create an AI superpower
Russia is by far the smallest of these four markets, and its statistics are the most complex.
According to iKS-Consulting, the Russian commercial market reached approximately 85,800 racks by the end of 2025. Performance estimates vary depending on whether they include commercial, enterprise, or mining data centers. More conservative estimates put the total power consumption of traditional Russian data centers at around 1.2 to 1.7 GW, with approximately 0.8 GW attributed to the commercial segment.
For comparison, RTK-COD, a leading Russian company within the Rostelecom group, operates approximately 27,800 racks with a total power consumption of 235 MW.
Russia possesses exceptional conditions for energy-intensive data centers: vast reserves of natural gas and coal, a large nuclear power industry, massive hydroelectric plants in Siberia, ample land availability, and, in many regions, a cold climate.
One example is Udomlja, located near the Kalinin Nuclear Power Plant, where a data center complex with a power consumption of around 48 MW has been built. IXcellerate is also planning campuses in Moscow and its surrounding areas with hundreds of megawatts of additional capacity.
Energy resources are not the same as readily available electrical power in the right location.
Approximately three-quarters to four-fifths of Russia's commercial data center infrastructure is concentrated in Moscow and its surrounding area. According to Vedomosti, network companies in that region began rejecting some new connections in 2026 due to limited available capacity, which was either fully utilized or reserved for the period of 2026–2028. The occupancy rate of data centers in Moscow is around 95 percent.
Russia faces the same paradox as the United States or Europe: a country may have vast amounts of energy, but not necessarily where data centers need it most.
A logical solution is to move new computing capacity closer to nuclear and hydroelectric power plants in northwestern Russia, the Urals, or Siberia, and connect them with a high-capacity optical network to Moscow and St. Petersburg.
However, the biggest problem for Russia is not electricity.
It's chips, servers, and capital.
Nvidia confirms in its regulatory documents that it ceased direct sales and operations in Russia after the imposition of sanctions. At the same time, U.S. export regulations restrict the supply of high-end accelerators, such as the A100 and H100 families, and related systems.
Therefore, while Russian data centers can be powered cheaply by nuclear or hydroelectric power, obtaining tens of thousands of state-of-the-art AI GPUs is significantly more difficult than in the United States, Europe, or China.
Furthermore, high capital costs and more expensive imported infrastructure are hindering construction. In 2025, the number of new commercial racks increased by only about 5.8 percent to 85,800, which represents a significant slowdown compared to previous years.
Russia has very good conditions for the energy component of AI infrastructure, but it has the weakest access to state-of-the-art semiconductors and the smallest capital market among the four regions.
So, who is in the best position?
If the metric is state-of-the-art computing power, then the United States clearly leads. It has chip manufacturers, the largest hyperscalers, the deepest capital markets, and the highest concentration of AI companies.
If the metric is the ability to quickly build a massive energy and data infrastructure, then China is the most serious competitor to the United States. Its advantage lies in its ability to coordinate the construction of power plants, transmission networks, optical routes, and data centers as a single entity.
The European Union stands among them. It has a sufficiently large economy, a high-quality electrical grid, strong nuclear and renewable energy sectors, cutting-edge research, and a crucial position in the semiconductor manufacturing chain. However, its weakness remains the lack of its own hyperscaler comparable to Amazon, Microsoft, or Google, and a strong dependence on foreign AI accelerators.
Russia may have better conditions than its current share of the global market suggests, in terms of the availability of energy resources, space, and climate. However, it has not yet been able to fully translate this energy advantage into a technological one.
The development of data centers shows one important change in the global economy: the future of artificial intelligence will not be determined solely by who creates the best algorithm. It will also be crucial who can manufacture chips, secure capital, and provide several gigawatts of stable electricity where the servers will be located.
In this new race, data centers are becoming less like office buildings and more like a new type of heavy industry in the digital age.