Artificial intelligence is rapidly becoming one of the most important new sources of electricity demand in the world. The International Energy Agency projects that electricity generation supplying data centers could more than double by 2030, creating an enormous challenge for utilities, grid operators, technology companies and energy developers. The question is no longer simply how many data centers can be built. It's whether the energy infrastructure required to power them can be built fast enough.
Artificial intelligence is transforming the technology industry, but behind every AI model, GPU cluster and hyperscale data center is something far more fundamental: electricity. The extraordinary investment flowing into AI infrastructure is creating an equally significant challenge for the energy industry as developers attempt to secure hundreds of megawatts — and increasingly gigawatts — of reliable power for new data center campuses.
According to the International Energy Agency's Energy and AI report, global electricity generation required to supply data centers is projected to increase from approximately 460 TWh in 2024 to more than 1,000 TWh in 2030 and 1,300 TWh by 2035 under its Base Case. Over the next five years, renewables are expected to meet nearly half of the additional electricity demand, followed by natural gas and coal, while nuclear power becomes increasingly important toward the end of this decade and beyond. The scale of this expansion means AI is no longer simply a technology story. It's quickly becoming one of the most consequential energy infrastructure stories of the decade.
Renewables Will Supply a Major Share of AI's Growth
Renewable energy is positioned to become the fastest-growing source of electricity supplying data centers through the end of the decade. The IEA projects renewable generation serving data centers to grow at an average annual rate of approximately 22% between 2024 and 2030, meeting nearly 50% of the growth in data center electricity demand during that period. Wind, solar PV and hydroelectric generation are expected to provide most of that expansion, supported in part by technology companies financing new renewable projects through power purchase agreements and direct investment in co-located generation.
This development makes economic as well as environmental sense. Technology companies have become some of the world's largest corporate buyers of renewable electricity, and large-scale solar and wind projects can provide attractive long-term energy costs. But renewable generation alone doesn't solve the AI power challenge because hyperscale data centers operate continuously and require extremely high levels of reliability, while solar and wind production varies according to weather and time of day. Battery storage can shift electricity across several hours, provide rapid-response capacity and help stabilize large loads, but hundreds of megawatts of continuous computing demand require a broader portfolio of generation, transmission, storage and grid infrastructure.
The result is likely to be an increasingly integrated energy system rather than one dominated by a single technology. Renewable generation can provide a growing share of low-cost electricity, while batteries, firm generation and the electric grid provide the capacity and reliability required to support continuous digital infrastructure.
Natural Gas Will Remain Important in the Near Term
One of the most important conclusions from the IEA analysis is that the rapid expansion of AI won't immediately translate into an entirely renewable electricity system. Natural gas and coal together are expected to provide more than 40% of the additional electricity required by data centers through 2030, using both increased utilization of existing generating assets and new power plants. This reflects the fundamental challenge facing the industry: electricity demand from AI is expanding faster than many new transmission, renewable and nuclear projects can realistically be developed.
The United States illustrates this dynamic particularly well. Natural gas currently supplies more than 40% of the electricity physically consumed by U.S. data centers, followed by renewables at approximately 24%, nuclear at around 20% and coal at roughly 15%. As demand accelerates, the IEA expects natural gas to provide the largest amount of additional U.S. generation through 2030, adding more than 130 TWh of annual generation, while renewables add approximately 110 TWh over the same period.
This helps explain why natural gas has become such an important part of the data center development conversation. Developers facing multiyear utility interconnection timelines can't necessarily wait for transmission upgrades or new generation resources to arrive. In regions with adequate gas supply and pipeline infrastructure, on-site or nearby natural gas generation may provide a pathway to firm capacity faster than traditional utility development timelines — the same time-to-power logic driving developers toward fuel cells and other on-site generation. The challenge will be balancing the need for reliability and speed with corporate decarbonization objectives, emissions requirements and the economics of building dedicated generation.
Nuclear Could Become a Major Piece of the Long-Term AI Power Mix
Perhaps the most significant longer-term change identified by the IEA is the growing role of nuclear energy after 2030. Nuclear generation already supplies approximately 15% of the electricity physically consumed by data centers globally, and the emergence of small modular reactors could potentially expand that role substantially. The attraction is straightforward: data centers require enormous quantities of electricity continuously, while nuclear facilities provide firm power with high capacity factors and comparatively low operational carbon emissions.
Technology companies have clearly recognized this potential. The IEA reports that technology companies have announced plans to support more than 20 GW of SMR capacity, with the agency's Base Case expecting SMRs to begin entering the data center electricity mix after 2030. If advanced nuclear technologies can overcome challenges involving permitting, financing, supply chains, construction schedules and commercialization, AI could become one of the most significant new sources of nuclear electricity demand in decades.
