Saad Ullah
⬤ Morgan Stanley Research is sounding the alarm about a massive electricity crunch headed for America's AI data centers. The numbers paint a stark picture: between 2025 and 2028, these facilities will need 69 gigawatts of power to keep up with exploding AI demand. The problem? Only 25 GW is realistically available—10 GW from dedicated generation projects and another 15 GW from whatever spare capacity the existing grid can offer. That leaves a gaping 44-gigawatt hole.
⬤ To put that shortfall in perspective, we're talking about the equivalent output of 44 nuclear power plants. Building out enough generation and grid infrastructure to fill that gap isn't cheap—Morgan Stanley estimates roughly $60 billion per gigawatt. Do the math, and you're looking at about $2.6 trillion just to solve the power problem.
⬤ But the spending doesn't stop there. The AI data centers themselves need to be built, which could run another $2 trillion. Combined, the industry faces a staggering $4.6 trillion investment requirement in the near term just to keep pace with demand.
⬤ Electricity is quickly becoming the bottleneck for AI growth. If power generation and grid capacity can't scale as fast as AI's appetite for compute, we're headed for delays, skyrocketing costs, and serious strain on regional energy systems. The infrastructure simply isn't ready for what's coming.
Saad Ullah
