The AI revolution is shifting from algorithm design to energy consumption, with industry leaders warning that the sector's next bottleneck is electrical capacity. Over the next several years, trillions of dollars will flow into what are increasingly called AI factories: sprawling campuses packed with advanced processors, networking gear, cooling systems, and total electrical demand rivaling that of major cities.
McKinsey projects AI-related infrastructure spending could exceed $5 trillion by 2030. At the same time, real estate and infrastructure firm JLL estimates developers may need roughly 100 gigawatts of new data-center capacity to keep pace with the technology's growth. This massive power requirement is already straining grids in key markets like Northern Virginia, where data centers account for a growing share of total electricity consumption.
Behind these numbers lies a rapid construction cycle. Developers are racing to secure land, permits, and long-term power purchase agreements, often competing with residential and industrial users for grid access. The buildout mirrors the scale of earlier energy-intensive booms such as the shale revolution or the internet backbone expansion of the late 1990s. Yet unlike those cycles, this one shows no signs of peaking soon.
Geopolitically, the shift tightens the link between energy security and technological leadership. Nations with cheap, abundant power — from the United States' shale gas advantage to the Middle East's solar potential — may gain an edge in attracting AI infrastructure investments. Meanwhile, countries facing grid constraints risk seeing capital flow elsewhere.
Critics counter that efficiency gains in AI hardware or a shift toward smaller, more specialized models could temper demand before the projected buildout fully materializes. They also point to permitting bottlenecks and rising transmission costs as potential brakes on the 100 GW goal.