U.S.-China Relations
NextFin News - Data center chief financial officers are preparing for a harder funding environment after the Democratic sweep of Congress in the 2026 midterm elections, as rising political scrutiny of AI infrastructure projects collides with more expensive debt and a weaker equity market. The rout that followed was swift and severe: the Nasdaq Composite fell 5.5% on Thursday, wiping out an estimated $2.5 trillion in market value and pushing the index into correction territory, nearly 12% below its record close on October 6. Every sector in the S&P 500 finished in the red. But for the CFOs running the balance sheets behind the AI buildout, the deeper question is what comes next — and the answer points to a longer, more expensive road to capital.
The combination of events matters more than either one alone. A project that looked financeable under one set of permitting assumptions can become a stranded-cost risk if approvals stall or local opposition forces redesigns. Add in a credit market that is no longer willing to underwrite speculative developers at single-digit rates, and the industry's central problem shifts from "how do we build fast enough" to "which projects survive the funding filter." That is a structural change in who gets to build, layered on top of what is likely a cyclical funding squeeze.
The New Political Reality for Data Center Capital
The midterm results delivered a direct political shock to the AI infrastructure trade. Ballotpedia tracked 50 local ballot measures related to data centers across 13 states in 2026, with 41 of them — 82% — appearing on the November 3 general election ballot. As of November 4, voters had decided 47 of the measures, and in 46 cases a majority favored greater limits on data center development, either by approving restrictions or by rejecting measures that would have supported development. Only Harwood, North Dakota, a town of 845 people, voted in favor of a new project: a $3 billion, 1.8-million-square-foot Applied Digital facility on its outskirts.
The local backlash is now being amplified at the state level. Newly elected governors in Ohio, Georgia, Michigan, New Mexico and Texas campaigned on a promise to stop — or at least pause — data center construction statewide, while those in New York and Oregon pledged to prolong pauses already in place. Texas alone accounts for more than 25% of planned US data center capacity by 2030, which makes Governor-elect Gina Hinojosa's campaign pledge of a construction moratorium the single largest swing factor in the national buildout. Before the election, many developers had expected existing restrictions to ease; the result was the opposite.
"The big election outcome was, especially with the Democrats winning a lot of the governorships, that the slowdown of the data center buildout is the real key theme that's going to stick," said Michael O'Rourke, chief market strategist at JonesTrading.
For CFOs, the political shift changes the risk calculus on projects that can take years and billions of dollars to complete. The transmission mechanism runs through three channels: permitting timelines, which stretch out and push revenue recognition further into the future; power interconnection, where audits and reviews can defer grid access; and the cost of capital itself, which rises as political risk gets priced into project finance. The result is a higher hurdle rate for new commitments and a longer timeline from announcement to shovels in the ground.
The Financing Squeeze: Debt Costs and a Wary Credit Market
The political shock arrived on top of a credit market that was already tightening. As of November 5, the ICE BofA US Corporate Index option-adjusted spread sat at 93 basis points, with BBB-rated credit at 116 basis points and the high-yield index at 348 basis points. CCC-rated debt — the deepest end of the speculative-grade market where many smaller data center developers live — was trading at a spread of 1,320 basis points, a level that signals a market pricing in meaningful default risk rather than just illiquidity. BB-rated credit stood at 213 basis points.
The repricing of AI-adjacent credit has been building for months. CoreWeave, the AI cloud specialist, received a B+ issuer credit rating from S&P in early November with $1.5 billion of debt rated B — a reminder that even the darlings of the AI infrastructure boom are being underwritten as speculative-grade borrowers. Earlier in the year, CoreWeave had to raise yields to more than 10% to entice investors into a $2.6 billion leveraged loan sale, offering coupons roughly 5.5 percentage points above benchmark SOFR rates. That deal, priced in July, now looks like a marker for how much the market's appetite has shifted.
For CFOs across the data center ecosystem, the message is clear: cheap money is no longer available to paper over execution risk. Companies that built business plans on low single-digit financing costs now face a world where debt service can consume a much larger share of project cash flows, forcing a reassessment of which projects clear the return threshold. The squeeze is most acute for developers whose projects are not yet cash-flowing — precisely the group that needs new capital most.
