The advent of ChatGPT and LLMs has plunged the world into a new phase of hegemonic competition. At the core of this competition is not military power or fossil fuel transport routes, but “AI processing capability” and “power supply capacity.”
Especially unless “AI processing ” requires enormous electric power, “power supply capacity” could be the most critical core component of AI Data Center.
As AI models grow larger, the high-performance GPU clusters required for their training and inference are beginning to demand electricity consumption on a national scale.
According to IEA statistics, the largest consumers of power within data centers are servers, including Central Processing Units (CPUs) and GPUs, accounting for approximately 60-75% of the total.
Data Center Power Consumption Breakdown:
- Servers: ~60%
- Storage: ~5%
- Network: ~5%
- Cooling Systems: ~7-30%
- Other
In 2024, global data center power consumption reached 415 TWh (terawatt-hours), which is about 1.5% of the world’s total electricity consumption and equivalent to roughly five months of Japan’s national power consumption. Moreover, this consumption has been increasing at an annual rate of about 12% over the past five years.
This article analyzes from various angles how the competition for power supply capacity for data centers is intensifying and becoming more severe.
Contents
- 1 AI Competition Becomes a Electric Power Competition
- 2 The Changing Role of Data Centers (From IT Facility to Strategic Infrastructure)
- 3 New Geopolitical Factors Brought by Energy Resources and Power Grids
- 4 AI-Companies and Countries Competition for Power
- 5 Power Outage Risks Threatening AI Systems
- 6 Five Key Points of the Hegemonic AI Struggle
- 7 Is control of power grids the real key to AI dominance?
- 8 Summary: Power Grids Decide the AI Era
AI Competition Becomes a Electric Power Competition
Data centers (DCs) were once discussed within the framework of corporate IT infrastructure. Today, however, they have evolved to the point where they can be described as “national strategic apparatuses that convert massive amounts of electricity into data (AI).”
While data itself crosses borders via networks, power supply is currently determined by physical infrastructure. Therefore, to seize leadership in the AI era, simply having servers or even deploying GPUs is not enough. Dominance in “securing power,” “improving power efficiency,” and “completing power transmission (grid) infrastructure” is the key to AI hegemony.
※This theory is also strongly advocated by NVIDIA CEO Jensen Huang. According to him, when looking at power supply capacity, the United States is already significantly lagging behind China.
Nvidia GTC：CEO Jensen Huang delivers keynote address（YouTube）
https://www.youtube.com/watch?v=t11IVAZhbg0
The competition for AI data centers is transforming into a scramble among nations and corporations over power and location, where geopolitical impacts will greatly influence each country’s AI hegemony.
※As a side note, viewing things from this perspective makes it clear that AI development in Japan, Germany, Korea, a country poor in energy resources, faces inherent limitations.
The Changing Role of Data Centers (From IT Facility to Strategic Infrastructure)
The rise of generative AI, LLMs, and image/video generation models has fundamentally changed the concept of power design in data centers.
While traditional data centers were built around CPU-based server clusters, AI models center on parallel computing and processing, leading to an incomparably dramatic leap in power density. Furthermore, because cooling facilities also require massive amounts of electricity, a chain structure of “power → cooling → processing computation” is formed in AI models.
Consequently, data centers have transformed into infrastructure that can no longer be discussed solely in terms of network or land costs. The criteria now placed at the top for evaluating data center locations include: “low electricity costs,” “stable power supply,” “cooling efficiency from the local environment,” “effectiveness of the transmission grid,” and “abundance of energy resources and low geopolitical risk.”
These power supply capabilities cannot be considered separately from national politics. In other words, the construction and operation of data centers is shifting from an extension of corporate competition to the building of spheres of influence among nations.
New Geopolitical Factors Brought by Energy Resources and Power Grids
As we have seen, implementing AI models requires an extremely vast amount of electricity.
Therefore, in this field, countries with “abundant energy resources,” “stable power supply capabilities,” and “influence over neighboring countries through their power grids” hold an overwhelming advantage and can become game-changers in the AI era. As mentioned earlier, while GPUs can be transported easily, power and transmission grids cannot.
Let’s look at some specific potentials.
1. Potential Rise of Energy-Rich Countries in the AI Era
Countries possessing fossil fuels such as oil, natural gas, and coal have an advantage due to their high self-sufficiency in electricity. Examples include Middle Eastern nations, the United States, and Russia. On the other hand, countries rich in hydropower, geothermal, and renewable energy like China, Brazil, and Canada also hold long-term advantages due to lower risks from fuel price fluctuations and trade sanctions.
Furthermore, regions like Nordic Norway, Central Asia, Kyrgyzstan, and Tajikistan, with their combined advantages of hydropower, geothermal resources, and cool climates, are gaining attention as areas with high potential in both electricity unit costs and cooling efficiency.
2. Electricity Exports and Power Control as New Diplomatic Tools
When AI companies plan data center construction, electricity itself can become a diplomatic tool. Countries can impose electricity export restrictions, offer preferential electricity prices to foreign companies, or grant/deny access to the power grid.
For example, if a structure emerges where “to operate AI in this region, you must use our country’s power grid,” even countries without AI capabilities can gain influence over AI companies and data flows by controlling the power supply.
