William Wheaton's stock-flow model says data centers have the least elastic supply curve in real estate

On December 4, 2025, the Starke County Planning Commission moved its meeting to the Knox High School auditorium because the planning office could not hold the crowd that showed up to oppose data centers. The commission's vote to recommend a one-year moratorium was unanimous. Starke was the fourth rural Indiana county to move against data centers in ten months.

There is a 1999 journal article that predicts this, and almost nobody in this industry has read it.

The machinery is 27 years old

William Wheaton spent more than three decades on the faculty of the MIT Center for Real Estate, which he helped found, and in 1999 he published "Real Estate 'Cycles': Some Fundamentals" in Real Estate Economics. The paper builds a stock-flow model of property markets. The stock of space adjusts slowly, because development takes years. The flow of new supply responds to prices, with that lag. And the paper's central result is that whether a market generates its own cycles depends on the elasticity of supply relative to demand. When supply is more elastic than demand, the lag turns price signals into overbuilding waves and the market oscillates endogenously; think of the office markets that hung a decade of vacancy on the 1980s. When supply is less elastic than demand, the self-generated oscillation dampens.

Dampens, not disappears. That distinction is where this gets expensive.

An inelastic supply curve kills the overbuilding cycle, because the market physically cannot flood itself. What it cannot kill is volatility from the demand side. When demand spikes against a supply curve that will not bend, the entire shock goes into price and availability, because quantity is not allowed to answer. Wheaton's stable market is stable in units. In dollars it can be violent.

The model makes a checkable claim

Run the model forward and it hands you a specific, testable claim. The property type with the longest development lag and the least elastic supply, facing the fastest-moving demand curve, should show the sharpest swings in pricing and availability anywhere in real estate. And if the supply curve's slope is set politically rather than by construction costs, the swings should arrive on political dates, not economic ones.

I think that property type is data centers, and I will go further: I think data centers now carry the least elastic supply curve in American real estate. Consider what sits between an AI demand signal and a delivered megawatt. A discretionary rezoning in most jurisdictions. A multi-year utility dependency for generation and transmission. And a county board that can zero out local supply with one resolution. No other asset class stacks all three, and the demand curve on the other side is the AI compute build, which moves in quarters while the supply side moves in election cycles.

Indiana steepened the curve four times in ten months

If supply elasticity is set by county boards, you should be able to watch the curve steepen in real time. Indiana obliged. Marshall County passed the state's first data-center moratorium in February 2025. White County followed on October 20, 2025, by a 2-1 vote. Putnam County adopted a one-year moratorium in November 2025 covering wind, solar, data centers, and small modular nuclear reactors. Then Starke's planning commission went unanimous on December 4, with a threshold that reads as a full ban: any data-center proposal over 5,000 square feet, and the commissioners' final adoption vote scheduled for December 15.

Four counties, ten months. Our read on a rural Indiana data-center filing is 38 out of 100, and the file calls the pattern what it is, contagion rather than coincidence. Bryce Gustafson of the Citizens Action Coalition, the organizer connecting these county fights, put the opposition's supply-side theory plainly: "Once the barn door gets open on that stuff, it makes it easier for the horses to get through." The lone White County dissenter, Commissioner Mike Smolek, made the opposite point from inside the vote: "We've had a moratorium on wind and solar for almost a year now. Has anybody come forward and said, Hey, this is what we need to change?" A moratorium, once passed, just sits there. That is precisely what makes it a supply-curve event rather than a news event. It does not need maintenance to keep megawatts off the market, and the state preemption bill that could undo it all, Indiana's HB 1333, entered the 2026 session unresolved.

Loudoun cut the curve at the core

Indiana is the frontier steepening. Loudoun County, Virginia, is the core doing the same thing, which matters more. On March 18, 2025, per Loudoun Now's coverage, the Board of Supervisors voted to end by-right data-center development in the heart of Data Center Alley. New applications now go through special-exception review, the discretionary Virginia path with mandatory public hearings and written findings. Existing applications were grandfathered. Holland & Knight's April 2025 analysis read the change as a fundamental rewrite of the county's entitlement framework, and Virginia's JLARC had put the policy groundwork on the record in its December 2024 data-center report.

In Wheaton's terms, the country's anchor data-center market deliberately cut its own supply elasticity at the exact moment demand went vertical. A by-right regime is an elastic supply curve; you file, you build. A special-exception regime is a steep one; our file puts 6 to 18 months on that discretionary path before anyone pours concrete, with an organized opposition apparatus waiting at every hearing. We score a new Loudoun site at 42. The grandfathered pipeline, meanwhile, became something Wheaton's model prices instantly: a fixed stock of the old elasticity, never to be manufactured again.

What the model says happens next

No classic bust. That outcome requires elastic supply, and this asset class has spent 2025 amputating its own. What the model predicts instead is scarcity pricing with political triggers: availability that gaps on ordinance dates, and every new moratorium quietly transferring value to whoever already holds permission. The dampened market is dampened in quantity. Somebody still absorbs the demand shock, and it will be the tenant, in price.

My bet, checkable by anyone with a transaction database: through 2027, the premium for entitled, powered data-center land over comparable unentitled land in constrained markets widens rather than compresses, and the single best predictor of local pricing gaps will be entitlement-regime changes like March 18, not construction costs.

Wheaton wrote down the machinery 27 years ago. The counties are setting the dials now. Watch what a grandfathered Loudoun application trades for this year, and tell me the supply curve isn't the whole story.