The AI boom, earnings fragility, and last week’s volatility
Corrado Tiralongo - Nov 12, 2025
Executive summary
- Nvidia’s strong earnings reaffirmed the resilience of AI-related demand, yet markets reversed sharply the following day, highlighting growing sensitivity to interest-rate expectations and the concentration of market leadership.
- Last week’s volatility reflected concerns that the valuation expansion in large-cap technology is becoming increasingly dependent on lower real yields, even as policy expectations shift toward a slower and shallower easing cycle.
- The financing behind the AI buildout is evolving, with companies relying more on private credit, structured vehicles, and vendor-financing arrangements as capital requirements scale into the trillions. This shift introduces new sources of fragility not present during the early, cash-flow-funded phase of the boom.
- Today’s cycle is less comparable to the dot-com era and more consistent with earlier periods of overbuilding in transformative technologies such as railways in the 1800s and the electrification and automobile boom of the late 1920s. These were real innovations with lasting economic benefits, but the speed and method of financing created vulnerabilities.
- Valuations in AI-exposed sectors remain elevated but more reasonable when adjusted for real yields. The sustainability of current multiples depends on the breadth of earnings delivery, the trajectory of AI capital expenditure, and the ability of firms to convert investment into durable profitability.
- From a portfolio perspective, a focus on the quality and durability of earnings, capital discipline, and diversified sources of return remains important as financing dynamics evolve, and the distribution of outcomes widens.
Introduction
Over the past year, artificial intelligence has shaped market leadership, sector performance, and investor psychology in ways few other themes have managed. The exceptional earnings power of a handful of technology firms, combined with the promise of broad productivity gains, has propelled a powerful market narrative. But recent volatility and a surge of client questions remind us that the character of the AI boom is not static. It is evolving.
A recent client note crystallized the debate: Are we still in the early stages of a transformative cycle, or have financing practices and market behaviour begun to resemble something more fragile? That question arrives at an important moment. Nvidia’s strong earnings report last week, the market’s unusual reversal the following day, and the broader discussion sparked by a recent Canadian podcast all point to meaningful shifts in how the AI cycle is being funded, interpreted, and priced.
On November 17, 2025, CBC’s Front Burner podcast aired an episode titled “If AI is a bubble, what happens when it pops?” featuring Paul Kedrosky, a partner at the venture capital firm SK Ventures and a research fellow at MIT’s Initiative on the Digital Economy. Kedrosky has followed technology booms and busts over several decades, and his remarks on the structure of AI-related financing provide a useful lens for thinking about where we are in the current cycle. This commentary integrates those developments, alongside historical context and our research, to assess where we are and what investors should be attentive to.
Nvidia’s earnings and the state of the fundamentals
The initial reaction to Nvidia’s latest results underscored how central the company has become to the AI narrative. Its report was stronger than expected, reaffirming a message consistent across hyperscalers: demand for AI infrastructure remains extremely strong, and visibility into medium-term spending remains supportive. Companies closest to the buildout continue to express confidence in multi-year demand for chips and data-centre investment, with little indication of a slowdown in earnings momentum or capital expenditure plans at this stage. These data points support the view that the AI rally remains grounded in fundamentals rather than purely sentiment. Strong earnings continue to underpin valuations, and while concerns about end demand remain valid, the companies best positioned to assess long-term trends continue to communicate confidence.
Why markets sold off despite strong results
The striking feature of last week was not Nvidia’s beat, but the market’s reaction the following day. The S&P 500 opened strongly, reflecting enthusiasm for Nvidia’s performance, before closing sharply lower. This type of reversal is unusual in the absence of negative news flow. The selloff was concentrated in the information technology sector and was even more pronounced in equal-weight indices than market-cap weighted ones. This pattern is consistent with rising concern about the sustainability of the AI-led rally, particularly given the extent to which a small cohort of companies has driven market gains since 2022.
The episode highlighted two dynamics now shaping the cycle:
- Valuations have become more sensitive to changes in interest-rate expectations.
Much of the valuation expansion over the past year reflected lower real yields. The rebound in yields, combined with more hawkish Federal Reserve commentary, helped destabilize sentiment. - Investors are increasingly questioning whether the pace of AI capex is sustainable.
With trillions of dollars in projected spending, the conversation has shifted from whether AI will deliver long-term productivity gains to whether the financing model can support the scale and speed of the buildout.
