AI and the Reinvention of Economic Power: From Resources to Ingenuity

Summary

• Five major tech firms spent over $400 billion on capital projects in 2025, the IEA says.
• Spending on that scale is expected to rise roughly 75% during 2026.
• Economists are asking whether competing demand for capital raises financing costs across other sectors.

The latest

Capital spending by five of the world's largest technology companies passed $400 billion in 2025 and is projected to climb about 75% in 2026, according to the International Energy Agency. Not all of it is earmarked for artificial intelligence, but the figures mark a shift in what technology firms now buy: land, power supply, chips and construction rather than software alone.

Details

• The numbers: The International Energy Agency puts combined capital expenditure by five major technology companies above $400 billion for 2025, with an expected increase of roughly 75% during 2026. The agency does not attribute the entire sum to artificial intelligence, treating it as a measure of the broader investment shift led by advanced computing.
• What changed: Technology companies historically scaled without heavy physical investment relative to revenues. That equation has shifted: giant data centers function as industrial plants, with chips as machinery and electricity as raw input. The practical consequence is competition for land, power contracts, chips and engineers.
• The spillover: The money does not stay inside computing. It converts into demand for construction, engineering, energy, equipment and services, generating jobs and projects far from the technology industry itself. On that reading, the build-out is reviving parts of the traditional industrial economy rather than replacing them.
• The capital question: Economists are weighing what happens to other sectors needing the same financing. Global savings are finite, and heavy borrowing demand could intensify competition for capital and push up its cost for other borrowers. Reuters analyst Mike Dolan and Axios both examined that competition in analyses published on 14 October 2026.
• The counterargument: The relationship is not a straight transfer from factories and households to AI firms. New investment generates income, income generates savings, and higher productivity may leave economies better able to finance future expansion — making the boom a test of the financial system's capacity rather than a zero-sum contest.
• The productivity bet: The economic case rests on measurable gains: factories cutting waste, hospitals shortening diagnosis times, logistics firms handling more shipments per unit of cost, and small companies doing work that once required large teams. Distribution of those gains, the analysis holds, would be neither automatic nor equal.
• The Gulf angle: Gulf states hold capital, energy, geography and a record of executing large projects. The larger return, the analysis argues, comes only if data centers anchor local companies, research, industrial and health applications and exportable computing services — not if the region limits itself to operating imported technology.
• The downside case: Some AI projects may miss profit targets, valuations may fall, data-center spending may slow, and fast-moving technology may strip value from assets already built. The historical parallel offered is the dot-com collapse and nineteenth-century railway bankruptcies: investors lost money, the infrastructure stayed.
• The next phase: Maturity would require corrected expectations, reallocated capital and a shift from racing to build capacity toward using it efficiently — the stage that would show where the real value sits.

Background

Earlier build-outs followed the same pattern: railways, electricity grids and telecommunications networks were expensive to construct and became the base for industries that could not have existed before them. The IMF's October 2026 World Economic Outlook is among the references cited in this assessment.

Between the lines

The $400 billion figure measures inputs, not outcomes. Because the IEA itself declines to attribute the full sum to AI, the number is better read as evidence that computing has become capital-intensive than as a verdict on returns. The productivity claims — faster diagnosis, leaner factories, smaller firms reaching global markets — remain projections, and the financing-cost debate stays unresolved because both sides of it are plausible.

What's next

Watch 2026 capital expenditure disclosures from the five firms against the projected 75% increase, corporate bond yields as a signal of financing competition, and whether Gulf data-center projects attract locally owned applications rather than operating contracts alone.

Source: International Energy Agency, International Monetary Fund, Reuters, Axios