Billions Spent, Profits Delayed: How to Read the AI Investment Surge

Global private investment in artificial intelligence reached about 344.7 billion US dollars in 2025, up 127.5% on the previous year, according to the Stanford AI Index 2026. Total corporate investment, which includes mergers and acquisitions, minority stakes and public offerings, reached about 581.7 billion. The Index notes that private-sector figures may understate what some countries, such as China, have spent, since government funding does not all appear in them. Yet these figures do not tell us where the profits will gather.

Four layers, four economies

The field is usually divided into four layers, a simplification of our own. The first is computing and energy, which is capital-intensive: Google reported more than 150 billion dollars of capital expenditure in 2025, according to the Index, and the International Energy Agency considers that about 20% of planned data-centre projects could be delayed unless grid risks are addressed (the consumption figures are in the article "The AI Race: Who Owns the Chips, the Energy and the Rules?"). The return on this layer depends on demand keeping pace with construction. The second is chips and their concentration, a question we examined in the article "Chips and Sovereignty: Who Holds the Keys to Computing?".

The third is models and applications, whose capabilities are converging and whose cost of use is falling, narrowing the margin of anyone who relies on a model owned by someone else; see "AI Startups: The Best Model Is Not Enough". The fourth is data and sector expertise, whose value is measured not by volume but by rarity and by its link to a decision that must be taken, as in "Data Is Not the "New Oil": Where the Comparison Holds and Where It Misleads".

The gap between spending and returns

The Index, citing McKinsey's survey of the previous year, reports that 88% of responding organisations use AI in at least one business function, against 78% a year earlier. McKinsey's 2026 survey, fielded between 4 May and 8 June 2026 among 1,719 participants in 97 countries, found that 37% attribute some effect on operating profit to AI, almost unchanged on the previous year, and that 44% say AI is scaling across their organisation, up from 38%. Use is wide; documented effect on profit is narrower. That need not mean returns are absent, since they are hard to measure, as we explained in "AI Return on Investment: How to Measure It Without Fooling Ourselves".

What regulators say

The record of the Bank of England's Financial Policy Committee (2 October 2025) says that equity valuations "appear stretched", particularly for technology companies focused on AI, and that "the risk of a sharp market correction has increased". It noted that the earnings yield implied by the cyclically adjusted price-earnings ratio (CAPE) was close to a 25-year low and "comparable to the peak of the dot-com bubble", although the forward price-earnings ratio, at 25 times, remained below that bubble's level. It added that the share of the five largest companies in the S&P 500 was the highest in 50 years, close to 30%, and that risks include disappointing adoption of the technology or stiffer competition. It is an assessment of risk, not a forecast of timing, and knowledge of the unseen belongs to God alone.

Five questions to test any claim

This is our own wording of the questions, not a recommendation:

• If the price of the technology falls to almost nothing, who keeps the profit in the chain?
• Does this asset survive a downturn, or does it rest on a flow of easy money?
• What is hard for a competitor to copy: unique data, deep integration with the work, or an algorithm that can be cloned?
• How exposed is it to geopolitics, energy and supply chains?
• Is the figure shown realised or projected, and who measured it and when?

This article does not recommend buying or selling any asset. Anyone who wishes to put money to work should consult a licensed specialist and check that the company's activity and financing are sound under Sharia, for not everything that grows is lawful. Provision rests with God, and part of taking the means is not to build a decision on a headline.