For most of 2025, the argument on Wall Street was binary: either artificial intelligence was a genuine platform shift or it was a repeat of the dot-com bust. By late July 2026, that framing had already become outdated. In a widely discussed Bloomberg Opinion column published July 31 titled “A Tale of Two Bubbles Defining a Third,” columnist John Authers argued that investors are no longer simply debating whether AI resembles the 1999 internet mania — they are increasingly weighing parallels to the 2008 credit crisis instead, a far more unsettling comparison because it implies leverage and financing risk, not just overvaluation.
From Overvaluation to Over-Leverage
The distinction matters. A 1999-style bubble is primarily a story about stock prices detached from earnings — painful when it pops, but largely a public-markets problem. A 2008-style bubble is a story about debt, counterparty risk, and financing arrangements that can freeze credit markets when confidence breaks. Authers’s column points to the growing use of debt-financed data center construction, vendor financing between chipmakers and their customers, and special-purpose financing vehicles as evidence the AI boom has picked up more 2008-like characteristics over the past year.
A Volatile Month for Tech Stocks
The timing of the column was not incidental. Bloomberg reported that the so-called Magnificent Seven megacap technology stocks shed roughly $797 billion in market value in a single stretch in late July as AI skeptics sold down positions, part of what Bloomberg separately described as the biggest Wall Street sector rotation since 2020. That volatility followed a string of earnings reports in which investors punished companies for capital-expenditure increases that lacked matching revenue proof, even as they rewarded firms like Microsoft and Amazon whose cloud units showed clearer monetization.
The Bull Case Still Has Defenders
Not everyone accepts the darker framing. Bulls note, correctly, that the AI infrastructure buildout is backed by real, growing revenue in a way the fiber-optic overbuild of 1999-2001 never was. Companies selling AI infrastructure — Nvidia, Broadcom, and the hyperscalers’ own cloud divisions — have posted some of the strongest quarters in their history, with demand for computing capacity outstripping supply rather than the reverse. From that vantage point, the current unease looks more like a healthy repricing of a real growth story than the unwinding of a fantasy.
Where the Skeptics Focus Their Concern
Skeptics counter that revenue growth alone doesn’t resolve the financing question. The worry isn’t whether AI is useful — it’s whether the capital structure underpinning today’s data center construction can survive a slowdown in demand growth, given how much of it now depends on debt, guarantees, and vendor-financing loops between chipmakers and their largest customers. Bloomberg’s own newsletter coverage in July carried the blunt headline “What If the AI Bubble Pops?”, walking through scenarios in which a capex slowdown cascades through credit markets rather than staying contained to equity valuations.
Crowding as an Underappreciated Risk
A less discussed but structurally important point from Bloomberg’s reporting is that AI itself may be amplifying the fragility of AI-linked trades. Research cited in Bloomberg’s coverage suggests that AI-driven trading systems make markets faster and more informed on average, but also more crowded and harder to de-risk quickly when sentiment turns — a dynamic that showed up in the rapid, synchronized selloff across megacap names in late July. That crowding dynamic helps explain why the drawdown moved so quickly beyond an isolated stock or two: because so many large funds hold overlapping positions in the same handful of AI-linked names, any shift in consensus sentiment tends to hit the entire cohort at once rather than unfolding gradually.
What Comes Next
None of this resolves the underlying debate, and that is arguably the point: serious, credentialed voices on both sides are now arguing from different first principles rather than different facts. The next data points investors are watching closely include Nvidia’s fiscal second-quarter earnings, expected August 27, and further capex disclosures from Amazon, Microsoft, Meta, and Alphabet, all of which raised 2026 spending plans in their most recent quarters. If those companies can show capacity constraints easing alongside continued revenue growth, the 2008 comparison may fade. If credit markets show more stress in AI-linked debt and financing vehicles, the darker analogy could gain currency fast — and it’s the latter scenario Wall Street strategists say would be far harder to unwind than a simple stock correction. For now, both camps agree on one thing: the answer will be determined by hard financial disclosures over the next several quarters, not by rhetoric, which is why earnings season has become the most closely watched recurring event on the market calendar.
Photo: MDGovpics / BY via flickr