The AI buildout was on the ballot this year, and it lost. Democrats won governorships in major data center markets, like Texas and Ohio, after campaigning for moratoriums or tougher development requirements. Voters in communities across the country also backed restrictions on new projects.1
Building out AI infrastructure was already a massive undertaking, with significant hurdles around power, permitting and financing. Growing political resistance adds to the difficulty, potentially bringing longer negotiations, higher costs and further delays. But how much this ultimately affects investment remains uncertain, given the economic incentives for states and localities to accommodate development.
More bark than bite, so far
Opposition to data centers had been building well before the midterms, reflecting bipartisan concerns about electricity bills, grid reliability and water consumption. Wood Mackenzie counts more than 400 state and local moratorium actions since January, against just 45 across all of 2023–2025.
Yet many restrictions are temporary, and statewide moratoriums have struggled to advance: of 16 states considering legislation, only one has passed a bill through its legislature. Meanwhile, data center capacity under construction grew nearly 50% in the first half of 2026.2
The potential economic benefits are also substantial. A U.S. Chamber of Commerce study estimates that a mid-sized 50 MW data center generates nearly $700 million in annual state GDP and supports more than 5,200 jobs per year during construction.3
Finding the right model for states and developers
Newly elected officials will likely take a firmer stance to deliver on campaign promises. Over time, however, local governments must weigh constituent concerns against the investment, tax revenues and jobs data centers can bring.
States are still determining what a fair deal with developers should look like. Utility regulators are seeking commitments to ensure households don't shoulder incremental grid costs. Developers, meanwhile, are offering more substantial community benefits: Amazon and Google have each announced roughly $1 billion in community investment initiatives. Many are also exploring dedicated power generation to reduce reliance on constrained grids.
These arrangements could help secure community support, but also raise costs and prolong negotiations. The evolution of public sentiment toward AI will influence how welcoming politicians can afford to be.
What does this mean for the AI trade?
Political resistance could reshape where capacity gets built, which projects move forward and which developers are best positioned to deliver.
- The buildout from 2028 onward is most exposed. Projects already under construction are generally less vulnerable to new restrictions. Future developments face greater uncertainty around permitting, power availability and financing, with delays and higher costs potentially pressuring returns.
- A shakeout of speculative projects may be healthy. Wood Mackenzie estimates more than 430 GW of proposed U.S. data center capacity, but just 59% is under active development.4 Utilities are already weeding out speculative power requests, and higher hurdles could accelerate this process. This should favor well-capitalized hyperscalers and developers with stronger economics, while reducing the risk of speculative overbuilding.
- Investment can move elsewhere. Extraordinary demand for compute should continue to fuel construction, potentially drawing investment toward more welcoming states or international markets. Developers will also likely pursue dedicated power generation, albeit at additional cost.
Ultimately, AI adoption, monetization and access to capital will remain the primary drivers of investment beyond the midterms. But how developers and policymakers navigate growing infrastructure challenges will help determine whether the next phase of AI growth remains constrained by energy shortages, or enabled by energy abundance.