3. Control Over Power Grids Dictates the Realization of AI Data Centers
Power grid systems directly connected to data centers exert far greater influence than pipelines or submarine cables.
Within these networks, those supplying electricity to servers directly are the most critical. After all, without power, neither data nor AI can function. In other words, controlling the power grid means controlling the foundation on which AI data centers are built.
Control over power grids dictates the realization of AI data centers.
AI-Companies and Countries Competition for Power
Next, let’s discuss the blurring line between corporations and countries (governments).
Recently, AI-related companies like GAFA (Google, Amazon, Meta, Apple), Microsoft, NVIDIA, and OpenAI are directly entering national-level power contracts, accelerating investments in power plant construction, acquisitions of renewable energy suppliers, and long-term land lease agreements.
This represents a move where corporations are securing power and locations ahead of government. While the traditional order was “country → company” for power supply, a reversal phenomenon of “company → occupation of the national power grid” is now occurring. The reason companies are investing in power infrastructure is not merely to reduce operational costs but because “power is the bottleneck that determines the outcome of the AI competition.”
Particularly, GPU clusters for AI companies have power densities per rack unit reaching several dozen times bigger than traditional levels. Even there are ones that consume as much electricity as an entire city.
These corporate actions directly intersect with national power policies (nuclear power resurgence, renewable energy investment, fuel export controls, international grid connections).
Power, AI, and the land of data center are becoming a trinity, forming a new business model where state intervention is involved.
Power Outage Risks Threatening AI Systems
In the new AI era, the superiority or inferiority of data center projects cannot be judged solely by code, algorithms, or chip quality.
At the core of a data center lies the “land” where cooling water flows and the physical, analog elements of “power plants” and “transmission grids” that generate massive electricity are paramount. Here, we discuss the greatest vulnerabilities faced by AI systems: power dependency and geographic concentration risks.
1. Power Outages Are More Critical Than Network Failures
Networks can be restored only if power is available, but restoring electricity is often uncertain and vulnerable. Consequently, areas equipped with emergency power solutions—such as large-scale UPS, private generators, or storage batteries—become highly attractive for investment.
2. Electricity Prices Reflect AI Competitiveness
Another vulnerability of AI systems lies in electricity cost.
AI development in countries with high electricity prices faces proportionally higher training expenses. Countries that cannot absorb these costs risk falling behind in global AI competition, because AI relies on massive amounts of electricity—making electricity prices a direct measure of AI combat power.
3. Geographic Concentration Risk of Data Centers
Currently, most data centers are concentrated in North America, Europe, the Middle East, and East Asia, making them highly exposed to cascading risks such as electricity export restrictions, climate hazards, geopolitical tensions, energy price volatility, and power grid disruptions.
Conversely, countries or regions with high energy self-sufficiency, efficient cooling, political neutrality, and a relatively unconcentrated data center landscape stand to gain the most in the upcoming AI race. In other words, unexpected emerging regions could rapidly rise as new data center hubs.
Five Key Points of the Hegemonic AI Struggle
Finally, here are key points for identifying potential winning projects in the AI data center era.
- Self-sufficiency/Supply Capacity of Power Resources (can the country absorb power generation costs domestically?)
- Stability of Power Supply (transmission grid, emergency measures, UPS, disaster resilience)
- Cooling Efficiency (climate conditions, power dependency of cooling technology)
- Influence of the Power Grid (electricity export controls, international connections, regulations, rights)
- Data Center Location Environment (control or management rights over locations for AI companies)
Some argue that capabilities and rights related to these power aspects are more appropriate data than corporate financial results or national trade balances for accurately reading hegemonic structures.
Is control of power grids the real key to AI dominance?
Countries and companies that secure strong power grid capacity and strategic locations may gain an edge in AI projects—even without owning advanced AI models themselves.
This shift in power dynamics has already begun. For example, Norway is positioning itself as a “Northern Data Hub,” leveraging abundant hydropower and a cold climate ideal for cooling efficiency. Meta and Google have built large-scale data centers there, signaling that the country’s role goes far beyond corporate investment—it’s a national strategy demonstrating geopolitical influence over AI infrastructure.
In Tennessee, USA, Google’s data center project faced delays after the local power company requested construction postponement due to insufficient grid capacity. As a result, Google has been forced into additional investment to meet energy demands.
Meanwhile, hydropower dams in the Amazon rainforest can fuel AI models in Silicon Valley, and solar power in the Middle East has the capacity to run autonomous driving AI systems in Europe. This emerging landscape creates new dependencies—and geopolitical tensions—between power suppliers and AI developers.
Summary: Power Grids Decide the AI Era
As it has been observed, the competitive advantage of AI data centers cannot be measured solely by AI model performance or GPU scale. The true foundation lies in power supply capacity and control over power transmission networks.
It is no exaggeration to say: “Those who control power control AI, and those who control AI transform countries.”
That is precisely why the next generation of “superpowers” is more likely to emerge not from AI model or GPU designers, but from those who dominate continental-scale power grids and strategic data center territory.
In the AI era, the hegemonic struggle will unfold not across algorithms, but across maps of energy and electricity—the most intense arena of competition ahead.