Nuclear and renewables could eventually complement one another particularly well. Renewable generation can provide increasingly inexpensive electricity when resources are available, while nuclear can provide the firm, around-the-clock generation profile required by large computing campuses. Combined with battery storage and grid connectivity, that portfolio could provide both reliability and significantly lower emissions than an energy system dependent primarily on fossil generation.
The Bigger Constraint May Be the Grid
Generating enough electricity is only part of the AI power challenge. Delivering that electricity to the precise location where a data center needs it may prove considerably more difficult. A new 500 MW or 1 GW campus doesn't simply require an equivalent amount of generation somewhere within the regional power market. It requires sufficient generation, transmission capacity, substations, transformers, interconnection infrastructure and local delivery capability to reliably move enormous quantities of electricity to one specific location.
This distinction is becoming increasingly important because a region can have substantial overall generating capacity while still lacking the transmission or substation capacity necessary to serve another major load. New generation can also spend years waiting in interconnection queues before it can connect to the grid, while transmission projects frequently require even longer development timelines. For an AI industry moving at technology-sector speed, traditional power infrastructure timelines are becoming a serious constraint.
The IEA's higher-growth Lift-Off Case demonstrates the potential consequences. Under that scenario, global electricity generation associated with data centers approaches 2,000 TWh by 2035, approximately 45% higher than the Base Case. Renewable generation expands significantly, but long grid-connection queues limit how quickly additional clean generation can reach data centers, causing fossil generation to supply a larger portion of the near-term increase.
This highlights one of the most important realities facing the industry: the future competition for data center sites may increasingly become a competition for deliverable power, rather than simply a competition for inexpensive electricity or inexpensive land.
Time-to-Power Is Becoming a Competitive Advantage
Historically, data center site selection focused heavily on fiber connectivity, tax incentives, land availability, water, workforce and electricity prices. Those factors remain important, but time-to-power is rapidly becoming one of the most valuable characteristics of a development site. A parcel with inexpensive land and attractive tax incentives may have limited value if the utility can't deliver the required capacity for five or six years.
For AI companies investing billions of dollars in GPUs, servers and supporting infrastructure, delaying a campus can create enormous opportunity costs. This changes the development equation because a more expensive location capable of delivering several hundred megawatts within a shorter period may ultimately be more economically attractive than a lower-cost site trapped behind transmission or interconnection constraints.
This dynamic could fundamentally change how data centers are developed. Instead of selecting land first and then determining how electricity will reach the site, developers may increasingly identify available power first and build the computing infrastructure around the energy resource. Existing power plants, retired industrial facilities, brownfields, transmission corridors, large substations and sites with existing utility infrastructure could therefore become increasingly valuable as AI development accelerates.
The next generation of data center development may consequently look less like traditional commercial real estate development and considerably more like integrated energy infrastructure development.
Behind-the-Meter Power Could Become Part of the Solution
Difficulty obtaining utility power is also increasing interest in behind-the-meter generation and hybrid energy campuses. Large AI facilities could increasingly combine utility electricity with dedicated natural gas generation, renewable energy, fuel cells, battery storage and eventually advanced nuclear or other firm-generation technologies. Instead of depending entirely on a single utility connection, developers could create a diversified energy architecture designed around reliability, cost and speed of deployment.
Battery energy storage could become particularly important in these configurations. AI workloads can create rapid changes in electricity consumption, and batteries can respond almost instantaneously. BESS can potentially provide power-quality support, manage load fluctuations, reduce peak demand, integrate on-site renewable generation, provide backup capability and interact with the grid when market conditions make doing so economically attractive.
Sophisticated energy-management systems could coordinate these resources based on electricity prices, computing loads, renewable generation, battery state of charge, grid conditions and reliability requirements. The future hyperscale data center could therefore operate less like a conventional commercial building and more like a sophisticated microgrid or independent power system connected to the larger utility network.
Geography Will Matter More Than Ever
The IEA analysis demonstrates how dramatically the data center electricity mix varies by region. In China, coal currently provides nearly 70% of the electricity physically supplying data centers, followed by renewables at nearly 20% and nuclear at close to 10%. By contrast, Europe is positioned for renewables and nuclear to supply the overwhelming majority of incremental data center electricity, with their combined share reaching approximately 85% by 2030. Japan and Korea are also expected to see renewables and nuclear provide nearly 60% of data center electricity by 2030.
The United States sits between these models with a diversified combination of natural gas, renewables, nuclear and coal. By 2035, however, the IEA expects low-emissions resources to provide more than half of U.S. data center electricity as renewable generation continues expanding and advanced nuclear begins contributing additional supply. These regional differences could eventually influence where AI infrastructure is built because locations with abundant generation, available transmission capacity, supportive permitting environments and access to multiple energy resources may be able to bring campuses online faster and at lower long-term energy costs.