Five CFOs, Five Different Pressure Points
The five companies flagged in the market's data center watchlist — Prologis (PLD), AIB Data Centers (AIB), Dropbox (DBX), CleanSpark (CLSK) and Lumen Technologies (LUMN) — illustrate how unevenly the pressure is being distributed. The same macro headwind produces five very different balance-sheet stories.
Prologis enters the slowdown from a position of strength. The industrial real estate giant, rated A2 by Moody's and A by S&P, started $1.3 billion in data center development projects in the first quarter of 2026, part of $2.1 billion in total new development starts. CFO Tim Arndt described the quarter as "fantastic," and the company has secured or advanced 5.8 gigawatts of data center power as of September 30, with 1.7 GW secured and 4.1 GW in advanced stages. For an investment-grade developer with deep pockets and access to both equity and debt markets, a financing slowdown is more of a headwind than an existential threat — it slows the pace of deployment rather than blocking it. Prologis can afford to wait.
AIB Data Centers tells a different story. The company, which changed its name from BlockchAIn Digital Infrastructure effective June 25, 2026, is a small-cap developer with no traditional debt and $52.8 million in cash as of June 30 — but also just $2.9 million in quarterly revenue, down 39% year over year, and a net loss of $3.5 million, or $0.07 per share. Its stock has lost 84% year to date and was trading around $0.72 in early November. For a developer at this scale, equity markets are the primary funding source, and a rout that cuts market capitalization by more than half directly impairs the ability to raise the capital needed to turn identified capacity into built facilities. AIB has a 65-megawatt, 15-year electric service agreement at its CLT-01 facility and roughly 570 MW of identified AI and high-performance computing capacity potential — but identified capacity is not built capacity, and the gap between the two is where the funding winter bites hardest.
CleanSpark represents the bitcoin-mining-to-AI-infrastructure conversion trade, and its CFO, Gary Vecchiarelli, has been explicit about the discipline required. "Capital stewardship remains central to how we allocate resources and evaluate growth," he said in the company's third-quarter fiscal 2026 results. CleanSpark closed $2.276 billion of 7.875% senior secured notes due 2031 in late September to fund its data center build-out — a deal priced at nearly 8% that the company says supports roughly $6.6 billion of contracted revenue over the initial term. The transition is capital-intensive, and the company posted a net loss of $378 million in the second quarter of fiscal 2026. With 1.8 GW of power under contract as of September, CleanSpark has the raw inputs; the question for its balance sheet is whether AI hosting margins can arrive fast enough to offset deteriorating bitcoin mining economics before the debt service burden becomes dominant.
Lumen Technologies is the fiber-connectivity play, and its CFO, Chris Stansbury, framed the moment as a pivot point. "This is a defining moment for Lumen, as we close the chapter on our financial past and pivot to an executable path to growth," Stansbury said at the company's 2026 Investor Day. Lumen is planning to expand its fiber network to 58 million miles by 2031 and has pushed strategic revenue to 53% of total business revenue, up from 51% in the first quarter. For Lumen, the AI buildout slowdown is a demand-timing issue rather than a funding issue: hyperscalers may delay new interconnection orders, but the underlying need for fiber backbone to connect distributed AI facilities remains. The risk is timing mismatch — revenue recognition tied to buildout milestones that may now arrive later than modeled.
Dropbox is the outlier. It is a mature cloud storage business navigating a CFO transition as Timothy Regan steps down, with Ross Tennenbaum taking over the finance function. The company announced a $1.5 billion buyback plan and is working to repay outstanding convertible senior notes due in 2026. Dropbox's exposure to the data center buildout is indirect, running through its own infrastructure spending and its attempt to pivot from storage to AI-enhanced productivity. Its challenge is less about financing new facilities and more about proving that AI products such as Dash can offset a slowly declining core before cash generation weakens.
Second-Order Effects: Who Actually Bears the Pain
The first-order read of the rout is straightforward: less capital, more scrutiny, slower construction. The second-order effect is more consequential, and it cuts against the consensus narrative. A slowdown in data center construction does not hurt all participants equally — it redistributes leverage within the stack.
