Consumers are slowly but surely learning the benefits of artificial intelligence, or as the federal government recently branded it "super intelligence." But every castle built of sand must face the coming tides and the backlash, and the false promises of AI are no different.
After three years of pouring capital into machines and ending careers at the expense of people's meaning and purpose, Wall Street faces a number it cannot talk its way around: a $4.9 trillion gap between what AI data centers must earn and what they are forecast to bring in. The promised riches of AI have not materialized, and the borrowing to chase those mythical profits is crowding the same bond market that finances the federal government. “Superintelligence” now sounds less like a breakthrough and more like a marketing ploy for a faltering mega-investment. A correction is coming, and the workers and small investors will be asked to absorb the cost.
Key points:
- Gartner estimates AI data center owners need $8.5 trillion in revenue through 2030 but will see no more than $3.6 trillion.
- Bain and Co. finds a $4.2 trillion shortfall by 2031 and says existing AI productivity gains cannot close it.
- AI-related debt is surging, with Goldman Sachs counting $489 billion in global supply this year.
- AI borrowing is competing with a federal deficit near $2 trillion, pushing long-term Treasury yields to a 25-year high.
- The hyperscalers can absorb the shock. Neoclouds, lenders and late-arriving retail investors may not.
- A dissenting public, struggling to retain their meaning, purpose, and traditional roles in society, are an obstacle to the dehumanizing aspects of AI.
The parallels to 2000 are uncomfortable. On Jan. 1 of that year, the world exhaled when the Y2K scare fizzled, and investors promptly chased dot-com stocks that defied common sense and, mostly, never turned a profit. Then came March. The run-up was exposed as one of history’s great investment bubbles, a generation of investors was burned, and the tech market needed years to recover. Even Microsoft was swept up in the mood shift. Columnist Charlie Gasparino argues the likeliest trigger was the instincts of hedge funds and institutions that do their own homework and began taking profits. Today’s pitch, groundbreaking technology that will enrich everyday investors and not only oligarchs, is nearly identical. A recent spike in rates after a bad inflation report, followed by an AI selloff, may be an early tremor.
The arithmetic nobody can spin
On Oct. 14, research firm Gartner put numbers to the unease. Owners of AI data centers would need $8.5 trillion in revenue through 2030 to cover announced spending, financing costs and expected returns, yet the enterprise and consumer AI revenue attributable to those facilities is forecast at no more than $3.6 trillion. Gartner analyst Will Sommer said that revenue will be “insufficient to support profitable returns” on planned spending. Subtract one figure from the other and a $4.9 trillion hole remains. Neocloud-related stocks fell sharply that same day.
Bain and Co. reached a strikingly similar verdict. It projects hyperscaler capital spending could reach $780 billion in 2026, a fivefold increase in three years, and annual AI infrastructure spending could hit $1.5 trillion by 2031. Paying for that would require an AI market worth $6 trillion a year, while today’s consumer and enterprise markets might reach $1.8 trillion, leaving $4.2 trillion unaccounted for. Bain’s conclusion is telling. Productivity gains from existing AI services are not enough, so the industry must conjure brand new markets, from autonomous vehicles to robotics. In plain English, the returns promised to investors depend on products that do not yet exist.
To be fair, this is not 2000 in every detail. Today’s hyperscalers sit on cash-generating core businesses in search, cloud and e-commerce that can service debt even if AI returns arrive late, whereas the dot-com era’s telecom builders leaned on speculative, junk-rated leverage. That cushion protects the giants. It does not protect the neocloud upstarts, the lenders behind them, or the retail investors who arrive last at every party.
Debt, deficits and the human variable
The financing is where the story turns from a tech problem into a national one. Goldman Sachs pegged global AI-related debt supply at $489 billion so far this year, already ahead of its full 2025 estimate of $322 billion, and Barclays data show AI debt has climbed to roughly 15% of investment-grade issuance. Alphabet even sold a rare century bond. All of it lands on the long end of a bond market that must also absorb a federal deficit running near $2 trillion a year, with roughly $1 trillion in interest costs. Treasury’s long-term borrowing costs have reached a 25-year high, and Bank of America economists say AI borrowing is “potentially crowding out long-end Treasury demand.” Other forces, including the Iran war and stubborn inflation, share the blame, and no one can pin down AI’s exact contribution. But the feedback loop is plain: higher yields raise the government’s interest bill, widening deficits and demanding still more borrowing.
That leaves the question no spreadsheet captures. AI can streamline tasks, and no honest critic denies it. But the investment case is not about streamlining tasks. It is about replacing the paychecks, professions and purpose of the very people who must buy, tolerate and vote on the arrangement. When a technology’s profits depend on hollowing out meaning and careers, resistance becomes a line item. Customers can refuse the chatbot, workers can refuse the pilot program, and voters can refuse to subsidize the buildout through higher borrowing costs. “Superintelligence” works best as a story told to keep capital flowing while cash flow lags.
History offers an old piece of advice: take your profits and do not get greedy. The pitch deck says the machines will make everyone rich. The math says someone is about to pay the bill for a bunch of hype, and it rarely lands on those who wrote the pitch deck.
Sources include:
Please contact us for more information.