A University of Pennsylvania poll found 61% of Americans now oppose new data center construction in their communities, up 12 points in a single year. It's reportedly become significant enough to be called a "sleeper issue" heading into the 2026 midterms.
At the same time, U.S. data center construction spending hit $75B in July alone, up 57% year over year. The industry is racing to build directly into a wall of growing public resistance.
Texas told its grid operator to stop approving new data center connections pending an audit. New York passed the first statewide moratorium on large new data centers. These aren't fringe events, they're becoming a pattern.
Why this doesn't touch Kilawatt Cloud
We don't build data centers. Kilawatt routes GPU workloads across RunPod, Vast.ai, Lambda, and Hyperstack, with automatic failover if one provider runs dry. Every GPU we route to already exists, already has power, already cleared whatever zoning and community approval it needed to get built.
That's not a workaround, it's a structural difference. While hyperscalers navigate grid interconnection queues, zoning boards, and now voter backlash, an aggregation layer over already-built capacity doesn't hit that same wall. We're not competing for the same scarce resource everyone's fighting over, we're routing around whoever already won that fight.
The honest caveat
This doesn't mean unlimited GPU supply forever. If community and political resistance genuinely slows new capacity coming online industry-wide, the existing capacity we route across becomes more valuable and more contested too. We're insulated from the fight, not immune to its downstream effects.
But there's a real difference between "our growth is capped by permitting delays" and "our growth is capped by how much of the internet's existing GPU capacity we can route to intelligently." The second problem is a much better one to have.
Curious if other infra-layer builders are thinking about this the same way, or if I'm missing a real exposure here.
Top comments (0)