Forbes: AI Data Centers May Cost $10 Trillion
Forbes reported that the U.S. is rushing into a wave of AI data center construction, with costs potentially exceeding $10 trillion, and could surpass the scale of the late 19th-century railroad boom as early as 2028. Money is being spent on land, electricity, water, and cement.
Author Giacomo Tognini added that there is over $1 trillion in debt financing, and the power needed within two years may exceed California's total electricity usage, with land area by 2030 potentially larger than Washington, D.C., of which only about 35% may actually be built. Beyond labs and new clouds, wealth is being generated in areas like wiring and concrete.
Calculations by Stijn Van Nieuwerburgh from Columbia University are used to compare with history. Brookings' related analysis totals the data center construction, power systems, networks, dedicated chips, and other equipment from 2025 to 2032 to about $10.3 trillion, averaging about 3.6% of U.S. GDP, which is a larger economic scale than previous construction waves like canals, railroads, electrification, highways, and telecommunications.
Interviewees named in the article include Ryan Alfred from Atrium, DataBank, and Kiran Raju. The major buyers remain Amazon, Google, Microsoft, Meta, and Oracle, all of which have announced capital expenditures in the hundreds of billions. A 35% completion rate means that announced expenditures do not equate to the amount of cement poured.
The comparison to the railroad boom only applies to the proportion of capital expenditures, not to the capacity already sold. At that time, track laying could precede freight, while this round of data centers can precede model revenues. The debt exceeding $1 trillion is the part that is more sensitive than cash reserves.
The buyers are large-scale cloud vendors needing power and land, while the sellers are contractors for land, power grids, water, and cement. The event is driven by expenditure forecasts, not by individual new contracts. Funding extends from chip orders to civil construction, benefiting wiring, concrete, and data centers with secured power, while pressuring the 65% that only appear in announcements and cannot access power and permits.
Source: Public Information
ABAB AI Insight
The late 19th-century railroad boom in the U.S. laid miles ahead of demand, leading to bankruptcies and reorganizations, with only companies occupying main lines surviving. Van Nieuwerburgh places this round at a larger proportion: averaging about 3.6% of GDP, higher than previous constructions from canals to telecommunications. Forbes has pushed the timeline to 2028, earlier than the spending window ending in 2032.
The capital path has extended from labs to heavy assets. Chips remain a bottleneck, but Tognini writes wealth in wiring, concrete, land, electricity, and water. Over $1 trillion in debt means the cycle is not solely determined by cash reserves. Amazon, Google, Microsoft, Meta, and Oracle are the main subjects of announced expenditures, while hosting providers like DataBank cater to demand for those not wanting to build themselves. A 35% completion rate is a constraint in this article, not an industry guideline.
Comparable is the fiber overbuilding of the late 1990s, and railroad companies in the 1880s used bonds to lay miles ahead of freight. The current position is in the announcement phase of an expansion period. California's two-year electricity usage and the land area of Washington, D.C. illustrate that physical constraints have surpassed model announcements.
Structurally, this is a reconstruction of the industry chain. Pricing power has shifted from model parameters to electricity access, water permits, and cement pouring. If only 35% is completed, valuations will remain at powered sites, while the remaining announced capacity will not recover debt. Railroads left main lines, not every track-laying company. This round will leave grid interfaces and completed data centers, not every capital expenditure plan.
ABAB News · Cognitive Laws
- Announced expenditures are not cement; the completion rate is.
- A bet larger than railroads still needs to first access electricity.
- Prosperity laid out by debt leaves only main lines during restructuring.