Powering the digital economy: The global expansion of data centres and its energy implications – CEPR

Introduction: The Physical Reality of a Digital World

Every time you stream a high-definition movie, join a video conference, ask a generative AI to write a poem, or simply scroll through your social media feed, you are interacting with one of the most significant and unseen industrial revolutions of our time. The seamless, ethereal experience of the “cloud” belies a profoundly physical reality: a global network of massive, windowless buildings packed with humming servers, intricate cooling systems, and an insatiable appetite for electricity. These are the data centres, the factories of the 21st century, and their explosive growth is the engine powering our digital economy.

This unprecedented expansion, however, presents a formidable challenge that is increasingly drawing the attention of economists, policymakers, and energy experts, as highlighted in recent analyses from institutions like the Centre for Economic Policy Research (CEPR). The very infrastructure enabling our digital future is placing an immense and accelerating strain on global energy grids and climate goals. As we move deeper into an era defined by artificial intelligence, the Internet of Things (IoT), and ubiquitous connectivity, we are confronting a critical paradox: the drive for digital progress is colliding with the physical limits of our energy and environmental systems. Understanding the scale of this expansion, its profound energy implications, and the innovative solutions required to mitigate its impact is no longer a niche technical concern—it is one of the defining economic and environmental issues of our age.

The Unseen Engine: What is Driving the Data Centre Boom?

The global demand for data centres is not driven by a single factor but by a powerful confluence of technological shifts that are fundamentally reshaping how we live, work, and interact. This boom is a direct consequence of the exponential growth in data generation, processing, and storage, fueled by several key trends.

The Great Migration to the Cloud

A decade ago, most large organizations managed their own on-premise server rooms—costly, inefficient, and difficult to scale. The rise of hyperscale cloud computing, pioneered by giants like Amazon Web Services (AWS), Microsoft Azure, and Google Cloud, changed everything. These companies built vast, centralized data centres, offering computing power and storage as a utility. This “cloud migration” allowed businesses of all sizes to offload their IT infrastructure, gaining flexibility, scalability, and cost savings. The result has been a massive consolidation and expansion of computing resources into ever-larger and more numerous hyperscale data centres, each of which can consume as much electricity as a small city.

The Data Deluge: From IoT to Streaming

The volume of data being created globally is staggering and continues to grow at an exponential rate. This “data deluge” stems from countless sources:

• The Internet of Things (IoT): Billions of connected devices—from smart home assistants and wearable fitness trackers to industrial sensors in factories and autonomous vehicle components—are constantly generating and transmitting data that needs to be collected, processed, and stored.
• Digital Media and Entertainment: The shift from physical media to on-demand streaming for video (Netflix, YouTube), music (Spotify), and gaming (GeForce Now, Xbox Cloud Gaming) requires enormous data centre capacity to store content and deliver it to millions of users simultaneously with minimal latency.
• E-commerce and Digital Services: Every online transaction, social media interaction, and digital service relies on a complex web of data processing that occurs in a data centre somewhere in the world.

This relentless creation of data necessitates a corresponding expansion in the physical infrastructure to house it, creating a perpetual cycle of demand for new and larger data centre facilities.

The AI Revolution: A New, Power-Hungry Paradigm

If the cloud and big data lit the fire of the data centre boom, the artificial intelligence revolution has poured jet fuel on it. The computational requirements for training and running advanced AI models, particularly large language models (LLMs) like those behind ChatGPT and other generative AI tools, are orders of magnitude greater than for traditional computing tasks. Training a single state-of-the-art AI model can consume gigawatt-hours of electricity, equivalent to the annual energy consumption of hundreds of homes.

This is because AI processing relies on high-density computing racks filled with power-hungry GPUs (Graphics Processing Units) that generate immense amounts of heat. A standard server rack might draw 5-10 kilowatts (kW) of power; an AI rack can easily draw 50-100 kW or more. This not only increases direct electricity consumption but also places extreme demands on cooling systems, which in turn consume even more energy. As AI is integrated into more applications, from scientific research and drug discovery to business analytics and consumer products, the demand for specialized, high-density, AI-ready data centres is skyrocketing, creating a new and even more energy-intensive phase of the expansion.

The Sobering Reality: Quantifying the Energy and Environmental Footprint

The digital world’s physical backbone has a voracious and rapidly growing appetite. While efficiency has improved, the sheer scale of the data centre expansion is creating a formidable environmental and infrastructural challenge that can no longer be ignored.

