Data Center Statistics (2026)

Data centers are the physical backbone of the internet and, increasingly, of artificial intelligence: warehouse-scale buildings full of servers, storage, and networking gear. There are now more than 11,000 of them worldwide, and the sector has shifted from steady growth into a full-blown boom as generative AI drives unprecedented demand for compute.

Two numbers frame the story. Globally, data centers used about 415 TWh of electricity in 2024, roughly 1.5% of world power, and that is set to more than double by 2030. Meanwhile the capital pouring in is staggering: McKinsey estimates the buildout could require about $6.7 trillion by 2030 (see the sections below).

The United States dominates. As of late 2025, aggregators counted about 5,427 registered data centers in the US, roughly half the global total and more than the next several countries combined (see the chart and table below). Germany (529), the United Kingdom (523), and France (322) lead Europe, which holds about 3,362 facilities across 45 countries.

These counts come from commercial directories (Cloudscene-based) and should be read as approximate, since definitions of a facility vary. Even so, the concentration is real: the US, a handful of European hubs, and China host most of the world's compute, and the gap has widened as American cloud giants race to build AI capacity at home.

Electricity is the defining constraint of the AI era. The IEA estimates global data center consumption at about 415 TWh in 2024, roughly 1.5% of world electricity, growing around 12% a year since 2017. In its central case, that more than doubles to about 945 TWh by 2030 and climbs toward 1,193 TWh by 2035 (see the chart and table below).

To put 945 TWh in context, it is close to the entire electricity consumption of Japan. Goldman Sachs Research is even more aggressive, projecting demand could rise 165% by 2030 versus 2023, to roughly 1,350 TWh. The US accounts for nearly half of global data center power, China about a quarter, and Europe about 15%.

In the United States, the DOE and Lawrence Berkeley National Lab found data centers used about 176 TWh in 2023, equal to 4.4% of national electricity, up from just 58 TWh (about 1.9%) in 2014. Growth accelerated from a 7% annual pace in 2014-2018 to 18% in 2018-2023 (see the table below).

The forecast is the headline: by 2028, US data centers could consume between 325 and 580 TWh, or 6.7% to 12% of all US electricity, translating to roughly 74-132 GW of power demand. That is a once-in-a-generation load addition to a grid that had seen flat demand for two decades.

The surge is overwhelmingly about artificial intelligence. AI-specific workloads made up an estimated 5-15% of data center power in recent years, but the IEA projects that could reach 35-50% by 2030. Electricity use in AI-accelerated servers (the racks of GPUs that train and run models) is growing about 30% a year.

The physics are unforgiving: a rack of AI accelerators can draw 10 times the power of a traditional server rack, and training a frontier model consumes as much electricity as thousands of homes use in a year. That density is why new AI campuses are measured in gigawatts, not megawatts, and why power availability now decides where they get built.

The buildout is concentrated among a small set of hyperscalers. Synergy Research counted 1,136 hyperscale data centers operational at the end of 2024, with 137 new ones opened that year and a known pipeline of 504 more. Total hyperscale capacity has doubled in under four years and is on track to double again (see the table below).

Ownership is remarkably concentrated: the US holds 54% of global hyperscale capacity, and just three companies, Amazon, Microsoft, and Google, control 59% of it. Meta, Alibaba, Tencent, Apple, and ByteDance round out the top operators. Generative AI is the main reason new facilities keep getting larger.

The capital numbers are historic. Combined Big Tech capital spending rose from about $162 billion in 2022 to roughly $448 billion in 2025, and analysts project about $725 billion in 2026, an increase of roughly 77% in a single year (see the chart below). Around three-quarters of it is tied directly to AI infrastructure.

For 2026, guidance points to Amazon near $200 billion, Google around $175-185 billion, Meta $115-135 billion, and Microsoft $110-120 billion. This is the largest private infrastructure buildout in modern corporate history, and it is why data centers have become a core theme for equity investors.

