US Data Centers Set to Double Power Draw by 2030 as Water Fights Slow Site Approvals

US Data Centers Set to Double Power Draw by 2030 as Water Fights Slow Site Approvals EIA's AEO2026 projects standalone server consumption accelerating through 2050 while community water disputes complicate interconnection timelines for new facilities. A single large data center can consume up to 5 million gallons of water per day, roughly 1.8 billion gallons annually and the equivalent of a town of up to 50,000 residents, and that figure is generating more immediate community pushback than the electricity demand sitting behind it.¹ That matters for electricity markets because the social friction is slowing approvals at exactly the moment demand is accelerating. The EIA's Annual Energy Outlook 2026 projects electricity consumed by data center servers increasing across the commercial building stock through 2050, with standalone data centers growing faster than all other facility types combined.⁴ Analyst estimates put data center power consumption at 35 gigawatts by 2030, more than double the 17 gigawatts recorded in 2022.² The IEA put global data center electricity consumption at approximately 415 terawatt-hours in 2024, with a projection of 945 terawatt-hours by 2030, a near-doubling in six years.⁵ The density of demand inside individual facilities is the sharper near-term signal. The IEA's 2026 update on energy and AI found that a single server rack within an advanced data center could carry a peak power demand equivalent to that of 65 households by 2027.⁵ A modern hyperscale campus houses thousands of racks. The evaporative cooling systems that manage their heat are the source of the water draw, and that consumption scales with both computational load and ambient temperature.¹ Google's proposed campus in Franklin, Indiana, which required rezoning more than 450 acres, illustrates what this looks like at the community level.¹ Local resistance tends to center on water permits rather than grid connections. Water limits are visible and tangible to residents in a way that megawatt interconnection queues are not. Investment pressure means the buildout will not slow for that resistance. Data center investment reached $197 billion in 2024, one of the fastest-growing segments of commercial real estate globally.² In 2021, the sector logged 209 deals worth more than $48 billion, a roughly 40% increase from 2020's $34 billion.² Capital is moving faster than permitting infrastructure can process it. Demand growth has followed distinct waves historically. Lawrence Berkeley National Laboratory data show U.S. data center electricity consumption rose roughly 90% between 2000 and 2005, then decelerated to 24% growth from 2005 to 2010, and moderated further to about 4% between 2010 and 2014, reaching an estimated 70 billion kilowatt-hours that year, roughly 1.8% of total U.S. electricity consumption at the time.³ The current AI-driven cycle differs from all of them: computational intensity per rack is higher, and the demand driver is a general-purpose consumer technology spreading faster than any prior enterprise buildout. AI adoption is the mechanism. Within two years of ChatGPT's November 2022 launch, around 40% of U.S. households reported using AI chatbots, according to the IEA.¹ That pace outran prior consumer technology cycles. The EIA AEO2026 shows server consumption still rising well past 2030, with the trajectory in standalone facilities steepening rather than plateauing.⁴ The watch point is whether state utility commissions begin attaching water-use conditions to grid interconnection approvals for new data center sites. If they do, the social friction problem becomes a permitting delay problem, and delay at scale shifts the timing of new generation capacity additions across the country.²