A Growing Share of Global Power

According to the International Energy Agency (IEA), data centres, transmission networks, and cryptocurrencies accounted for approximately 2% of global electricity consumption in 2022. While this figure may seem modest, its rate of growth is alarming. Projections vary, but many experts predict this share could double or even triple by 2030, driven largely by the proliferation of AI workloads. In some regions with a high concentration of data centres, the impact is already acute. In Ireland, for instance, data centres now consume nearly 20% of the country’s total grid electricity, a figure projected to rise to nearly 30% by the end of the decade, straining the national grid and raising concerns about energy security.

This demand creates a direct challenge for utility providers, who must invest billions in new power generation and transmission infrastructure to keep pace. The long lead times for such projects often lag behind the rapid construction cycles of data centres, creating a potential bottleneck for digital growth.

Beyond Kilowatts: The Water-Energy Nexus

A data centre’s environmental impact extends beyond its electricity meter. Cooling the thousands of servers operating 24/7 generates a tremendous amount of heat, which must be constantly dissipated. Many traditional data centres rely on evaporative cooling systems, which use vast quantities of water. A single large data centre can consume millions of gallons of water per day, comparable to the consumption of a small town. This creates a significant strain on local water resources, particularly in the arid and semi-arid regions where many data centres are located, such as Arizona and Utah. This “water-energy nexus” means that the digital economy’s growth is intertwined with two of our most precious and increasingly scarce resources.

The Limits of Efficiency: Understanding PUE

The industry has made significant strides in improving energy efficiency, a metric often measured by Power Usage Effectiveness (PUE). PUE is the ratio of the total energy used by a data centre to the energy delivered to the computing equipment. A perfect PUE of 1.0 would mean no energy is “wasted” on overheads like cooling and power distribution. While older facilities had PUEs of 2.0 or higher, modern hyperscale data centres operated by major tech companies often boast impressive PUEs of 1.2 or even lower.

However, these efficiency gains are being outpaced by the sheer velocity of growth. A 10% improvement in efficiency is rendered moot if the total computational demand doubles in the same period. Furthermore, the rise of high-density AI racks is challenging traditional cooling methods, potentially threatening to reverse some of the hard-won gains in PUE as more energy-intensive cooling solutions become necessary.

The Global Hotspots: Mapping the Digital Infrastructure Race

The global distribution of data centres is not uniform. It is concentrated in specific regions, or “hotspots,” that offer a confluence of favorable conditions: robust fiber optic connectivity, reliable and affordable power, a skilled workforce, and supportive government policies. However, the very success of these hubs is now creating significant challenges.

“Data Center Alley” and the Megaclusters

The most famous of these hotspots is Loudoun County in Northern Virginia, often called “Data Center Alley.” It is the largest concentration of data centres in the world, handling an estimated 70% of global internet traffic on any given day. Other major global hubs include:

• Silicon Valley, California: The historic heart of the tech industry remains a key hub.
• Dublin, Ireland: A primary gateway to Europe, attracting hyperscalers with favorable tax policies and a cool climate.
• Singapore: The premier data hub for Southeast Asia, prized for its political stability and connectivity.
• Amsterdam, Netherlands: A central connectivity point for Europe, home to one of the world’s largest internet exchanges (AMS-IX).
• Frankfurt, Germany: A major financial and connectivity hub in continental Europe.

These megaclusters have created significant economic benefits, generating jobs and tax revenue. However, they have also become victims of their own success.

Regulatory Headwinds and Community Pushback

The intense concentration of data centres in these hotspots has led to mounting pressure on local infrastructure, particularly energy grids. In 2022, Dominion Energy, the utility provider for Northern Virginia, announced that it could not meet all new data centre connection requests due to constraints in its transmission capacity. Similarly, Dublin has faced grid alerts, and EirGrid, Ireland’s grid operator, has placed restrictions on new data centre connections in the Dublin area.

This has led to a wave of regulatory pushback. Cities like Amsterdam and Singapore have implemented moratoriums on new data centre construction to allow infrastructure and sustainability policies to catch up. Local communities are also raising concerns about the noise from cooling systems, the consumption of land and water, and the visual impact of the massive, non-descript buildings. This is forcing the industry to look for new frontiers.