Hyperscalers and the largest cloud providers, which have been the biggest spenders on AI infrastructure, may actually benefit from a slower buildout. If smaller rivals and speculative developers are forced to delay or abandon projects, the largest players gain relative share of scarce power and permitting capacity. The constraint shifts from "who can raise money" to "who has the balance sheet to wait." That dynamic favors the Prologis and Lumen types of the world — investment-grade operators with long-duration capital — over the AIB types racing to monetize announcements. The same headwind that threatens a small developer can consolidate advantage for a large one.
The Anthropic IPO, filed in October 2026 with a disclosed $80 billion first-half loss and 900% net revenue retention, sits at the center of this tension. The filing gave the market its first clear look at the capital intensity of the frontier-model race: SpaceX has committed to spend up to $1.25 billion per month on Anthropic's cloud compute capacity at data centers near Memphis, Tennessee. If the IPO market cools alongside the equity rout, the most capital-hungry players may find themselves forced to rely on private funding at less generous terms — or to lean harder on the hyperscalers that control both capital and distribution. The market has already shown it wants a more attractive valuation to commit to such offerings, with some investors viewing a recently discussed $1.5 trillion valuation as being at the high end of expectations.
The Counter-Thesis: This Is a Pause, Not a Reversal
The strongest argument against the "tougher road" narrative is that demand for AI compute has not disappeared. OpenAI CFO Sarah Friar said the company will press ahead with its US data-center expansion despite the midterm sweep, and the hyperscalers have signaled continued commitment to capacity growth. The ballot measures are local and procedural — they slow projects, they do not ban the technology. If power purchase agreements remain signed and tenant demand holds, the buildout resumes once political uncertainty clears. On this view, the rout is a cyclical overreaction, and the mean-reversion trade — buying the beaten-down names once the political noise fades — is the obvious play.
That argument has force, but it conflates the direction of demand with the cost of capital. Even if every delayed project eventually gets built, the companies that financed at the peak of optimism face a permanent impairment: debt raised at 8% to 10% cannot be refinanced away if the underlying cash flows took two years longer to materialize. Equity raised at a depressed market capitalization locks in dilution that no subsequent recovery can undo for existing holders. The slowdown does not need to be permanent to be damaging — it needs only to be long enough to break the refinancing assumptions embedded in current balance sheets. A two-year delay is not a pause for a company whose debt matures in three.
What to Watch
The next signals will come from three places. First, the credit markets: if CCC spreads widen materially beyond 1,320 basis points, smaller developers will find themselves locked out entirely, and the funding winter becomes a solvency event rather than a timing issue. Second, the permitting pipeline: a cluster of state-level actions in Texas, Ohio and Georgia would confirm that the ballot measures were the opening move rather than the whole game, and would put the more than 25% of US planned capacity in Texas directly at risk. Third, the Anthropic IPO pricing — a successful float at a strong valuation would reopen risk appetite for AI infrastructure capital; a pulled or deeply discounted offering would confirm that the window has closed for the most capital-intensive players.
The base case is a two-speed market: investment-grade operators with secured power and long-duration capital continue building, while speculative developers face a funding winter that separates viable projects from announcements. The downside case is a broader credit event in AI-adjacent debt that forces asset sales and write-downs across the stack. The upside case is a rapid political de-escalation that restores permitting certainty and brings equity markets back to life.
The central judgment: this is a cyclical funding squeeze layered on top of a structural shift in who gets to build. The demand for AI infrastructure is not in question; the question is which balance sheets survive long enough to serve it. The market has already answered part of that question — the 84% year-to-date drawdown in the smallest developers is not a pricing anomaly, it is a repricing of survival odds. The falsifying signal would be a sustained tightening of CCC spreads back below 800 basis points alongside a return of permitting approvals in Texas and Ohio: that combination would indicate the squeeze was cyclical noise rather than a regime change. Until then, CFOs are right to prepare for a tougher road.
Explore more exclusive insights at nextfin.ai.
