Artificial Intelligence and Macroeconomics: Dangerous Liaisons

Records are made to be broken. Nvidia seems to have taken this saying to heart, judging by the evolution of its stock price. Its surge has pushed the company past the $5 trillion market-cap threshold—roughly equivalent to 20% of U.S. GDP—just a few months after having crossed the $4 trillion mark. OpenAI, another flagship AI company, is considering an IPO to raise $60 billion at an implied $1 trillion valuation. Not bad for a company virtually unknown to the general public before the launch of ChatGPT three years ago. One can easily imagine the frenzy among bankers preparing what could be the biggest deal ever and propelling the company straight into the top 10 U.S. market caps, shoulder to shoulder with Berkshire Hathaway!

While waiting for this much-anticipated IPO, announcements in the sector are almost daily—so frequent that even specialized analysts struggle to keep their models up to date. The result is that the AI ecosystem now resembles a tangle of partnerships, contracts, and above all, agreements (which are legally less binding). What should we make of all this? Above all, the dizzying surge in investment projections.

The hyperscalers* took advantage of the recent earnings season to outdo one another with CapEx announcements. Goldman Sachs analysts compiled the numbers: spending is expected to rise from $400 billion this year to $533 billion in 2026. The phenomenon is so powerful that it now shows up in U.S. macroeconomic statistics: business investment is being driven primarily by AI-related equipment, growing at an annualized pace of 12% in the first half of the year (four times faster than in 2024), while the rest of the economy—residential real estate in particular—lags behind. On an even broader scale, the World Trade Organization reports that semiconductors, servers, and telecom equipment accounted for nearly half of the increase in global goods trade in the first half.

A new phenomenon has emerged: investment levels are now so high that hyperscalers’ own cash flows are sometimes insufficient to fund them. Like any other industry, they are therefore naturally turning to capital markets to raise debt. Meta has been the boldest, issuing a jumbo $30 billion bond (four times oversubscribed), and total industry issuance in 2025 is nearing $120 billion—around 15% of U.S. investment-grade corporate bond issuance (excluding financials). Even though the market absorbed this wave of paper, it triggered early warning signs among credit analysts: Oracle—whose financial situation** is atypical within the sector—has seen its CDS*** spike sharply over the past few weeks, and more broadly, credit spreads for this group of five companies have widened to 80bp (+20bp).

Beyond the short-term positive impact of this investment wave on U.S. growth, the medium-term holy grail remains a meaningful increase in productivity and therefore potential growth. While it is widely assumed that such gains will materialize—albeit on an uncertain timeline—a recent study from the Massachusetts Institute of Technology cast doubt: among 300 corporate pilot projects examined, only 5% show an acceptable return on investment. Once again, the latest earnings season offered insight into how industry leaders are tackling the issue. In this regard, an example shared by Patrick Pouyanné, CEO of TotalEnergies, is instructive: he believes AI models will optimize the utilization rates of production assets, leading to increased output and, consequently, stronger cash-flow generation.

But the ongoing technological revolution poses a threat to another major component of the economy: employment. Indeed, the avalanche of announcements is not limited to sales—management teams are also acting swiftly on “human resources.” Two examples among many: Amazon plans to cut 14,000 support-function roles, while Accenture intends to do the same with consultants deemed not “AI-compatible.” The phenomenon is not confined to the tech sector: Walmart’s CEO recently stated that the retail giant expects to expand over the next three years without any increase in headcount.

While most of the macroeconomic effects of the AI wave lie ahead, the impact on equity markets is a very current issue. Some commentators even draw parallels with the Dotcom bubble, focusing on market concentration and valuations. The Fed Chair himself commented that equity markets appear “fairly highly valued.” On the opposite end, the sector’s permabulls argue that circular cash-flow debates and uncertainties around investment returns justify the low risk of a bubble forming.

Beyond historical comparisons and rhetorical arguments, two factual observations stand out. The equal-weighted U.S. stock index (SPW) currently trades at a significant discount (25%) to the S&P 500, whose valuation is skewed by a limited number of high-multiple stocks****. And the concentration of the index has led to historically unusual market behavior: it has fallen even on days when a majority of stocks were rising—twenty times this year. In such an unusual setup, investors already thinking about next year’s asset allocation will likely need to adopt a contrarian mindset, just as the “momentum” factor shows signs of fatigue…

Amazon, Google, Meta, Microsoft, and Oracle
** The company has a net-debt-to-EBITDA ratio of around 2.5x
*** Since late September, its credit default swap—a derivative that measures the cost of insuring against default—has doubled to exceed 100bp
**** Price/earnings ratio of 17x vs. 23x

Pierre Pincemaille, General Secretary of Investment at DNCA. This article was finalised in November 28th, 2025.

This promotional document is a simplified presentation and does not constitute a subscription offer or an investment recommendation. No part of this document may be reproduced, published or distributed without prior approval from the investment management company.

DNCA Investments is a trademark held by DNCA Finance