Zoom out and the totals become almost abstract. McKinsey estimates that meeting compute demand could require about $6.7 trillion of data center capital by 2030, of which $5.2 trillion is for AI-capable facilities and $1.5 trillion for traditional IT. That is built on a projected 156 GW of AI data center capacity demand by 2030.

Goldman Sachs frames it in power terms: it expects data center demand to grow about 50% to 92 GW by 2027 and forecasts a 17% compound annual growth rate through 2028. Whether these figures fully materialize is the central debate, but the direction, a multi-trillion-dollar decade, is broadly agreed.

As a market, data centers are large and compounding. Research firms peg the 2025 global market at roughly $365-418 billion and project it reaching $517-692 billion by 2030, with compound growth rates clustered around 10-12% a year (see the table below). Estimates differ because firms define the market differently.

These dollar figures are proprietary research estimates rather than government data, so treat them as directional. The consistent signal across every firm is double-digit annual growth, driven by hyperscale cloud expansion, enterprise cloud migration, and the pivot to AI and machine-learning workloads.

Geography inside the US is lopsided too. Northern Virginia, the original 'Data Center Alley,' is the largest data center market on earth, with about 4,040 MW of inventory (CBRE) and a record 1,102 MW of net absorption in 2025, up 144% year over year. It alone represents roughly 13% of global operational capacity (see the table below).

Demand so outstrips supply that Northern Virginia's vacancy rate sat at 0.5% with just 21.5 MW available at year-end, pushing large-deal asking rates to $155-185 per kW. Growth is now spilling into power-rich markets: Atlanta more than tripled to about 1,279 MW, and Dallas-Fort Worth passed 1 GW.

Power is not the only resource. Cooling all those servers takes water: Google's data centers used about 6.1 billion gallons (23.1 billion liters) in 2024, up 20% year over year, and Microsoft's water use rose 34% from 2022 to 2024. A single large data center can draw up to 5 million gallons a day, comparable to a town of tens of thousands of people.

Efficiency has improved even as scale exploded. Power usage effectiveness (PUE), the ratio of total facility energy to IT energy, averages about 1.58 industry-wide but 1.09-1.12 at hyperscalers (Google reported 1.09 in 2025). Water usage effectiveness averages about 1.9 liters per kWh, but Google (0.80) and Microsoft (0.49) run far leaner.

This growth lands on an electricity system that was not built for it. US data center load could add the equivalent of a mid-sized country's demand within a few years, and the DOE's 74-132 GW projection for 2028 would be one of the largest new sources of grid load in decades. Utilities are extending gas plants and delaying coal retirements to keep up.

The IEA notes the 2024 energy mix behind data centers was about 60% fossil fuels, 27% renewables, and 15% nuclear, and it sees roughly 38 GW of captive gas projects in development, about a quarter earmarked for data centers. Nuclear, including small modular reactors, is increasingly part of hyperscaler power deals.

Data centers accounted for about 0.5% of global energy-related CO2 emissions in 2024, and the IEA's central case sees that reaching roughly 1% by 2030 (1.4% in a faster-growth scenario). The footprint is modest globally but concentrated locally, where a single AI campus can rival a small city's demand.

Over time the IEA expects the mix to shift toward roughly 60% clean power by 2035 as renewables, nuclear, and storage scale up alongside gas. For now, the tension between AI's compute appetite and clean-energy timelines is one of the defining infrastructure challenges of the decade.

The data center boom touches far more than cloud stocks. The value chain runs from chipmakers and networking to power generation, utilities, cooling and electrical equipment, real estate investment trusts that own the buildings, and the natural-gas and nuclear suppliers that feed them. A multi-trillion-dollar buildout creates winners and stranded bets across all of it.

The open question is whether AI revenue will justify roughly $725 billion of annual capex, a debate that will define returns for the whole theme. For a thesis-driven investor, the practical move is to define what you actually believe (compute demand, power bottlenecks, or efficiency gains) and build a basket around that view rather than chasing a single headline name. Walnut is not an investment adviser, and none of this is a recommendation to buy or sell any security.