The AI Engine... Investing in the Energy Needs of Artificial Intelligence
Data centers consumed an estimated 485 terawatt-hours (TWh) in 2025, accounting for approximately 1.5 percent of global electricity consumption (according to the International Energy Agency). However, more significant than the current share is the pace of growth. The IEA’s main projections estimate that consumption will more than double, reaching around 945 TWh by 2030—slightly more than Japan’s current total consumption—and rising to 1,200 TWh by 2035.
The United States accounts for nearly 45 percent of global data center electricity consumption, followed by China at approximately 25 percent, and Europe at around 15 percent, according to the IEA. While estimates regarding US data center electricity consumption by 2030 vary, they all point to a strong upward trend. The Lawrence Berkeley National Laboratory projects that data centers could represent about 11.8 percent of total electricity demand (within a range of 9.5 to 15.3 percent), while the Electric Power Research Institute (EPRI) estimates the share between 9 and 17 percent. McKinsey, meanwhile, forecasts a range of 11 to 12 percent.
The paradox is that these projections start from a current baseline of only 3 to 5 percent, reflecting the sector’s exceptional growth speed. What was once a marginal share of national electricity demand just a few years ago is now poised to represent roughly one-eighth of total electricity demand in the United States in the coming years.
Despite the wide range of forecasts and the final outcome depending on various factors, the baseline and most likely scenario indicates that consumption will multiply several times over in the next four to five years.
This demand is transforming the industry from an abstract, virtual sector into a purely physical one. Operating artificial intelligence at scale requires power generation, transmission lines, transformers, cooling systems, water, and land, thereby expanding investment beyond a narrow group of technology companies.
In terms of power generation, no single fuel source dominates the landscape. The IEA expects renewables to meet about half of the growth in data center demand, adding more than 450 TWh by 2035, driven by short construction periods and lower costs. Natural gas covers most of the remaining share, particularly in the United States, where it is supported by policy and cheap supply. Nuclear energy will contribute over a longer timeframe, as the sector revives interest in Small Modular Reactors (SMRs)—modular units built in stages—which are expected to begin appearing by 2030 (according to the IEA). Beneficiaries are distributed among utility companies, independent power producers, gas and nuclear supply chains, and equipment manufacturers.
Electricity grids have also become a major bottleneck to expansion, benefiting entities capable of providing effective solutions to these challenges. Developing new transmission lines in advanced economies takes between four and eight years, while wait times for transformers and cables have doubled over the past three years.
Meanwhile, equipment that had previously constituted a minor component of industrial manufacturing has suddenly become scarce. Utility companies, long regarded as defensive assets with slow growth, are being revalued as demand growth returns for the first time in decades. Artificial intelligence can also help alleviate some of the resulting pressure; the International Energy Agency estimates that AI-assisted management of power grids could free up to 175 gigawatts of transmission capacity without the need to build new lines.
The scale of spending is reshaping how these projects are financed. The estimated $3.9 trillion over the next five years is too large to be covered by corporate budgets alone.
This gap between ambition and internal cash flows is attracting external capital. Here, the artificial intelligence theme intersects with private markets, becoming more significant for investment portfolios than any demand forecasts. According to Vanguard data, the average annual total debt issuance of the top five hyperscalers was approximately $35 billion annually between 2020 and 2024, before jumping to $93 billion in 2025 and reaching around $132 billion in the first seven months of 2026.
Broader estimates, which include chip manufacturers, developers, and utility companies, are several times higher. Companies that enjoyed structural cash surpluses just a few years ago are now among the largest borrowers in the market.
New financing structures are emerging alongside debt. In August 2026, Nvidia announced partnerships with several global asset managers to launch platforms aimed at attracting over $500 billion in investor capital, providing cloud operators and enterprises with dedicated funding resources for systems based on Nvidia technologies. These arrangements treat compute capacity as productive infrastructure capable of generating long-term, usage-linked revenues. Although implementation is still in its early stages, it demonstrates that this expansion is attracting capital traditionally associated with the energy, transportation, and real estate sectors.
For diversified investment portfolios, the most significant exposure to this sector may lie in infrastructure, private credit, and physical assets, rather than being limited to publicly traded technology stocks.
This is not merely an American story. The shift toward overseas markets directly impacts investors in this region, as the International Energy Agency expects more than half of the energy consumed by data centers to be outside the United States, as operators seek energy, land, and greater control over computational resources.
Here, the Gulf states emerge as among the largest beneficiaries of this global shift. Saudi Arabia’s HUMAIN company is aiming to develop AI data center capacity reaching 1.9 gigawatts by 2030, scaling up to 6.6 gigawatts by 2034. Meanwhile, the United Arab Emirates has announced the creation of a 5-gigawatt complex in Abu Dhabi, one of the largest AI data center projects outside the United States. Qatar has also allocated substantial investments to expand its large-scale data center capabilities. The region enjoys several competitive advantages, including relatively low energy costs, advanced telecommunications infrastructure, and a strategic geographic location that provides access to a broad segment of the global population. For regional investors, these developments represent an exceptional opportunity to capitalize on one of the most significant global structural trends, which is forming and growing in their immediate vicinity.
However, this growth trajectory is not without challenges, necessitating a careful assessment of the associated risks.
The first risk lies in the level of demand itself. Projections of electricity consumption linked to artificial intelligence rely on assumptions regarding technology adoption rates, revenue-generation potential, and operational efficiency levels. These variables may change over time, particularly if rapid advancements in chip design and intelligent models reduce energy consumption at a faster pace than anticipated.
The second risk pertains to execution efficiency. Challenges related to connecting projects to power grids, lengthy lead times for essential equipment supply, and a shortage of specialized skills and labor could delay the implementation of many projects. Estimates from the International Energy Agency indicate that a significant portion of the planned additional data center capacity faces delay risks due to these factors.
The third risk is financial. AI-supporting infrastructure requires massive capital investments, making future returns highly sensitive to any deviation from core assumptions. There is also a social and regulatory dimension of equal importance: if the rapid growth in electricity demand is not managed efficiently, it could lead to higher energy costs for households and businesses, potentially necessitating further regulatory intervention and oversight.
Despite the significance of these risks, they do not alter the overall growth trend currently witnessed by the sector; rather, they underscore the need to treat it as a long-term investment opportunity in infrastructure. Consequently, a more prudent approach would be to build diversified investments across various segments of the value chain, rather than focusing on direct bets on a limited number of companies that may lead the AI race.
• Global demand for data center electricity is expected to more than double, rising from approximately 485 terawatt-hours in 2025 to around 945 terawatt-hours by 2030. The United States is projected to account for 9% to 17% of its national energy consumption by 2030, compared to the current 4-5%.
• Meeting this energy demand will rely on renewable energy, natural gas, and nuclear power in the future, reviving utilities, energy equipment, and grid infrastructure sectors as growth areas after decades of stagnation.
• The scale of investments required through 2030 exceeds what technology companies’ balance sheets can bear alone, opening the door for infrastructure investors, private credit, and physical assets, rather than limiting exposure to listed technology equities.
• The Gulf region is emerging as a key hub as capacity shifts outside the United States, giving regional investors access to a global trend close to home.
• Risks are real: broad uncertainty around AI demand volumes, network and equipment bottlenecks, and high capital intensity all argue for a long-term, diversified approach rather than concentrated, direct bets.
