- AI-related stocks have accounted for 75% of S&P 500 returns, 80% of earnings growth, and 90% of capital spending growth since ChatGPT launched in November 2022, according to JP Morgan Asset Management data.
- Goldman Sachs CEO David Solomon, Jeff Bezos, and OpenAI CEO Sam Altman all publicly warned of overinvestment on the same day, with Altman stating “people will overinvest and lose money.”
- MIT research found that 95% of 52 organizations achieved zero return on investment despite spending $30-40 billion collectively across 300+ generative AI initiatives.
- At Yale’s CEO Summit, 40% of 150+ executives surveyed said AI hype had driven overinvestment and anticipated a market correction.
What Happened
Jeffrey Sonnenfeld, Senior Associate Dean for Leadership Studies at Yale School of Management, and Stephen Henriques, Senior Fellow at the Chief Executive Leadership Institute, published an analysis in Yale Insights (originally in Fortune) outlining three scenarios for how the AI investment bubble could burst. The October 2025 paper drew on JP Morgan Asset Management data showing that AI-related stocks have driven 75% of S&P 500 returns, 80% of earnings growth, and 90% of capital spending growth since ChatGPT’s launch.
The concentration is extreme. AI capital expenditures surpassed U.S. consumer spending as the primary economic growth driver in early 2025, contributing 1.1 percentage points to GDP. That level of dependence on a single technology sector has drawn comparisons to the dot-com bubble and the 2008 financial crisis.
Why It Matters
Three of the most prominent figures in technology and finance issued warnings on the same Friday. Goldman Sachs CEO David Solomon said he expects significant capital deployment failures. Jeff Bezos characterized the environment as “kind of an industrial bubble.” Sam Altman, the CEO of OpenAI — the company that triggered the AI investment wave — cautioned that “people will overinvest and lose money.”
When the heads of Goldman Sachs, Amazon, and OpenAI all flag the same risk simultaneously, the signal carries weight. At Yale’s CEO Summit, 40% of the 150-plus executives surveyed agreed that AI hype had driven overinvestment, anticipating a correction. These are not outsider critics — they are the people deploying the capital.
Technical Details
Sonnenfeld and Henriques identify three distinct bubble-burst scenarios. The first is concentration contagion: the major AI players — OpenAI, Nvidia, Microsoft, and Google — are deeply interdependent. A revenue shortfall at one could cascade through the others, similar to how interconnected mortgage-backed securities amplified losses in 2008.
The second is governance failure. The AI sector operates with minimal regulatory oversight, which the authors compare to cryptocurrency’s early years. Anthropic CEO Dario Amodei’s estimate that there is a 25% probability of AI development going “really, really badly” is cited as evidence that even insiders acknowledge systemic risks.
The third scenario is technological disruption. Advances in quantum computing or novel semiconductor architectures could render hundreds of billions of dollars in current AI infrastructure investments obsolete before they generate returns.
Who’s Affected
The concentration risk falls most heavily on passive index fund investors, who are exposed to the Magnificent Seven technology stocks whether they intended to be or not. Pension funds, 401(k) holders, and target-date funds all carry significant indirect AI exposure through S&P 500 tracking.
MIT research adds a sobering data point: 95% of 52 organizations studied achieved zero return on investment despite collectively spending $30 to $40 billion across more than 300 generative AI initiatives. If large enterprises cannot generate ROI, the revenue assumptions underpinning AI stock valuations become difficult to sustain.
What’s Next
The Yale analysis does not predict when a correction will occur, only that the structural conditions for one are in place. The authors note that AI capital spending would need to translate into measurable productivity gains and corporate revenue growth to justify current valuations. If the MIT findings on near-zero enterprise ROI prove representative rather than exceptional, the gap between investment and returns could trigger the repricing that economists increasingly expect.
Sonnenfeld and Henriques stop short of calling a timeline, noting that bubbles can persist longer than skeptics anticipate. The critical variable, they argue, is whether AI companies can demonstrate revenue growth at a pace that closes the gap between current valuations and actual earnings — a gap that widens with every quarter of heavy capital expenditure that produces limited measurable returns.
