Data Center Expansion Faces Scrutiny Amid Shifting Political Landscape

Data as of November 4, 2026.

The vote is over, the zoning is not

The 2026 midterm elections have concluded with control of both chambers changing hands, and one consequence is already visible in prices: the dollar fell to its weakest level in a month and global equities slid, with the heaviest selling concentrated in artificial intelligence names. Data centers, the physical layer underneath those names, are now exposed to a political process that had been content to approve them.

The question is what actually changed. Federal elections do not rezone a substation, set a utility's interconnection queue or issue a building permit. Those decisions sit with county boards, state utility commissions and grid operators, and they turn on local arguments about water use, ratepayer costs, diesel generators and tax abatements rather than on party control in Washington.

That division of authority is the reason the sector's political risk is real but slower than the market's reaction implies. A new compute tax or a federal efficiency mandate would require legislation, and legislation requires a coalition that does not currently exist for either. A county board declining a conditional use permit requires four votes and a meeting room.

What is genuinely at stake federally

Two federal levers matter more than rhetoric. The first is tax treatment: the expensing provisions and credits that determine the after-tax return on a multi-billion-dollar facility are set in statute, and a Congress willing to revisit them changes the hurdle rate for every project in the pipeline.

The second is the power market itself, since wholesale electricity rules and transmission planning are regulated federally in ways that determine whether new load can be served at all. A facility that cannot get firm capacity is not a facility, regardless of what the tenant has signed.

State and local governments hold the third lever, and it is the one that binds first. Property tax abatements, sales tax exemptions on equipment, water rights and the approval of on-site generation are all negotiated project by project. Those negotiations are where the cost of this boom is being redistributed, and they are invisible in a national election result.

How that redistribution fits into the wider contest over industrial policy is traced in currency as an instrument of policy, which covers the ways governments compete for capital and the costs of doing so.

The market repriced faster than the policy changed

The reaction on the day after the vote is the ordinary behaviour of a market that had priced a regulatory status quo and now has to price a range of outcomes. Semiconductor and infrastructure names fell hardest because they carry the longest duration: their valuations depend on cash flows a decade out, and a decade is long enough for a tax change, a permitting regime or an interconnection rule to alter the arithmetic.

The parallel case is instructive. Developers who own powered land rather than chips, such as Goodman Group, sell capacity rather than compute, and their contracts are signed with tenants whose own capital budgets respond to the same political signals. The exposure travels through the lease, not through the share price alone.

What the market cannot price is the sequencing. A compute tax announced in 2027 would affect projects begun in 2028, while a permit denied this month stops a project this quarter. The nearer risk is local and the larger risk is federal, and the two move on different clocks.

The constraint that outranks politics

Behind the political argument sits a physical one. The binding constraint on data center construction in most American markets is not capital, land or public opinion. It is the time required to connect new load to a grid that was not planned for it. Interconnection queues in several regional markets run to multiple years, and the equipment needed to expand transmission has its own backlog.

That constraint does double duty for the industry. It slows competitors as much as incumbents, which protects the returns of facilities already energised. It also raises the political temperature, because every large new load is a cost that existing ratepayers may be asked to share, and rate cases are decided by commissioners who answer to voters rather than to shareholders.

The financing side of the same problem is a rate question. The cost of capital that determines whether a project clears its hurdle is set by central banks, and the divergence between them is covered in four central banks in one week. A facility underwritten at one set of rates can be uneconomic at another without a single rule changing.

Who pays for the wires

The ratepayer argument is the strongest one available to opponents of data center expansion, and it is grounded in how American utilities recover costs. When a large new load requires a substation upgrade, a new transmission line or additional generating capacity, the cost enters the utility's rate base. The utility earns a regulated return on that investment, and the customers who pay for it are the ones on the system, including households that never use the facility.

Large-load customers often negotiate special contracts with dedicated facilities and minimum take obligations, which shifts some of that cost away from residential ratepayers. The negotiation is where the politics lives, and it is conducted in front of state commissions rather than in Congress.

Three variables decide the outcome. The first is the size of the load relative to the existing system, since a facility that doubles a utility's demand changes the planning assumptions entirely. The second is the contract structure, specifically whether the large customer pays for capacity it reserves but does not use. The third is the generation mix, because a system that must build new capacity to serve a single customer spreads that cost differently than one with surplus.

The same logic applies to water, where cooling requirements compete with residential supply in stressed aquifers, and to local tax abatements, where a county forgoes revenue in exchange for jobs that a highly automated facility may not provide in the numbers promised.

The financing of the underlying grid follows the rate cycle rather than the political one, and the divergence between major central banks is tracked in four central banks in one week. A transmission project approved at one cost of capital can be uneconomic at another, which is why utility commission cases have become the venue where this debate is actually resolved.

What to watch

Three things will tell us more than the election coverage did. The first is state-level legislation in the next session: abatement caps, water use limits and on-site generation rules either pass or they do not, and they are the fastest-moving instruments available.

The second is interconnection timelines published by the large regional grid operators. A shortening queue lowers the political temperature because it reduces the ratepayer subsidy argument; a lengthening queue raises it.

The third is the composition of the AI trade itself. If the heaviest selling continues in names whose cash flows are furthest out, the market is pricing regulatory duration risk. If the selling rotates into the electrical equipment and utility names that supply the buildout, the market is pricing something different and more physical: the possibility that the power simply will not be there on schedule.