The Forecast Signs the Mortgage

A forecast changes character when enough contracts are written around it. Before a data center is built, demand for autonomous software is a proposition. After land, chips, grid connections, and loans have been committed, low demand becomes a loss allocated among specific institutions. The forecast is no longer merely something to assess; utilization becomes something that exposed parties have reasons to produce.

They do not share one mortgage. The owner seeks tenants; the lender seeks repayment; a utility that expanded generation or transmission may need revenue from the projected load; a government that granted concessions may need visible jobs; a vendor with reserved capacity needs customers to consume it. Their incentives are separate, conditional, and sometimes opposed. Contracts nevertheless align them on one narrow outcome: the new capacity should not remain idle.

This matters beyond the companies that own the machinery. Cloud discounts, minimum-spend agreements, bundled software, vendor road maps, and consulting practices transmit the pressure outward. An employer may redesign work around automation without owing a cent on the data center. Yet the infrastructure can still change the employer's menu of prices and products until using machine labor appears cheaper than declining it. The server hall does not defeat an experienced worker's judgment; the commercial system built to fill it can lower the amount of proof required to ignore that judgment.

Railways did this earlier. Builders predicted traffic, then debt and fixed track gave promoters, towns, and industries reasons to route commerce toward the line. Some routes helped create markets that justified the gamble. Others made subsidy and preferential treatment look like evidence that the original traffic forecast had been right. Once capacity exists, demand is partly discovery and partly recruitment.

That ambiguity genuinely limits the indictment. General infrastructure can be a rational purchase of options: cheap, abundant capacity may enable valuable uses no planner named in advance. If unexpected demand pays for the asset, the initial forecast may have been wrong while the investment was sound. Repayment pressure is not proof of waste.

It does, however, corrupt a common test. Rising adoption after a giant buildout cannot by itself validate the belief that supposedly justified the buildout, because adoption is now being cultivated by parties exposed to low utilization. The institutional danger is not simply a bad prediction. It is a network of contracts that makes an uncooperative future too expensive to permit. A future reorganized to use the asset is no longer an independent verdict on why the asset was built.