It’s fascinating how bubbles have routinely confounded the finest minds since the dawn of modern financial markets. From the Dutch tulip mania of the 17th century, through the South Sea and Mississippi madness of the 18th, to the railway speculation of the 19th, then the Japanese asset bubble of the 20th, and the dot-com, subprime, and crypto frenzies of the new century — each time, without exception, the global financial elite of the day parroted the “this time is different” mantra, right up until it turned out it wasn’t.
They’ve mapped the anatomy of bubbles, labelled each stage – displacement, boom, euphoria, profit-taking, panic – and yet they still fail to recognise one when it’s happening in real time. Now, once again, talk of another bubble is all the rage on Wall Street, and hence the rest of global finance.
This time, it’s the AI boom, of course. And while the crowd is still divided between genuine productivity transformation and dangerous mispricing of future promises, the structure and tone of the debate itself echo a familiar past. Goldman Sachs, hardly prone to hyperbole, is already drawing comparisons with the late 1990s. They don’t call it a bubble just yet. But they’re not ruling one out either.
Their strategists have pointed out at least five risk signals from the dot-com era that are showing up again – rising investment to unsustainable levels, declining profitability despite soaring equity prices, aggressive corporate debt issuance, rate cuts that stoke risk-taking, and early signs of widening credit spreads.
Back then, non-residential tech and telecom investment peaked at 15 percent of US GDP before collapsing. Now, a fresh wave of capital expenditure is building, led by today’s tech heavyweights. Amazon, Meta, Microsoft, Alphabet, and Apple are collectively on track to spend US$349 billion in 2025. That kind of figure would have looked like fiction just two years ago.
Corporate profits, meanwhile, have yet to decline. In fact, S&P 500 net margins for Q3 are holding around 13.1 percent, well above the five-year average. But Goldman notes that in the late ‘90s, macro-level profitability began deteriorating years before the actual crash. Markets just didn’t care until they had to. Some would argue that’s precisely what a bubble does – it seduces the system into ignoring signals that would otherwise trigger caution.
Debt levels are another area drawing scrutiny. Goldman flags that, during the dot-com period, companies levered up heavily before profits turned. This time, firms appear more responsible on the surface, with most large tech capex still financed through cash flows. But scratch the surface, and a different picture emerges. Big Tech has raised US$93 billion through bond issuance this year alone — more than the previous three years combined. And some, like Meta, have resorted to creative off-balance-sheet financing. In late October, it struck a US$27 billion deal with Blue Owl Capital to fund a data centre in Louisiana without disturbing its A+ rating. On paper, it’s clever. Structurally, it pushes leverage out of sight all right, but does it also push it out of risk?
Oracle is more aggressive still. Its capex next year is expected to be 138 percent of operating cash flow. Meta, the runner-up, comes in at 84 percent. Oracle’s bond spreads have widened the most among its peers this year, reflecting growing discomfort among creditors. Meta’s own 25-year note for the Louisiana project pays 6.6 percent — a full percentage point above its comparable corporate bonds. Despite the credit rating, the pricing resembles a junk bond. That may be rational in a market where balance sheet transparency is eroding.
Still, this isn’t 1999. Not quite. Then, valuations were more extreme, retail participation was more frenzied, and many tech firms had no revenue at all. Today’s megacaps generate actual earnings. But their growth expectations — anchored in the pursuit of artificial general intelligence — are increasingly dependent on speculative timelines and unprecedented capital outlays. At some point, the bill will need paying.
Interestingly, and instructively for students of international markets, the US bond market, for now, has not lost its cool. But it is not asleep either. Credit spreads remain tight historically, but they are no longer compressing. The ICE BofA US High Yield Index spread climbed to 3.15 percent recently, up from 2.76 percent in late October. Subtle, but notable. The market is still pricing risk – but with enough liquidity and demand to keep the wheels turning.
This is also where the Fed re-enters the conversation. In the late 1990s, rate cuts helped extend the rally and paper over cracks. This year, after a long hiking cycle, the Fed cut rates by 25 basis points in October and is expected to cut again in December. Some, including Ray Dalio, have warned that further easing could pump the bubble even more, rather than containing it.
But even if this isn’t a classic bubble yet, the structure of current financing is risky enough. The global AI infrastructure buildout is projected to cost nearly US$3 trillion through 2028. Cash flow will cover half at best. That leaves the rest to markets. And with firms rushing to raise funds through both traditional bonds and private credit structures, pressure on credit markets is building right in front of everybody. And the risk now isn’t just market correction. It’s the potential spillover into the real economy if these investments underdeliver, or if tighter credit disrupts further capex cycles.
In recent weeks, the AI debate has quietly shifted. Less about the philosophical questions of super intelligence, more about the mechanics of paying for it. There’s even early talk of whether OpenAI may one day need public funds to stay solvent. It’s not mainstream thinking yet. But it’s already slipped into the discourse. And that, too, is a signal.
If the past teaches anything, it’s that bubbles don’t announce themselves, and that the smarter they are on Wall Street, the more blind they are to them when they’re inflating. They appear logical, even inevitable, until the moment they aren’t. Whether this is that moment remains unclear. But when corporate borrowing accelerates, off-balance-sheet debt becomes fashionable, and credit markets begin to distinguish between the headline and the footnotes, the system is no longer priced for perfection. It’s pricing for something else.
Whatever that turns out to be, it won’t stay hidden for long.
Maybe this time will be different after all.
Copyright Business Recorder, 2025
The writer can be reached at [email protected]
Comments
Comments are closed for this article.