For decades, data centers operated as predictable electricity consumers, connecting to the grid through long-term utility agreements as internet demand grew. Artificial intelligence has disrupted that model almost overnight. Today, operators are rethinking how power is sourced, delivered, and managed as AI workloads push infrastructure beyond what traditional energy systems were built to handle.
The scale of change is dramatic. Worldwide data center electricity consumption is projected to rise from 448 terawatt-hours in 2025 to 980 terawatt-hours by 2030, according to Gartner. Within that surge, AI energy consumption is the fastest-growing driver. AI-optimized servers alone are expected to see electricity usage jump from 93 terawatt hours in 2025 to 432 terawatt hours by 2030.
Those projections highlight the industry’s central challenge. The grid was not designed to expand at AI speed, and energy infrastructure has become a major constraint on data center growth.
The Rise of Bring Your Own Power
By 2025, operators had begun shifting from passive electricity buyers to active energy planners. Instead of waiting for utilities to expand capacity, developers are designing sites with integrated on-site power generation and storage.
McKinsey estimates that 25 to 33 percent of incremental data center demand through 2030 will be met by behind-the-meter power, representing up to 33 gigawatts of new BTM generation deployed over the next five years.
Industry surveys show the shift is already underway. The Foley 2026 Data Center Survey found that 56 percent of developers are exploring co-located or on-site generation, while North America absorbed nearly 6,000 megawatts of data center capacity in 2025 alone, roughly 50 times the volume absorbed a decade earlier.
Capital is following the shift. According to PwC, U.S. power and utilities deal value reached $141.9 billion across 35 transactions in 2025, up from about $28 billion in 2024, as companies raced to secure dispatchable power for AI workloads.
Investors increasingly recognize that the next phase of AI infrastructure investment depends less on real estate and more on a reliable energy supply. JLL’s 2026 Global Data Center Outlook estimates the sector will require up to $3 trillion in infrastructure investment by 2030, with roughly 100 gigawatts of new capacity coming online between 2026 and 2030, equating to about $1.2 trillion in real estate asset value creation.
Hydrogen Enters the Conversation
Demand for clean energy data centers adds another challenge. Operators must expand compute capacity while meeting sustainability commitments.
Gartner analysts note that emerging technologies, including green hydrogen, geothermal, and small modular reactors, could become viable fuel sources for data center microgrids by the end of the decade.
Whitaker B. Irvin Jr., President and CEO of Q Hydrogen, believes hydrogen could play a role in that model.
“Currently, 96% of hydrogen production uses fossil fuels,” Irvin explains. “As governments globally provide billions to fund clean hydrogen hubs, much of the new technology in development that is innovative, carbon-free, and affordable has yet to be realized.”
Q Hydrogen has taken a different approach. Its process produces hydrogen from water without fossil fuels, electrolysis, or government subsidies. The company has operated a test facility in Kamas, Utah, since 2016.
“In 2016, Q Hydrogen opened its operational test facility in Park City, Utah,” Irvin says. “Today it can produce between 10,000 and 50,000 kilograms of H2 daily.”
Unlike solar or wind, hydrogen generation is not weather-dependent. That reliability matters for hyperscale facilities where uptime requirements drive data center power costs.
A New Infrastructure Model
Q Hydrogen’s first commercial facility will open soon at the former paper mill site in Groveton, New Hampshire.
“The generator initially will produce around 10 megawatts of energy daily, scalable to more than 100 megawatts, and the campus will require less than five acres of land,” Irvin says.
For developers facing grid constraints, that compact footprint aligns with the distributed power models emerging across the industry.
AI has triggered a fundamental rethink of data center energy. Operators who once relied on the grid now plan energy alongside computing, moving toward distributed power systems designed to support the next generation of AI data center energy demand.
