Technology stocks have come under heavy pressure in recent weeks as investors reassess the sustainability of the ongoing AI boom, particularly the massive capital spending and the short lifespan of the hardware powering it. Despite strong earnings from major players like Nvidia, markets remain jittery. The broader Nasdaq 100 has slipped 6.3% in the past few weeks, while the Technology Select Sector SPDR Fund has fallen more than 9%, reflecting a deeper shift in sentiment. Even Nvidia, which reported solid third-quarter results on November 20, saw its shares close down 1% on Friday, despite being up 29% year-to-date. This divergence between strong fundamentals and falling stock prices signals a growing belief that AI-driven spending may be approaching a breaking point.
A central worry fueling the selloff is depreciation — the pace at which GPUs and AI servers lose their value as newer and more powerful chips arrive. Few observers have been as vocal as famed short-seller Michael Burry, who argues that major tech companies are dramatically underestimating the speed of technological obsolescence. In a widely discussed post on November 9, he warned that hyperscale cloud providers could understate depreciation expenses by as much as $176 billion between 2026 and 2028. According to Burry, companies currently assume a five- to six-year useful life for servers and GPUs, while the real-world lifespan is closer to two to three years. If he is right, the industry’s cost structure could be far more burdensome than investors realize, potentially squeezing profits even at the largest firms.
Adding to these worries, Kai Wu of Sparkline Capital estimates that annual depreciation could jump from around $150 billion today to $400 billion within five years as companies accelerate GPU replacements to keep pace with competitive and performance demands. Wu argues that once depreciation is properly accounted for, today’s AI infrastructure expansion already rivals — and in some cases surpasses — the scale of the dot-com and telecom buildouts. Adjusted for today’s faster innovation cycle, he says AI spending relative to GDP exceeds previous peaks in major tech investment eras. This raises the specter of overbuilding, an oversupply of computing capacity, and a painful period of returns lagging far behind expectations.
The financial strain of this rapid buildout is becoming more apparent across the sector. U.S. companies have issued more than $200 billion in investment-grade bonds this year specifically to finance AI-related infrastructure — roughly 13% of all investment-grade issuance through October. This represents a dramatic increase in debt financing compared to historical norms. From September to October alone, tech giants raised $75 billion in new debt, more than double the sector’s typical annual levels over the past decade. The five largest hyperscalers — Amazon, Microsoft, Meta, Alphabet, and Oracle — have collectively issued $121 billion in debt year-to-date, compared to an average of just $28 billion annually in the last five years. Analysts at J.P. Morgan project that these firms may need as much as $1.5 trillion in new bonds over the next five years, with about $300 billion expected in 2026 alone. This extraordinary debt expansion reflects both the urgency and costliness of the global race to dominate AI infrastructure.
But not all companies are positioned equally in this increasingly capital-intensive environment. Oracle stands out as the most financially vulnerable among the major hyperscalers. Its free cash flow over the past 12 months has fallen to negative $5.9 billion — its lowest level in at least 23 years. Investors have taken notice: Oracle’s credit spreads have widened by 48 basis points since September, signaling rising perceived risk. On November 6, S&P Global Ratings downgraded Oracle’s outlook to negative, warning that if its current capital spending pace continues, the company’s cash reserves could be depleted by November 2026. This is a stark contrast to the healthier balance sheets of its peers.
By comparison, Amazon, Microsoft, Meta, and Alphabet remain in far stronger financial shape. Their credit ratings fall between AA- and AAA, reflecting confidence in their stability and long-term earnings potential. Their trailing twelve-month cash flows — ranging from tens of billions to more than $300 billion — give them significantly more flexibility to finance aggressive capital spending without jeopardizing financial stability. Analysts describe these companies as the “Mount Rushmore” of corporate credit: massive, durable, and unlikely to face short-term liquidity strains, even amid sharp increases in spending. Bloomberg Intelligence credit analyst Robert Schiffman noted that the combined capital expenditures of these companies relative to their free cash flow is still far below the levels seen during the late 1990s tech bubble, reinforcing the argument that while risks exist, the situation is not a carbon copy of past market excesses.
Despite their strength, even the biggest players are not immune to scrutiny. Investors are increasingly questioning whether the industry’s infrastructure buildout — from GPUs to data centers to advanced networking — can maintain its current pace without eventually hitting a wall. The worry is not only that the hardware becomes obsolete quickly but that spending outpaces revenue growth. Capital spending has surged across the tech sector, but it may take years for AI-related services to generate enough profits to justify the unprecedented investment. Many AI applications are still in early commercial stages, and monetization remains uncertain.
At the heart of the anxiety is a broader debate about the durability of the AI investment thesis. Proponents argue that AI will transform every major industry, fueling productivity gains and new revenue streams, and that early spending will pay off enormously over time. Critics counter that the sector is repeating historical mistakes: pouring capital into infrastructure faster than demand can sustainably absorb. They fear that companies are building GPU clusters and data centers at a pace driven more by competitive pressure than by measured economic returns.
Still, it is important to note that today’s environment differs in significant ways from previous bubbles like the dot-com crash. Many AI leaders have enormous cash reserves, highly diversified revenue streams, and profitable business models, making them better equipped to absorb rising depreciation and debt costs. Strong underlying demand for cloud computing, AI tools, and digital services also supports the argument that long-term fundamentals remain intact. The question for markets is not whether AI has a future — few doubt that — but whether the current spending and growth expectations are priced correctly.
Ultimately, the recent selloff reflects a recalibration rather than a collapse. Investors are beginning to account for the realities of depreciation, leverage, and potentially slower monetization cycles. The AI industry still has significant promise, but the path forward may require more disciplined spending and a clearer match between investment and long-term returns. As companies navigate the tension between innovation and sustainability, markets will continue to weigh the risks of an overheated AI arms race against the transformative potential of the technology itself.
