In 2025 the S&P 500 ended the year up approximately 17%, making it the third straight year of double-digit gains.1 This bull market has been powered by the stocks closely associated with artificial intelligence theme. Yet beneath the surface, this rally is anything but broad-based.
AI hype has funneled capital into a narrow cohort of beneficiaries, leaving the majority of the sectors behind. The so-called Magnificent 7 stocks (Alphabet, Amazon, Apple, Meta Platforms, Microsoft. Nvidia, and Tesla) contributed approximately 45% of S&P 500’s total return. In short, the rally's "strength" is propped up by mega-caps; peel back the layers and find that the median stock is up close to 6%.2
Defenders of this concentration argue that it is earned given the Magnificent 7’s strong balance sheets, higher profitability, and earnings. Skeptics, on the other hand, argue that their high PE embeds optimistic assumptions about AI’s payoff, and any shortfall in earnings could trigger a market correction given how popular this theme is.
Is AI Making a Workplace Impact?
The initial hype around AI is maturing into a period of profound pragmatism. The fundamental question has shifted: we are no longer just marveling at its capabilities but asking what its earning potential is. This transition is a healthy one, moving us away from the hope-and-hype cycles of the past and toward a market defined by real-world utility.
According to a recent paper, close to 36% of US workers are using Generative AI tools to some capacity in their occupation.3 However, the use and application of ChatGPT and other Generative AI tools in workplaces varies significantly by industry and demographic. A Gallup poll from June showed that while around 40% of US workers claim to use AI at least a few times a year, only 8% of those surveyed were using it daily.4
There hasn’t been much evidence yet that AI is resulting in large productivity gains, as the data is hard to quantify. One report from Goldman Sachs indicated that AI boosts labor productivity by 27-31% on average, but we wouldn’t put too much stock into this number as the sample size was small and included company anecdotes, which are likely upwardly biased.5
Although limited, the data looking at the impact of AI on youth employment clearly suggests a disproportionate effect. Recent analysis shows that hiring for entry-level roles (ages 22-25) for AI-exposed roles has declined 16% compared to roles that are less exposed to AI.6 This effect appears to be isolated, as the study showed that older workers were insignificantly impacted.7
In sum, while generative AI has achieved meaningful penetration across the U.S. workforce, its adoption remains uneven and its productivity benefits difficult to measure at scale. As AI capabilities continue to evolve, the data suggests that its influence on work will be gradual, underscoring the need for continued monitoring, adaptation, and targeted workforce development.
AI and Equities: A Bubble or Not?
The central question for 2026 is whether the massive capital investments made over the last several years can finally translate into sustained corporate earnings. Investors argue whether this is the beginning of a multi-decade productivity revolution or is it a high-stakes financial bubble nearing its peak. Let’s take a look at the core arguments from market participants on both sides of the AI bubble debate.
Why AI is a bubble?
The Capex vs. Revenue gap: There is a widening disconnect between the large amount that is being spent on chips and data centers and the revenue being generated by AI software. Critics worry that ROI will never materialize to justify the investment.
Circular Financing: There is a growing concern regarding circular deals within the AI industry, where the major players invest in their own customers and suppliers. Skeptics warn that this could create an artificial house of cards where reported revenues are inflated by companies essentially passing the same capital back and forth rather than generating it from outside demand.
Scaling Laws: There has been a suggestion that the idea that returns for more data and computing are diminishing. If models stop getting significantly smarter, the case for greater capital spend begins to fall apart.
Leverage: Most of the market value is concentrated in a few companies that have strong balance sheets. But there are companies that are not taking on significant debt to buy GPUs, creating pockets of risk if the hype cools.
Why AI is not a bubble?
Financial Strength: AI leaders are high margin companies that generate large amounts of free cash flow and are making investments with existing profits rather than debt, making them more resilient to market shocks.
Real World Utility: There is measurable efficiency in software development, legal document review, and high-volume customer service. Many AI companies are generating revenues from enterprise-grade customers.
Reasonable Valuations: Valuations are high but nowhere near the extremes of 1999 and much of the growth has been supported by earnings growth.
Where do we stand in this debate?
While we think that the AI sector is not a bubble yet, we acknowledge that there are signs of froth in certain pockets. We like the theme and believe in its potential to enhance productivity but recommend diversifying across the various layers of the AI stack (infrastructure, platforms, and applications).
