How the US midterm results could 'cascade through' the AI and data centre industry

The midterm elections turned data-centre growth into a political issue. Voters overwhelmingly rejected local proposals connected with data-centre development, and several candidates supporting tighter rules won governor races. This matters because the United States has more data centres than any other country after rapid growth in the past year.

Ohio recorded the largest number of votes, with 21 measures rejected across 18 jurisdictions. In Texas, governor-elect Gina Hinojosa campaigned for stronger regulation and promised a construction moratorium. Georgia also elected Democrat Keisha Lance Bottoms, while Ohio elected Amy Acton. These governors can influence permits, planning rules and construction policy.

The immediate result is greater uncertainty for companies planning AI infrastructure. Investors reacted sharply, with technology shares falling amid concerns about regulation and financing. Tapas Strickland said Texas and Georgia represent about one-third of AI data-centre spending and projects, so delays there could spread through the wider AI sector.

A data centre is a secure facility containing servers, storage systems and networking equipment. It processes, stores and sends digital information. AI companies use data centres to train models, run applications and serve results to users. The article does not provide a formal definition, so this explanation uses established computing knowledge.

AI systems need far more computing power than ordinary software tasks. Their calculations run across many specialised chips, which must operate reliably and continuously. Data centres provide the equipment, cooling, electricity and network connections needed to keep that work running. Nvidia, AMD and Micron produce parts used in AI computing, according to the article.

This infrastructure is now central to the industry’s growth. The United States has more data centres than any other country after explosive expansion over the past year. If projects face tighter rules, higher funding costs or public opposition, AI companies may struggle to add capacity as quickly, while investors reassess the sector’s growth expectations.

Ballotpedia tracked at least 50 local ballot measures connected with data centres. They appeared across 13 states, including Arkansas, California, Florida, Illinois, Ohio, Texas-adjacent? No, Texas was not among the listed states, along with several others. The votes showed strong local resistance to proposed development.

Ohio had the largest concentration. Voters rejected 21 measures across 18 jurisdictions. The article also identifies one exception: Harwood, North Dakota, where voters rejected an additional review of industrial projects, including data centres. In other words, Harwood voters opposed extra scrutiny rather than rejecting a data-centre project itself.

The article says all but one of the ballot measures failed, while three measures were still due to be decided later in the year. Therefore, the safest count is at least 50 considered and at least 49 rejected, with the final total potentially changing. The results signal a broad political challenge to rapid data-centre expansion.

Governors who favour stronger regulation can make data-centre construction harder to approve. They may support moratoriums, stricter reviews or new operating requirements. These measures can extend timelines, raise costs and create uncertainty for companies deciding where to build. The article links the election results directly to investor concerns about regulation and financing.

Texas governor-elect Gina Hinojosa campaigned for increased regulation and promised a moratorium on data-centre construction. Georgia elected fellow Democrat Keisha Lance Bottoms. Ohio also elected Amy Acton, who supports moratoriums on large data-centre construction. These positions could affect infrastructure rollout through state policy and administration.

The potential impact is substantial. Tapas Strickland said Texas and Georgia together make up about one-third of AI data-centre spending and projects. If both governor’s offices slow approvals, companies may delay projects, redirect investment or face higher costs. Strickland said the effects could cascade through the broader AI sector.

Data centres are the physical base for AI computing, so fewer projects can reduce demand for chips, servers, cloud capacity and related electricity infrastructure. The article does not quantify separate effects on cloud companies or energy producers, but those industries are connected to building and operating large computing facilities. The effects would likely vary by company.

The article already shows how investors reacted. Nvidia shares fell at least 9 per cent, while AMD and Micron dropped more than 17 per cent. The Nasdaq fell at least 5.5 per cent. Analysts cited concerns about data-centre financing and regulation, while Omkar Joshi described the market move as reality setting in after AI overspending.

A prolonged slowdown could delay orders and weaken growth expectations for chipmakers and cloud providers. It could reduce expected electricity demand from new facilities and make investors more cautious. However, the article does not say the AI bubble has burst. Joshi called the move a reset rather than definitive proof of a collapse.

If Texas, Georgia or other states impose tougher rules, operators may look for jurisdictions where permits are easier, construction is allowed and infrastructure can be delivered quickly. They could also expand existing facilities instead of starting entirely new projects. The article does not identify specific alternative states, so these are general industry possibilities rather than reported plans.

The mechanism is straightforward. Data-centre companies compare regulation, available land, electricity, network connections, financing and local opposition before committing capital. A moratorium or lengthy review can make one location less attractive. Projects may then move to another state or country, although relocation can itself be expensive and slow.

The pressure to find alternatives could grow because Texas and Georgia account for about one-third of AI data-centre spending and projects, according to Tapas Strickland. The article only says stricter rules could slow infrastructure rollout. It does not confirm that operators have already chosen replacement locations or that any particular market will benefit.

AI systems process enormous quantities of data and calculations. Training and running them requires many specialised chips working together, plus storage, networking and cooling. Data centres bring these components into one dependable operation. The article refers to the parts needed for AI computing and to the rapid expansion of data centres, but it does not explain the technical foundations in detail.

Electricity keeps the chips and cooling systems operating. Financing pays for buildings, equipment, power connections and expansion before companies earn returns. That makes AI growth dependent not only on software breakthroughs but also on physical infrastructure and access to capital. Nvidia, AMD and Micron are among the companies supplying important AI-computing parts.

The scale creates new risks. The United States has more data centres than any other country after explosive growth over the past year. Investors now worry about funding costs, regulation and possible overspending. Omkar Joshi said the market was experiencing a reset of reality, rather than necessarily a bursting bubble.