The current AI market exhibits a frightening fragility masked by impressive index gains. According to J.P. Morgan analysts, since ChatGPT's launch in late 2022, a mere 42 companies in the S&P 500 have driven 65% to 80% of the index's total earnings, revenue, and investment. This abnormal capital concentration has redrawn the financial landscape: today, the ten largest U.S. stocks account for 40% of the S&P 500's market capitalization. For context, their share did not exceed 17% in 2015. We are not witnessing broad economic growth, but rather the inflation of a top-heavy elite.
Technical Indicators of Overheating
Technical red flags are multiplying, echoing the fever of the dot-com era. J.P. Morgan highlights several alarming factors:
An explosion in margin lending for stock purchases. Retail investor dominance in chipmaker options trading. A fivefold increase in leverage impact within semiconductor ETFs since the start of 2024.
While Nvidia currently holds 85% of the AI accelerator market, this dominance is not permanent; analysts expect its share to slide to 75% by 2026.
Cloud giants like Google and Amazon are aggressively deploying proprietary chips, aiming to slash operating costs by 30–40% and eliminate the so-called "Nvidia tax."
The ROI Problem and the Open-Source Threat
Institutional skepticism is mounting due to the massive gap between infrastructure spending and actual software returns. While revenues at OpenAI and Anthropic are climbing, skyrocketing compute costs keep the question of long-term profitability wide open. Free cash flow for cloud providers is shrinking as debt loads rise—the market is effectively doubling down, betting on a miracle in the software layer.
The situation is further complicated by Chinese open-source models offering top-tier performance for pennies, stripping Western developers of their ability to maintain high margins. The hardware market is overheated, and the software sector has yet to prove it can foot the bill.