AI Infra Winners and Losers from U.S. Midtern Elections

Generally speaking, voters in the 2026 midterm elections seemed to send a message about new AI data centers Late Tuesday evening, 33 ballot measures to restrict data centers were passed by voters or were leading in ongoing vote counts, according to  Data Centers on the Ballot. 

The issue is what that might portend for future AI infrastructure development. Broadly, we might argue that:

• The cost of developing new capacity increases

• Project schedules become less predictable

• The geographic distribution of investment changes

• Costs escalate

• Political risk increases.

The cost of adding new units of compute seemingly is destined to increase, as a result, despite countervailing trends that have been pushing costs down. 

It is too early to argue that the new rules will halt new data center construction. Indeed, most of the rules seem to focus on conditions for such development rather than outright bans. But it might be rational to expect some level of delay in capacity expansion. Projects take longer to approve and complete, but ultimately get built.

Where they get built might well change, though. Companies probably will have to  spend more on power, transmission, cooling, land and community compensation. So areas where terms and conditions are more favorable might get a boost. 

Although all suppliers of AI training and inference operations will face higher obstacles, one might argue that the neocloud suppliers face the most danger. CoreWeave and other neocloud suppliers have less financial flexibility than the hyperscalers; few or no other existing sources of revenue other than available compute capacity and so face cash flow issues if new supply is delayed. 

Company	Construction delays	Higher infrastructure costs	Financing sensitivity	Ability to relocate or adapt
CoreWeave / neoclouds	Very high	High	Very high	Limited to moderate
Amazon	High	High	Moderate	High
Google	High	High	Moderate	High
Oracle	Very high	High	High	Moderate
Microsoft	High	High	Moderate	High
Meta	High	High	Moderate	High

One might also hypothesize that the hyperscalers actually could benefit from the opposition, as smaller developers will face relatively greater hurdles, in the form of:

• More demand for established data center operators with permitted sites

• Greater value for companies with existing power agreements

• More consolidation among neoclouds

• Greater bargaining power for hyperscalers who generate their own power

• Higher barriers to entry for new AI infrastructure competitors.

So among the possible outcomes is some strengthening of the market position of the hyperscalers, at the expense of neocloud providers.