Voters Push Back, But AI Buildout Continues
TREE NEWS reports: On November 3, voters across the United States were asked 41 times about data center construction in their communities. The answer was nearly unanimous: 76% of exit-poll respondents opposed building new data centers in their own neighborhoods. Democrats won the Texas governor’s race—home to the largest planned computing hub in the country—and took control of Georgia’s utility regulator. The market’s initial reaction was to treat this as a fatal blow to the AI buildout cycle. That framing is wrong.
Projects already under construction will proceed. What is being repriced is the forward supply slated for 2028 and beyond. This makes existing computing facilities with secured power more valuable, not less. The election did not end the AI trade; it changed who gets paid and when.
Two Distinct Battles
There are two separate fights unfolding simultaneously. The state-level battle over who pays for grid expansion can ultimately be settled with money, likely by mid-2027. But the local community battle over where facilities are sited cannot be bought off. That resistance will keep actual supply for 2028–2030 below planned levels.
Construction Won’t Stop: 2027 Projects and the Texas Slowdown
All projects scheduled for delivery before the end of 2027 are already in active construction. Nationwide, about 87 GW of computing capacity is under construction, with a net addition of 7 GW in October alone. The first cracks appeared in Texas, where new construction starts plunged from 3.4 GW in September to just 0.6 GW in October.
Forward Supply Delays: 2028–2030 Pipeline Repricing
The real risk is in the 2028–2030 pipeline. Our estimate ranges from a 10% delay in a mild scenario to 28% if local moratoriums are severe; we lean toward the lower end. The first industry model revision after the vote cut forward deliveries by about 11%.
Stranded Capacity Hits Record High
Stranded capacity—projects stalled by permitting and queue backlogs—has reached a record 61 GW, about 13% of the total planned pipeline, jumping 9 GW in October alone. Investors often overlook that only about 30% of headline announcements were ever likely to be delivered by 2030. The nominal numbers were never the real supply.
Most of the damage is delay: of 51 GW announced in January for 2027 delivery, 31 GW has slipped to 2028 or later, and about 9 GW has been formally withdrawn, canceled, or shelved.
Public Acceptance: New Sites vs. Existing Communities
For new sites, public acceptance is largely lost: 76% oppose new data centers in their communities. But in mature communities with existing data centers, support outweighs opposition at 49% to 42%. When projects commit to self-supplied power, community benefits, and local veto rights, they win 64% support in head-to-head comparisons.
The biggest open risk is whether projects already under construction could be halted. In Ashville, Ohio, nearly 82% of voters demanded the revocation of an already-under-construction campus agreement. One town is not a national trend, but it is the first real stress test for projects in progress.
Who Captures the Value: Repricing of Existing Power
Hardware stocks fell sharply as the market priced in reduced 2028 construction and slower capex. On November 4, Vertiv, Quanta, and GE Vernova dropped 9% to 13%, with some infrastructure contractors losing nearly half their market value in four weeks—clearly beyond what pure engineering cutbacks would explain. Sentiment has shifted to risk-off.
Who ultimately profits? The fight over who pays can be solved with money, but the community conflict over where facilities are built cannot be subsidized away. This limits future supply and gives mature assets with power in hand higher monopoly value and pricing power.
