Igor Pejic, a Reuters analyst covering AI-driven finance, has warned that a correction in artificial intelligence stocks may not reach the severity of the dot-com collapse but could still leave investors with few places to hide. His argument rests on a simple observation: the capital pouring into AI infrastructure has become so concentrated among a handful of firms, and so intertwined with debt markets, that a slowdown would ripple far beyond Silicon Valley. The question is not whether AI spending will cool, but how much damage the cooling inflicts on sectors that appear unrelated.
Spending Commitments Buried in Filings
The clearest evidence of how deeply major technology firms have committed to AI sits in their regulatory filings. Annual 10-K reports, quarterly 10-Q updates, and event-driven 8-K disclosures filed through the EDGAR database reveal the scale of capital expenditure pledges by the so-called Magnificent Seven and their key supply-chain partners. These documents go beyond press releases and earnings-call talking points. They contain audited financials, formally disclosed risk factors, and liquidity analyses that bind companies to specific commitments.
What stands out in recent filings is the risk-factor language. Hyperscalers have begun flagging their dependency on sustained AI demand as a material uncertainty, a legal acknowledgment that the revenue these investments are meant to generate has not yet fully materialized. That gap between committed spending and proven returns is where Pejic’s thesis gains traction. When companies disclose that their business outlook depends on demand that is still forming, investors reading those filings through the EDGAR login portal can see the tension between ambition and evidence.
These filings also show how commitments extend over multiple years. Long-term purchase agreements for chips, networking hardware, and cloud capacity appear as contractual obligations stretching well beyond the current business cycle. Even if management wanted to slow spending abruptly, the legal and financial penalties for breaking these agreements would be steep. That rigidity raises the stakes of any slowdown in AI-related revenue, because companies cannot easily dial back the very investments they have told shareholders are essential to future growth.
Concentration Risk Beyond Tech Stocks
A correction in AI equities would not stay contained within the technology sector. The International Monetary Fund’s Global Financial Stability Report has framed the AI investment boom as a potential source of systemic risk, citing asset concentration and spillover channels that connect tech valuations to broader financial markets. The IMF analysis addresses how a sharp repricing in a narrow group of stocks can drag down index funds, pension allocations, and credit markets that have indirect exposure to the same names.
This is the mechanism that makes Pejic’s “few safe havens” framing credible. When a small cluster of companies accounts for an outsized share of market capitalization, passive investment vehicles that track major indices carry concentrated bets whether their holders realize it or not. A downturn in AI sentiment would hit not just chipmakers and cloud providers but also the energy firms building power capacity for data centers, the real estate investment trusts financing those facilities, and the lenders extending credit against projected AI revenues.
Because these flagship companies sit at the core of major benchmarks, their performance influences everything from target-date retirement funds to sovereign wealth portfolios. Even investors who believe they are broadly diversified may find that a single theme (AI infrastructure) dominates their returns. That hidden concentration is what turns a sector-specific correction into a market-wide event.
Debt Hotspots in the Data Center Boom
Pejic’s analysis, published through Reuters coverage, identified specific debt pressure points in the AI data center supply chain. The financing structures behind new data center construction often rely on projected lease revenue from hyperscalers, creating a chain of obligations that assumes AI workloads will keep growing at current rates. If that growth slows, or if a single major tenant scales back its commitments, the debt service burden falls on developers and their lenders rather than on the tech firms themselves.
This dynamic echoes patterns from earlier infrastructure booms. Telecom companies in the late 1990s built fiber-optic networks on the assumption that bandwidth demand would grow exponentially. The demand eventually arrived, but not before a wave of defaults wiped out investors who had financed the buildout. The AI parallel is not exact, because today’s hyperscalers have far stronger balance sheets than the telecom startups of that era. But the downstream financing, the special-purpose vehicles and project loans funding physical construction, carries similar structural fragility.
Another vulnerability lies in how these projects are packaged for investors. Securitized debt backed by long-term data center leases can look safe on paper, with investment-grade tenants and predictable cash flows. Yet those projections often assume continuous renewal at high utilization rates. If AI training intensity moderates or efficiency gains reduce compute needs, cash flows may undershoot expectations, putting stress on structures that were sold as low risk.
Why a Full Dot-Com Repeat Is Unlikely
Pejic has been careful to distinguish the current AI cycle from the dot-com era on several fronts. The companies driving AI spending are profitable, cash-rich, and already generating revenue from cloud services and enterprise software. Unlike the dot-com startups that burned through venture capital with no path to profitability, the Magnificent Seven can absorb writedowns without facing existential threats. Their obligations are documented through SEC filer management systems that support ongoing disclosure and oversight.
That financial resilience limits the depth of any correction but does not eliminate its breadth. A repricing of AI expectations could shave significant value from portfolios without triggering the cascading bankruptcies that defined the dot-com bust. The pain would be distributed differently: less concentrated in equity wipeouts, more spread across credit markets, supply chains, and the emerging market economies that host semiconductor manufacturing and rare-earth mineral extraction.
Moreover, AI adoption is already embedded in mission-critical workflows across industries, from software development to logistics and healthcare analytics. That installed base provides a floor under demand for compute and cloud services. The risk lies less in AI disappearing than in growth rates falling short of the heroic assumptions currently embedded in some valuations and capex plans.
What Retail Investors Should Watch
For individual investors trying to gauge their exposure, the first step is understanding how much of their portfolio tracks AI-heavy indices. The SEC’s investor education site offers guidance on reading corporate disclosures and assessing risk-factor language. Paying attention to the gap between capital expenditure commitments and realized revenue in quarterly filings provides a more grounded signal than analyst price targets or media sentiment.
The second step is recognizing that diversification within equities may not provide the protection it normally would. If AI names dominate major indices, then holding a mix of index funds still leaves an investor heavily weighted toward the same cluster of companies. True diversification in this environment means looking at asset classes and geographies that are genuinely uncorrelated with U.S. tech spending, a harder task than it sounds when global supply chains tie so many economies back to the same set of firms.
Investors who want to look under the hood of specific companies can use the SEC’s online forms portal to trace new registration statements, shelf offerings, and other capital-raising activities linked to AI initiatives. A surge in debt issuance or preferred stock tied to infrastructure buildouts may signal rising financial leverage just as enthusiasm peaks.
The Gap Between Innovation and Valuation
None of this means AI is a bubble in the traditional sense. The technology is real, the productivity gains are measurable in specific applications, and enterprise adoption continues to expand. Pejic’s argument is not that the underlying innovation is illusory, but that markets have a long history of overpaying for real breakthroughs. When expectations race ahead of what customers are willing to spend, even transformative technologies can deliver disappointing returns to shareholders.
The gap between innovation and valuation is widest where narratives dominate over cash flows. Companies that promise future AI monetization without clear pricing models or margins are most vulnerable if sentiment turns. By contrast, firms that can already show recurring revenue from AI-enhanced products are better positioned to weather a shift in market mood, even if their share prices prove sensitive in the short term.
Pejic’s warning ultimately centers on discipline. Investors, he suggests, should treat AI like any other capital-intensive technology cycle: by scrutinizing balance sheets, questioning extrapolated growth curves, and reading the fine print in regulatory filings rather than relying on hype. If the AI boom cools, those habits may not prevent losses, but they could determine who merely endures a correction and who discovers that there were fewer safe havens than the headlines implied.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
