By Madison Mills | Axios | October 3, 2025

Tech’s latest gold rush is being bankrolled by record levels of debt, private lending, and elaborate off-balance-sheet financial vehicles. As Meta, Oracle, Nvidia, and other tech leaders pump billions into AI infrastructure and data centers, investors and market analysts are raising red flags over the sustainability of the boom—and the risk of a devastating bust reminiscent of previous market bubbles.

The New Face—or Mask—of Big Tech Debt

Debt has long been an early warning sign for financial crises. Economists point to excessive mortgage debt preceding the 2008 financial crisis, and runaway corporate leverage during the dotcom bubble. In 2025, the surge in borrowing by AI-focused tech giants is sparking fresh worries. Their strategy isn’t always visible to investors: much of this borrowing unfolds via private lenders, special purpose vehicles (SPVs), and off-balance-sheet financing arrangements, obscuring the true scale of leverage.

“That debt—how it’s structured and hidden—is almost an acknowledgement that this is getting out of hand,” said Dario Perkins, managing director of global macro at TS Lombard.

For instance, Meta is reportedly seeking $29 billion from private capital partners to fuel its AI data center ambitions—a staggering sum that, due to the use of SPVs, wouldn’t appear as direct debt on Meta’s balance sheet. Perkins warns these tactics echo the subprime mortgage era, when risk was shifted off the books to give investors a false sense of security.

Oracle, meanwhile, is tapping more traditional public markets, issuing $18 billion in bonds—one of the largest tech bond sales ever—to finance its own AI and cloud infrastructure buildout. This flood of cheap capital is being poured into data centers, AI chips, and next-generation computing power at unprecedented speed.

Exuberance Outpaces Economics

Despite the massive investments, some analysts argue there’s little evidence the current pace of AI-related spending will offer commensurate returns anytime soon. In a sign of the mounting risk, Perkins notes that some major tech players no longer even require a clear return on their AI capital expenditures, intent instead on outspending rivals in hopes of dominating tomorrow’s market.

“They say they don’t care whether the investment returns anything because they’re in a race,” Perkins said. “Surely that in itself is a red flag.”

Paul Kedrosky, a venture investor and technology analyst, echoed these concerns on the Plain English podcast. “There’s no way these companies can recoup the value of this capital spending,” he warned, pointing to the circular nature of some investments—such as Nvidia’s $100 billion stake in OpenAI. In such cases, large tech firms directly or indirectly fund their biggest customers, fueling a boom in their own hardware sales, but raising questions about the sustainability and transparency of the financial loop.

Boardroom Warnings: Bezos and Solomon Join the Chorus

It’s not just analysts sounding alarms. Amazon founder Jeff Bezos and Goldman Sachs CEO David Solomon are among the corporate chieftains cautioning against ‘industrial bubble’ dynamics in AI. Both have recently cited overheated investment, speculative fervor, and soaring tech stock indexes—the Dow up 10%, S&P 500 up 14%, Nasdaq up 18% year-to-date—as signals the market may be running well ahead of fundamental value.

Concerns run particularly high regarding ‘hidden debt’—the kind that is shuffled into SPVs or other entities to keep it off a company’s main financial statements. Investors worry that this growing pile of obligations could surface suddenly if market sentiment shifts, pushing stock prices downward and hampering companies’ ability to raise further funds.

A Trillion-Dollar Data Center Bet

The race for AI supremacy has spurred a historic expansion in data center investments. Nvidia, OpenAI, Google, and Amazon are leading a trillion-dollar building spree in North America and Europe, fueling exponential demand for high-performance chips, vast facilities, and climate-controlled server farms. Goldman Sachs now predicts global AI data center spending could surpass $1.3 trillion by 2027, with growth accelerating despite energy and supply chain concerns.

These numbers are staggering—and break the mold for capital investments in recent memory. However, it’s not clear whether the high costs will translate to durable profits or sustainable returns. According to market research firm Gartner, less than 20% of current AI data center capacity globally is being used for economically viable applications, increasing fears of a bubble-and-bust cycle if revenue fails to materialize.

Can the Party Last? Market Rationality Put to the Test

Many strategists point to a famous adage attributed to economist John Maynard Keynes: “The market can stay irrational longer than you can stay solvent.” Even with clear warning signs, history shows bull markets driven by technological innovation can persist for years before fundamentals catch up—or the music stops. TS Lombard’s Perkins, however, draws a sharp distinction between bubbles of the past and today’s AI frenzy. “We’re much closer to 2000 than 1995,” he noted, referencing the infamous dotcom crash.

He also points out several late-stage bubble characteristics emerging now:

• Borrowing funneled through opaque structures
• High levels of insider selling
• Investment ‘recycling,’ where money keeps circulating between the same companies
• And aggressive earnings growth that increasingly relies on unproven or hypothetical AI returns

Nevertheless, Perkins and other macro experts believe the core U.S. economy remains less entwined with AI capital spending than some bulls suggest, providing a cushion for diversified investors outside the tech sector should the bubble burst. “Investors in broader market indices, with exposure beyond AI, could weather an AI contraction better than most,” he said.

The Bottom Line: Caution Lights Flashing on Wall Street

As Silicon Valley’s race to own the future of AI reaches breakneck speed—and billions of funding dollars flow in through ever-murkier channels—the warning signs are becoming impossible to ignore. If even the world’s most profitable tech companies are finding it necessary to obscure giant debts to fund their AI bets, it signals a lack of confidence that near-term returns will justify today’s unprecedented spending spree. For investors and market watchers, the message is clear: While potential remains high, so too does the risk, and clear-eyed due diligence is more important than ever as the cycle matures.