AI is conceived as a tool of domestic power and a weapon of domination toward the outside world. I have no hesitation in saying that even before it becomes truly operational, it has already been militarized. It is an arms race to conquer minds, format them, dominate them, and seize control. For some, it is colonization; for others, it is sovereignty.In other words, it is a national priority — and from that point on, anything goes. The public’s savings, pension funds, and insurance reserves will be plundered. BlackRock itself says so, and in any case, it is presented as mandatory.

— Lisa Xclugny (

@xclugny)
May 25, 2026

U.S. stock market authorities are already changing the rules to give easier access to public and institutional savings in a distinctly Enron-like manner — and they are not even hiding it.

They are not only allowing the listing and inclusion in major indices of companies that have provided no proof of sustainable profitability, but also companies with proven records of falsified accounting.

The Economist, which has become skeptical about the AI boom after having been one of its most ardent promoters, has just published an interesting analysis.

An Accessible Analysis of Massive AI Investments and Their Accounting Risks

The five major AI players (Amazon, Google, Meta, Microsoft, and Oracle) will spend approximately $800 billion in real cash this year to build artificial intelligence infrastructure: servers, GPUs, data centers, power systems, networks, etc.This is a colossal amount, comparable to historic spending levels during major economic booms.

Yet in their income statements — the famous “profit” figures everyone watches — these investments are barely visible. Why? Because accounting rules treat them as assets, not as expenses.These companies capitalize the expenditures on their balance sheets. Depreciation (“accounting wear and tear”) only begins once the assets are put into service and is spread over several years — typically 4 to 6 years for servers and GPUs.What the Accounts Really Show

• The income statement remains “clean.” Reported profits continue to rise. Investors are happy.
• The cash flow statement tells an entirely different story. There, the truth is glaring: massive amounts of cash are flowing out of the companies’ coffers — and it is gone for good.

These giants could devote around 40% of their revenues this year to capital expenditures (Capex). That is higher than the oil industry during the shale boom or telecoms during the dot-com bubble of the early 2000s.

In short: we have moved from companies that “produced money” (strongly positive free cash flow) to companies that consume and burn cash to finance AI.

Some analysts even forecast negative cash flow quarters for Amazon, Meta, and Microsoft.

Why This Recalls Enron. Enron remains the emblematic scandal of the early 2000s, where the company used accounting tricks to hide debts and inflate profits. Off-balance-sheet structures were created to conceal financial reality.

The biggest financial scandals have always been linked to accounting subtleties, particularly the capitalization on the balance sheet of fictitious or highly subjective values.

Since 2009, U.S. and international accounting rules and practices have become “hedonic” — meaning sensitive items are marked in a more or less fanciful manner to present a flattering image. Former mark-to-market rules have been forgotten. Laxity has taken hold with total complicity; it is now a prevailing mindset. Special-purpose vehicles that obscure the real situation are not merely tolerated — they are encouraged.

In the current case, the practices are not considered fraudulent (the accounting “rules” are followed), but the mechanism is similar: massive expenditures do not immediately weigh on reported profits.

The pain is deferred to the future through slow depreciation. As long as AI revenues explode, everything looks fine. But if demand does not materialize quickly enough, or if AI models fail to generate the expected revenues, future depreciation charges will suddenly hit the bottom line — and the party will stop abruptly.

The insiders and banks will have sold their shares, but the public will be left holding the worthless bags .

AI is an enormous bet. Its financing should logically be borne by those who are the wealthiest and best able to take risks. That is the ABC of financial logic: capitalists take risks and are rewarded with exceptional returns, while the public, which cannot afford such risks, settles for modest remuneration. That is fair and honest.What is scandalous is the socialization of risks that is already underway — the exact opposite of that logic.

Large capital holders gorge themselves with limited risk, skim the best opportunities, pocket the commissions, benefit from insider information, and are preparing to offload all the duds and worthless assets onto the public.

We are returning, by anticipation, to the logic of the 2008 crisis: profits for capital, losses for the public.

The famous “risk dispersion” that lies at the heart of financialization.Sacrificing today’s cash flows for technological domination tomorrow is a rational gamble — but only for the ultra-wealthy, not for retirees, insured individuals, or small savers.

AI has enormous potential, but capitalism is not about producing goods and services to meet needs. It is a system for producing profit for capital. Nothing guarantees that AI will be the Eldorado, the gold mine that capital and its governments are currently trying to sell us.

Companies are spending because demand for AI cloud services is very strong today. But building data centers costs fortunes in energy, land, and equipment — not to mention the waste that always accompanies manias.

If model efficiency improves rapidly (achieving the same results with fewer calculations or less capital), a large part of these investments could become oversized.

During the internet bubble, fiber-optic cables were laid everywhere. Much of it remained “dark fiber” for years. Here, we are building computing power. The risk of overcapacity exists, even if AI seems more “real” than the promises of the 2000s.

Corporate leaders, who are heavily incentivized through equity and stock-based compensation, are behaving somewhat like central planners — but with other people’s money.

The system is profoundly asymmetrical: they are betting hundreds of billions on an uncertain future. It is fascinating, but risky. Shareholders tolerate it… as long as the stock price keeps rising.

The Economist summarizes: The five big firms will spend a staggering $800 billion in real cash this year on AI infrastructure. Yet their income statements barely reflect these investments, since depreciation only begins after construction and then occurs slowly.When a company buys AI servers, GPUs, buildings, power systems, and networking equipment, accounting treats them as assets, not as immediate expenses.Thus, the income statement does not immediately show the negative impact of these $800 billion in outlays.But the cash flow statement reveals the truth more directly: the money has already left the company.What is alarming is the scale: these companies will devote about 40% of their revenues to capital expenditures this year — more than the oil industry during the shale boom and more than telecoms during the internet bubble.— The Economist, May 2026