Electricity availability could therefore become an increasingly important component of economic development policy. States, utilities and countries competing for AI investment will need to consider not only tax incentives and real estate development but also transmission planning, generation capacity, permitting timelines and the ability to provide hundreds of megawatts of reliable power without compromising affordability or reliability for existing customers.
AI Is Becoming an Energy Infrastructure Industry
One of the most important conclusions from the IEA analysis is that AI development can no longer be separated from energy development. Building a hyperscale AI campus may require billions of dollars of computing equipment, but those servers have little value without reliable electricity. As individual campuses approach several hundred megawatts and potentially exceed a gigawatt, the infrastructure supplying the electricity becomes a project of comparable strategic importance.
This creates opportunities throughout the energy value chain. Utilities will need additional generation, substations and transmission infrastructure. Renewable developers will need new projects and PPAs. Battery manufacturers and integrators will have opportunities to provide storage and grid-support systems. Natural gas developers could see additional demand for firm generation, while nuclear developers may gain an entirely new class of corporate customers willing to support first-of-a-kind projects. Grid operators will simultaneously have to determine how to integrate enormous new loads while maintaining reliability and controlling costs for everyone already connected to the system.
The boundaries between the technology industry and the energy industry are therefore beginning to disappear. AI companies increasingly need to understand generation development, PPAs, interconnection, transmission, storage and utility planning, while energy companies increasingly need to understand the unique load profiles, deployment schedules and reliability requirements of hyperscale computing infrastructure.
The Solution Will Be an Energy Portfolio, Not a Single Technology
Energy discussions are often framed as competitions between technologies: renewables versus natural gas, batteries versus nuclear, or grid power versus behind-the-meter generation. The scale and speed of AI electricity demand suggest that this framing may increasingly miss the point. The IEA's projections show renewables providing the largest portion of incremental global data center electricity demand while natural gas and coal remain important near-term sources and nuclear becomes increasingly significant after 2030. Across the agency's various demand scenarios, renewables remain pivotal, but firm generation continues to play an important role in meeting rapid load growth.
The most successful AI energy strategies may therefore combine several resources. Solar and wind can provide increasingly inexpensive renewable electricity, batteries can provide flexibility and extremely fast response, natural gas can provide firm near-term capacity, and nuclear could eventually provide large quantities of continuous low-emissions generation. The electric grid connects these resources while providing access to a much broader portfolio of generating assets.
Rather than searching for one technology capable of solving the AI power challenge, the industry may need to focus on designing energy systems capable of intelligently coordinating multiple technologies. The competitive advantage will increasingly come from delivering the right combination of reliability, cost, scalability, sustainability and — perhaps most importantly — speed.
Conclusion
The AI revolution is rapidly becoming one of the most consequential energy infrastructure stories of the decade. The IEA projects electricity generation supplying data centers to increase from approximately 460 TWh in 2024 to more than 1,000 TWh by 2030 and 1,300 TWh by 2035 in its Base Case. Renewables are expected to meet nearly half of the additional demand through 2030, but natural gas and other conventional resources will remain important as utilities and developers race to build infrastructure fast enough to support the digital economy.
Worth noting: the IEA has since published updated analysis showing electricity consumption bottlenecks across the value chain are moderating the most aggressive near-term scenarios, even as booming investment continues — its most recent projections see data center electricity consumption roughly doubling from about 485 TWh in 2025 to 950 TWh by 2030, with AI-focused demand specifically tripling over that period. The direction of travel hasn't changed; if anything, it confirms the core thesis of this piece even more precisely than the original forecast.
The longer-term energy mix could look considerably different. Renewable generation will continue expanding, battery storage could become increasingly integrated into data center power systems, and advanced nuclear may provide another source of firm, low-emissions electricity after 2030. At the same time, transmission constraints and interconnection delays could accelerate behind-the-meter generation, microgrids and integrated energy campuses that combine multiple resources rather than relying exclusively on traditional utility service.
For the data center industry, the message is becoming increasingly clear: compute capacity is ultimately constrained by power capacity. AI developers can purchase more GPUs and construct more buildings, but those investments can't produce value without reliable electricity. The companies, utilities, energy developers and communities capable of solving the time-to-power challenge will therefore have an enormous competitive advantage as AI infrastructure expands.
The next generation of artificial intelligence will not simply be built around faster chips, larger models and more powerful servers. It will be built around the generation, storage, transmission and energy-management systems capable of keeping all of that computing infrastructure running.
Written by Chris Kalowes, founder of WattThe?! — 15+ years in utility-scale battery energy storage (BESS), renewable energy, and AI infrastructure, across utilities, IPPs, EPCs, developers, and technology providers.
Source: International Energy Agency, "Energy and AI" (2025) and "Key Questions on Energy and AI" (2026).