The midterms appear to have changed little about federal AI policy in the near term, but approvals of measures restricting datacenter construction across the country could spark a material shift in state and city level policies.
Going into the election day, Datacenters appeared on local ballots in many communities. As of this writing, of the 47 local initiatives seeking to restrict datacenter development voters had approved 44 measures and lead in two others that were yet to be determined. The opposition seen across states including California, Illinois, Maine, Ohio, and New York, mirrored a CBS News poll this summer that found that Americans oppose a new local data center by more than two to one.
Concern about the buildout has not belonged to one state or one party alone: Texas, which is expected to host 25%-30% of nationwide datacenter capacity by 2030, saw it’s Republican governor, Greg Abbott, suspend new approvals even before the election. Gina Hinojosa, who proposed a temporary construction ban during her campaign, defeated Governor Abbott, with estimates that as much as 13 gigawatts of Texas capacity could stall. In Georgia, Governor-elect Keisha Lance Bottoms has said she intends to sign an executive order pausing new construction and to review the state’s data center tax incentives. In Iowa, Governor-elect Rob Sand campaigned on ending tax breaks and blocking nondisclosure agreements. The question many states now face is whether incentives and utility cost allocation will tighten, and whether a middle ground can be found that supports investment while earning the confidence of local voters.
The economic stakes are meaningful. ING’s chief international economist estimates that technology investment, including computing equipment, software, and data center construction, accounts for about one-third of US GDP growth in 2026, after adjusting for imported hardware. It is also notable that data centers and power generation are the only construction categories currently expanding, while other non-residential business investment declined year-over-year for six consecutive quarters through the first quarter of 2026. Economists differ on the precise share, in part because of imported hardware and the difficulty of separating AI spending from other technology spending. It seems reasonable to expect that a slower buildout would moderate headline growth.
However, there is a silver lining for technology companies supplying the semiconductor chips and systems for these datacenters. While current resistance may impact the timing to construct the buildings and power systems necessary for the equipment, a majority of near-term capital expenditures – expected to top $1 trillion over the next year – is expected to be for hardware going into datacenters built over the last few years.
Several developments will help show which path the AI economy takes: whether Georgia’s executive order and Texas’s permit freeze hold, whether states revise tax incentives, and whether utility commissions shift grid costs onto developers. A slower buildout could raise costs for AI infrastructure companies, and could also bring more capital discipline to a sector that some investors already viewed as running hot. Reasonable investors may read these developments in different ways, and much will depend on how states balance local concerns with economic opportunity.
Investing in securities involves the risk of loss. Past performance is not indicative of future results. The opinions expressed herein are subject to change at any time without notice.
