US Data Centers Set to Triple Grid Share as Tech Emissions Jump
Government estimates project that data centers' share of US electricity consumption could nearly triple by 2028 from 4.6% in 2024, according to figures cited by the IEA, as artificial intelligence workloads push into energy territory that is straining utility capacity and corporate sustainability pledges alike.4
The gap between what companies promised and what their infrastructure actually draws has become hard to ignore. Google's greenhouse gas emissions jumped nearly 50% even as the company kept its 2030 carbon neutrality target in public communications; Meta's rose more than 60%, Amazon's 33%, and Microsoft's more than 23%.4 Google has since described its clean-energy target as a "moonshot."4
The fuel mix behind US data centers makes the problem concrete. Natural gas supplied more than 40% of the electricity consumed by US data centers in 2024, while coal accounted for 30% globally, the IEA said.4 Those figures undercut the argument, common among hyperscalers, that the buildout is a green economic transition by another name.
AI adoption is driving demand faster than grid infrastructure can absorb it. The IEA found that within two years of ChatGPT's launch in 2022, about 40% of households in the US and United Kingdom were already using AI chatbots.3 The EIA's Annual Energy Outlook 2026 projects that server electricity consumption alone will keep rising through 2050, with standalone data centers accounting for more of that growth than distributed facilities inside commercial buildings.5 Some analysts estimate nationwide US electricity use could rise as much as 20% over the next decade, with data centers a primary driver.4
The local-level friction is visible at places like Franklin, Indiana, where Google sought to rezone more than 450 acres for a data center campus.3 That pattern — a hyperscaler negotiating directly with municipalities over power access, zoning and transmission — is repeating itself across multiple US states and has brought state and federal lawmakers into a debate they had largely left to utilities to manage.
Utilities have responded by pursuing scale. NextEra Energy announced on Monday, May 18 (2026-05-18), a $67 billion deal to acquire Dominion Energy in an all-stock transaction that would create what the two companies described as the world's largest regulated electric utility.7 NextEra's shares fell more than 5% after the announcement; Dominion's rose just under 10%.7 Analysts at Deloitte noted that scale has become critical for utilities seeking capital access and the ability to execute large transmission projects.6
The investment market has also moved. Fluence Energy, a battery storage manufacturer, saw its stock climb 98% in a single week in May 2026 following a record order backlog disclosure and new supply agreements with two major hyperscaler customers.1,2 Management reaffirmed a 2026 revenue target of $3.2 billion to $3.6 billion, saying 85% of the midpoint was already contracted.1 Fluence's surge reflects a broader rotation into companies positioned to supply grid-scale storage and power infrastructure for AI campuses.2
The sequencing problem is straightforward. Cleaner baseload capacity — nuclear, long-duration storage — carries lead times of years to decades. Data center construction cycles run on 18-to-24-month timescales. Natural gas fills the interval. Companies that committed to clean-power procurement targets in 2020 and 2021 are now managing facilities that came online into a grid that cannot yet match the pace of their expansion.4
For lawmakers, the difficulty is that the energy demand is real and growing regardless of what targets any company publishes. A regulatory approach that throttles data center growth faces pushback from economic development interests; one that waves it through risks straining grids and locking in gas capacity for another decade. Companies say they must stay flexible as they build facilities that can consume more power than entire cities, according to the Fortune reporting.4
Fluence's Q3 results — expected to reflect deferred Q2 shipments as delivery schedules normalise — will offer one near-term read on whether storage investment is scaling fast enough to matter.1 The more durable signal is whether any legislative framework emerges to price the grid costs of new large loads, or whether that question gets left to interconnection queues and utility rate cases to answer indirectly.