Data centres, golf courses and the new American capital stack

On 30 June 2026, two stories crossed the wire that, on the surface, have nothing in common. One placed US data-centre construction spending above the combined federal outlay on airports, marine terminals and mass-transit systems. The other described the redesign of an existing course into a 7,660-yard, par-72 layout with expanded practice areas. Read separately, they are trivia. Read together, they describe a country whose private capital is now financing the new infrastructure of the AI economy while politically connected projects continue to attract public sympathy.

The thesis is plain. American fixed-asset investment has tilted decisively toward compute, real estate adjacent to compute, and the amenities that compute-adjacent communities demand. The headline-grabbing electricity, water, fibre, and zoning fights of 2024 and 2025 — chronicled in trade press and on Wall Street research desks — were the prelude. The capital is now arriving at scale, and it is doing so without a comparable public-works counterweight.

The data-centre boom, in one sentence

A 30 June X post by the market-data account Unusual Whales cited the underlying figure: US data-centre construction spending now exceeds what the country spends on airports, marine terminals, and mass-transit systems combined. The framing — that a single category of private buildout has outrun an entire tier of traditional public infrastructure — does the analytical work in a phrase. It reframes the AI buildout not as a tech story but as a civil-engineering story with consequences for electricity markets, water rights, and rural tax bases.

It also clarifies who is footing the bill. The hyperscalers, the REIT-financed developers, and the utilities are paying. They are paying with paper — investment-grade debt, sale-leasebacks, and preferred equity — that ultimately relies on the future cash flows of a handful of tenants whose computing bill has yet to be stressed by a recession, a regulatory shock, or a model-training plateau. The public balance sheet is not on the hook directly. The public risk, however, is on the hook indirectly: through rate-base expansions that flow into consumer electricity bills, through grid interconnections that prioritise large single loads over distributed resilience, and through municipal concession deals that trade long-term tax stability for short-term construction payroll.

The golf project, in one sentence

A separate 30 June X post by the same account noted a proposed layout redesign: par-72, 7,660 yards, 18 holes, plus a short pitch-and-putt course and expanded practice areas. A sporting project, in other words, packaged inside the same private-capital machinery that supports any premium real-estate amenity. Golf redevelopment is, in 2026, less a leisure story than a property-and-rights story — water allocations, course-restoration contracts, agronomy supply chains, and club-equity real-estate trusts that bundle recurring dues into securitised income.

Why pairing them matters

The temptation in editorial coverage is to treat each story in isolation, because each has its own actors, its own permitting regime, and its own political coalition. That is the framing the American public has been offered. But the two stories share a deeper logic: capital in the United States is increasingly being allocated to a narrow band of physical-asset categories — compute, power, premium amenities — that the existing public-works architecture was not built to finance. The traditional instruments (municipal bonds, federal grants, Highway Trust Fund–style mechanisms) served an older industrial map. The new map is being drawn by debt covenants and zoning boards.

The structural pattern is familiar. When private capital outruns public infrastructure, the public good migrates into private hands through three predictable channels. First, price — electricity, water and amenity access rise because supply cannot be quickly expanded on the public side. Second, priority — grid and permitting attention flow to the highest-paying customer. Third, exposure — when private assumptions fail, the public absorbs the residuals, whether through tariff increases, stranded-asset recovery, or local-government backstops.

The counter-narrative worth steelmanning

There is an honest argument against this framing, and it deserves airtime. The data-centre boom is, in this telling, a productivity miracle. Training and inference compute are general-purpose technologies on par with electrification and the internal combustion engine; restraining their buildout would be a 21st-century equivalent of opposing rural electrification in 1936. The boom also creates the tax base that, eventually, allows for new public investment — the same way the railroad boom of the 1880s funded the eventual rise of the Interstate Highway System in the 1950s. Pairing a data-centre story with a golf-course story is misleading because the two operate on different time horizons: compute is infrastructure; golf is amenity. The first compounds. The second depreciates.

The defense has real force. But it also assumes that the public revenues from the compute boom will be efficiently redistributed, and that the political economy of municipal tax abatement will not siphon them before they arrive. The history of 20th-century US industrial policy — from the Mills of the 1920s through the semiconductor clusters of the 1980s — is a history of partial redistribution: some regions captured rents, others captured residuals. The compute buildout looks similar at scale, only faster, only more geographically concentrated, and only with sharper conflicts over electricity and water than the 20th-century booms ever produced.

What remains uncertain

The sources cited here do not specify how much of the data-centre capital stack is debt-financed versus equity-financed, nor the share backed by investment-grade hyperscaler tenants versus speculative developers. They do not specify the geographic distribution of the boom, the status of interconnection queues at the relevant utilities, or the contractual terms of municipal concessions. They do not name the owner or sponsor of the golf-course redesign. That gap matters: the difference between a routine course refresh and a politically connected project is precisely the kind of detail that the unstressed reporting cannot resolve. The factual scaffold here — compute outpaces transit, golf courses are being redesigned at premium lengths — is solid. The interpretive overlay is offered with that caution in place.

Stakes

If the trajectory continues, two things are likely. First, US electricity bills will continue to rise faster than the headline inflation rate, because rate-base expansion flows through retail tariffs with a lag and interconnection queues are already deep at most relevant utilities. Second, the political salience of large compute campuses will rise, because the same voters paying higher bills will be the voters being asked to approve the next round of concessions. The 2026 midterms will be the first cycle in which this becomes a dominant kitchen-table issue in enough districts to register in national margins. The pattern is structural, and it is now visible in the headlines.

This publication framed these two stories as a single capital-allocation signal. The wires covered them separately; Monexus treats the convergence as the story.