The Race for New Frontiers

In response to saturation and regulation in established hubs, data centre developers are actively seeking new locations. This expansion is moving in several directions:

• Secondary Markets: Cities like Madrid, Warsaw, and Milan in Europe, and Phoenix, Atlanta, and Chicago in the U.S., are emerging as the next tier of data centre hubs.
• Cold Climates: The Nordic countries (Sweden, Finland, Norway) are attracting investment due to their cold climates, which allow for “free air” cooling for much of the year, and an abundance of renewable energy, primarily hydropower.
• Emerging Economies: There is a growing push to build data centres in Latin America, Africa, and other parts of Asia to serve growing local populations and reduce latency for digital services in these regions.

This geographic diversification is essential for the continued growth of the digital economy, but it also exports the same energy and infrastructure challenges to new communities around the globe.

The Path Forward: Balancing Digital Ambition with Sustainable Reality

The collision course between digital growth and energy constraints is not an unsolvable problem. A combination of technological innovation, strategic investment, and smart policy is charting a path toward a more sustainable digital infrastructure. The industry’s leading players, aware of the reputational and logistical risks, are at the forefront of this shift.

The Push for 24/7 Carbon-Free Energy

The single most impactful strategy for decarbonizing data centres is to power them with renewable energy. Major technology companies like Google, Microsoft, and Amazon are among the largest corporate buyers of renewable energy in the world. They actively sign long-term Power Purchase Agreements (PPAs) with wind and solar farm developers to match their energy consumption.

The next frontier is achieving 24/7 carbon-free energy, which means ensuring that a data centre’s operations are matched with clean energy generation on a local grid, every hour of every day. This is far more challenging than simply buying renewable energy credits, as it requires a mix of renewables (solar for the day, wind for the night), energy storage solutions like batteries, and advanced grid management to handle the intermittency of these sources. Achieving this goal at scale is a primary focus for the industry’s sustainability leaders.

Innovation in Efficiency and Cooling

As the power density of server racks increases, particularly with AI, traditional air cooling is becoming insufficient and inefficient. This is driving rapid innovation in more advanced cooling techniques:

• Liquid Cooling: This involves bringing a liquid coolant directly to the hottest components (direct-to-chip cooling) or submerging entire servers in a non-conductive dielectric fluid (immersion cooling). Liquid is far more effective at transferring heat than air, allowing for denser server configurations and dramatically reducing the energy required for cooling.
• Waste Heat Reuse: The vast amount of low-grade heat generated by data centres is typically vented into the atmosphere as waste. Innovative projects are now capturing this heat and using it for district heating systems, warming nearby residential buildings, offices, or greenhouses. A data centre in Stockholm, for example, provides heat for thousands of local apartments.
• AI for AIOps: Artificial intelligence is being used to optimize data centre operations themselves. AI-powered management systems can analyze real-time data on server loads, electricity prices, and ambient temperatures to dynamically adjust cooling systems and shift non-critical workloads to times of lower energy cost or higher renewable energy availability.

The Role of Policy and Strategic Planning

Government and regulatory bodies have a crucial role to play in guiding the sustainable development of data centre infrastructure. Effective policy can include:

• Grid Modernization: Investing in upgrading national and regional electricity grids to be more resilient, flexible, and capable of integrating both large, concentrated loads like data centres and variable renewable energy sources.
• Streamlined Permitting for Green Projects: Fast-tracking approvals for data centres that meet high sustainability standards, such as commitments to use 100% renewable energy, employ advanced cooling technologies, or engage in waste heat reuse projects.
• Zoning and Land Use Planning: Proactively identifying and zoning specific areas for data centre development where power and fiber infrastructure can be efficiently co-located, preventing haphazard development that strains residential areas and unprepared grids.

Conclusion: Building the Future on a Foundation of Concrete, Steel, and Silicon

The digital economy, for all its virtual convenience, rests on a colossal and growing foundation of physical infrastructure. The cloud is not an amorphous entity in the sky; it has a physical address, a power cord, and a significant environmental footprint. As the analysis from institutions like CEPR underscores, the global expansion of data centres represents a monumental achievement in engineering and economics, but it also forces a critical reckoning with our energy and climate realities.

The path forward is not to halt digital progress but to build it more intelligently and sustainably. The future of the digital economy depends on a holistic approach that pairs technological innovation in computing and cooling with massive investments in renewable energy and grid modernization. It will require collaboration between tech companies, utility providers, governments, and local communities to ensure that the factories of the digital age can power our future without compromising it. The choices made today in the planning, construction, and operation of these unseen engines will determine the true cost—and sustainability—of our increasingly connected world.