The AI appetite is getting bigger
The headline here is simple: AI workloads are turning data centers into power-hungry beasts. According to S&P Global Market Intelligence, US data center capacity is expected to more than double between 2026 and 2030, climbing from 62,242 MW in March 2026 to 151,734 MW by the end of the decade.
That’s not just a chart-friendly forecast. It’s the kind of number that makes utility planners, grid operators, and infrastructure investors sit up a little straighter. Because chips are great, but they still need electrons.
Why investors should care
If this demand surge plays out, the winners and losers won’t be limited to hyperscalers and AI chipmakers. The ripple effects could hit:
- utilities and independent power producers chasing new load growth
- grid equipment makers dealing with bottlenecks and upgrade cycles
- data center developers trying to secure enough land, permits, and power hookups
- local markets where AI campuses suddenly become the biggest thing in town
In other words, the AI story is getting less like a software fairy tale and more like an industrial buildout with very real plumbing problems.
The bottleneck isn’t the dream — it’s the wires
A lot of the AI trade has been built on the idea that demand is endless. Fair enough. But demand without transmission lines, substations, and generation is just a very expensive wish list.
That’s why this forecast matters: it suggests the next leg of AI growth may be constrained less by chip supply and more by whether the grid can keep up. And if it can’t, expect delays, higher costs, and a lot more competition for power access.
Big picture: the AI boom isn’t just computing its way forward anymore — it’s running headfirst into the laws of physics, and the power bill is arriving fast.