Data Center Opposition Meets Wall Street: Why $198B in Stalled Projects Is Now an IPO Problem

Data center opposition has stopped being a public-relations nuisance and become an underwriting input. In the second quarter of 2026, at least 45 U.S. projects worth roughly $68 billion were blocked or delayed, according to the research group Data Center Watch, a project of the AI intelligence firm 10a Labs. The same quarter delivered a quieter signal from capital markets: at least three data center-adjacent issuers postponed or shelved their listings, and one nuclear power supplier halted its IPO indefinitely.

Those two developments are usually reported in separate columns. Placed on the same page, they reshape the problem. Local permitting risk and cost-of-capital risk are now arriving at the same balance sheet in the same quarter, which makes the IPO queue a sharper test of the AI build-out thesis than any model release.

A $198 Billion First Half

Data Center Watch's second-quarter tally is not an outlier. Between January and March, at least 75 projects worth about $130 billion were disrupted. Across the first half of 2026 the count reaches at least 120 projects and roughly $198 billion in combined value. The coverage matters more than the total: Q2 disruptions accounted for more than half of every new large-scale data center the firm tracked during those three months.

Opposition has become institutional. Data Center Watch counts 843 identified opposition groups spread across 49 states, with Hawaii the sole exception. Roughly 30 statehouses introduced or adopted legislation, resolutions or executive actions covering siting, electricity, water and infrastructure cost-sharing. Miquel Vila, the firm's lead analyst, tracks both the arrival of new groups and the growth of online petition signatures.

California moved from study to statute. Governor Gavin Newsom signed seven data center bills that his office billed as the most comprehensive data center laws in the nation. None of them ban construction. They add enforceable reporting on energy and water consumption and require operators to carry their share of utility costs. The mechanism is cost allocation rather than prohibition, which converts local resistance into a recurring line item instead of a one-time permitting fight.

Some communities are not waiting for an application to arrive. Moratoriums are being passed preemptively, in places where no developer has filed a permit or expressed interest. For a developer that inverts the usual sequence: feasibility work now begins with a political read of the county rather than a land option.

State intervention is shifting from symbolic resolutions toward enforceable siting, power and cost-allocation rules. That shift is what makes local resistance durable, because a resolution can be ignored and a tariff schedule cannot.

Compute Overtakes Housing

The macro picture explains the stakes. Inflation-adjusted spending on information processing equipment, a category that includes data centers and computer hardware, reached $752 billion in the second quarter, up 51% from early 2021. Real private residential fixed investment was $748 billion over the same span, down 18% from its early-2021 peak, according to Bureau of Economic Analysis data.

Adam Shapiro, a vice president at the San Francisco Fed, frames the crossover as a shift in what drives U.S. growth, with investment moving out of housing and into computers. The two categories respond very differently to borrowing costs. The 30-year mortgage rate sits near 7% and the 10-year Treasury yield is at its highest since 2007, which holds residential construction down. AI infrastructure spending has been far less rate-sensitive, even as hyperscalers issue debt to top up their cash reserves. Alphabet reported negative cash flow earlier in 2026. Treasury Secretary Scott Bessent has pointed to how readily AI companies borrow regardless of the cost.

Hyperscaler infrastructure spending is heading toward $1 trillion a year. That is the demand side of the ledger. The supply side is a county board meeting.

Where Data Center Opposition Meets the IPO Queue

Public markets are where the two risks collide. SB Energy, a subsidiary of Japan's SoftBank, pushed back an offering that had been scheduled for this month; the deal was framed around a $50 billion valuation, and investor questions centered on that number. Holtec, a nuclear power company positioning itself to supply electricity for AI operations, said it would halt its IPO plans indefinitely and cited uncertainty around data center development. A third power provider whose main customers are AI data centers slowed its listing, though it could still price next month.

Taken together, the delays are not a verdict on AI compute demand. They are a verdict on the financing structure built on top of it. An issuer raising against contracted future power sales needs two things to hold at once: a predictable permitting timeline and a stable cost of capital. Both moved against it this quarter.

The rate backdrop sharpens the valuation question. With the 10-year Treasury yield at its highest since 2007, the discount rate applied to a power plant's contracted cash flows rises. An issuer with a permitting calendar it does not control has less room to argue for a premium multiple.

What the Two Risks Cost

Permitting risk and rate risk interact in a way that makes each one worse. A delayed project does not disappear; it sits in the queue while interest accrues and while ordered equipment depreciates on someone else's books. A higher cost of capital then raises the hurdle rate that the delayed project must clear to justify completion. Each extra quarter of local review makes the financing more expensive, and each increase in financing cost makes a marginal project easier to abandon.

That is where the two stories converge into a single thesis. Community veto power has graduated from a rounding error to a line item in AI infrastructure underwriting. Operators that treated data center opposition as a communications problem now have to model it as a schedule risk with a dollar figure attached, and the states are supplying the enforcement mechanism. California's reporting and cost-sharing requirements hand other states a template that is harder to attack as a ban, which makes it easier to copy.

For enterprise buyers, the consequences arrive through procurement rather than press coverage. Power-intensive capacity in contested jurisdictions carries longer lead times and higher cost-sharing obligations, and those costs surface in colocation contracts and utility tariffs instead of headline prices. Abbas Jaffery, a principal advisory director at Info-Tech Research Group, has argued that data center capacity has moved from a routine IT planning decision to a combined energy, community and economic question.

The counterargument deserves a hearing. Demand for AI computing power is not in question, and the build-out case rests on multi-year contracts that do not expire because a county delays a hearing. Several of the projects disrupted in Q2 will return with revised terms. The disruption figure counts delay alongside outright rejection, so $198 billion measures friction rather than value destroyed.

Friction compounds when it meets leverage, though. The clearest evidence is not in the project tallies. It is that companies whose entire pitch is selling power to data centers are choosing to stay private rather than test a public valuation.

Why this matters

The build-out thesis is now priced by two calendars at once: the permitting calendar and the rate calendar, neither of which any single operator controls. For the companies driving the $1 trillion-a-year infrastructure wave, site selection, community negotiation and capital structure can no longer run as separate workstreams, because each one now feeds directly into the cost of the other. The next meaningful data point is not a benchmark score but whether the issuers waiting behind SB Energy and Holtec can find a valuation that survives a permitting timeline they do not set.

Photo by Supradoc on Unsplash

Related Articles

• AI data center backlash is Big Tech's costliest constraint
• Anti-Data-Center Protests Go National With 142 Demonstrations Across 42 States
• OpenAI IPO Delay: Inside the $1 Trillion Standoff Reshaping AI Finance

✔Human Verified

Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.