These concerns contributed to a broader reassessment of where we are in the cycle, particularly given the intense concentration of earnings and price momentum in a handful of firms.
Valuations, earnings growth, and where fragility sits
Despite last week’s volatility, our recent analysis suggests valuations may not be as stretched as price action implies. When adjusting for lower real yields, valuations in the major tech sectors look less extreme than headline multiples suggest. Excess earnings yield metrics remain above levels reached during the dot-com period, and expectations for long-term earnings have risen meaningfully.
The stronger earnings trajectory across the sector since 2022 support higher valuations. While this does not eliminate risk, it reinforces the idea that the AI leadership is not built solely on speculative momentum.
At the same time, valuations are increasingly exposed to policy expectations. Markets may be overestimating the likelihood or pace of rate cuts next year, and any recalibration in the path of real yields could create further volatility in valuation-sensitive sectors.
Are we nearing the turning point of an AI bubble?
This question has been gaining traction, particularly after Kedrosky’s remarks on Front Burner. His argument is not that AI lacks utility. It is that the financing behind the buildout has evolved in a way that creates systemic vulnerability.
In the early phases, the boom was largely funded from free cash flow generated by the sector’s largest players. That kept the risks contained. But as the capital requirements have surged into the trillions, companies have increasingly turned to:
- private credit,
- special-purpose vehicles,
- lease-back arrangements,
- vendor-financing structures that resemble past periods of speculative overbuilding.
Kedrosky’s concern is that these structures shift risk away from the balance sheets of the firms best able to bear it and into a wider ecosystem of lenders and investors, often through opaque vehicles. That echoes elements of prior cycles in telecoms, housing finance, and other capital-intensive booms. It also aligns with our research suggesting that, if and when an AI-driven correction occurs, the path will be shaped at least as much by financing conditions as by technology fundamentals.
The risks today arise not from the technology itself, but from the intersection of capital intensity, financing complexity, and narrow earnings leadership.
Why the dot-com analogy falls short
Although comparisons to the dot-com era are common, the analogy is misleading. The dot-com cycle was driven by high expectations and low earnings. Today’s environment is characterised by high expectations and exceptionally high earnings. This distinction matters. It is why I prefer to frame the current period as an earnings bubble, rather than a purely speculative one.
Historically, such cycles resemble episodes like the railway mania of the 19th century or the electrification and automobile boom of the late 1920s. These were grounded in real technologies with lasting economic benefits, yet they were accompanied by periods of overbuilding, excessive leverage, and ultimately painful corrections when the financial underpinnings failed to keep pace with technological optimism.
The lesson from those episodes is that transformative technologies can coexist with poor capital allocation. The long-term winners ultimately prosper, but the path can be uneven, especially when financing becomes the weak link.
Portfolio considerations
From a portfolio perspective, the recent volatility and the evolving financing backdrop do not undermine the long-term potential of AI. Instead, they highlight the importance of focusing on:
- the quality and sustainability of earnings,
- capital discipline among companies participating in the AI ecosystem,
- valuations in the context of real yields, and
- diversification across mandates that may respond differently as the cycle matures.
My preference is to frame these as observations rather than recommendations. Investors may want to ensure that their exposure to AI is supported by mandates that emphasize capital discipline and free-cash-flow sustainability. They may also want to identify managers who take a balanced view of growth opportunities and financing risks, and to retain exposure to strategies that offer diversification in an environment where concentration, liquidity, and financing dynamics could become more important drivers of returns.
Conclusion
The AI boom remains one of the most significant drivers of market performance and economic optimism. Nvidia’s earnings reaffirm the strength of the underlying fundamentals, while last week’s volatility, and the concerns raised by observers like Paul Kedrosky, remind us that the financing model, not the technology, may ultimately determine the path forward. We are not forecasting an imminent end to the cycle. But we are attentive to the way financing structures are evolving, the sensitivity of valuations to real yields, and the growing divergence between a narrow group of earnings leaders and the rest of the market. Recognizing where we are along this continuum helps manage expectations and maintain a disciplined, forward looking approach. As developments continue to unfold, we will provide updates on how the balance of risks is shifting and what it may mean for investors.
Corrado Tiralongo (he/him)
Vice President, Asset Allocation & Chief Investment Officer
Canada Life Investment Management