Thesis: The AI profit boom is a house of cards built on concentration
Recent headlines celebrate AI‑powered earnings growth, yet the underlying structure is alarmingly narrow. J.P. Morgan’s latest note flags a cluster of vulnerabilities that could turn today’s exuberance into tomorrow’s correction.
Evidence: Numbers that betray the hype
According to The Decoder, only 42 AI‑related firms listed in the S&P 500 generate between 65 % and 80 % of the index’s total profit. That concentration means a handful of companies dictate the majority of earnings, leaving the rest of the market thinly supported.
The same report points to a semiconductor rally whose technical patterns echo those of the dot‑com era. Leveraged chip exchange‑traded funds have amplified their market weight five‑fold since early 2024, suggesting that investors are piling on debt‑linked exposure to the same hardware that fuels AI models.
J.P. Morgan identifies three layers of concentration risk: market‑level (few firms dominate earnings), infrastructure‑level (chip supply chains are over‑leveraged), and macro‑level (the broader economy could feel a shock if AI spending stalls).
Context: What the broader AI push looks like
Even as the financial metrics tighten, AI firms are expanding aggressively. TechCrunch AI reported that OpenAI hired the former Uber India chief on June 26, 2026 to steer its largest market outside the United States. The move underscores OpenAI’s intention to deepen its footprint in a region with massive growth potential, but it also adds to the pressure on data‑center capacity and talent pipelines.
At the same time, Anthropic’s Claude is stealing paid users from ChatGPT, according to a June 25, 2026 TechCrunch AI story. While ChatGPT still leads, the shift indicates a competitive market for premium AI services, each vying for the same compute resources.
Beyond pure profit, AI chatbots are entering sensitive domains. A June 26, 2026 article from the Competitive Enterprise Institute argues that free‑market AI tools can serve mental‑health needs without government oversight. If providers scale up to meet demand, the strain on GPU farms and networking gear will intensify, further tightening the infrastructure supply chain.
Counter‑Arguments: Why some see a sustainable surge
Proponents point to the sheer size of AI‑related revenue and the global appetite for intelligent services. They argue that the semiconductor rally reflects genuine demand for faster, more efficient chips, not just speculative buying. The hiring spree at OpenAI and Anthropic’s user gains are cited as proof that the market is still expanding, with new applications and geographies opening fresh revenue streams.
Critics of J.P. Morgan’s view note that concentration is not unusual in tech‑driven cycles; the internet era also saw a small group of firms dominate early earnings before broader diffusion. They contend that leveraged ETFs simply provide investors with a way to capture growth, and that the five‑fold increase since 2024 is a response to real, not imagined, demand.
Prediction: A tightening squeeze could force a recalibration
If the concentration risk materializes, we may see a two‑phase adjustment. First, a correction in chip‑related equities as investors unwind leveraged positions, echoing the dot‑com crash’s “sell‑off” phase. Second, AI service providers could confront higher infrastructure costs as supply constraints drive up GPU and networking prices.
In that scenario, firms that have diversified revenue streams beyond pure AI services—or that own parts of the hardware stack—will weather the shock better. Companies like OpenAI, which are expanding into emerging markets, may find new growth pockets, but they will also need to secure local data‑center capacity, a costly endeavor.
Conversely, if demand continues to outpace supply, the market may simply absorb the higher costs, passing them to enterprise customers. That would keep profit margins high for the 42 dominant AI firms, but it could also widen the gap between them and smaller players, reinforcing the concentration J.P. Morgan warns about.
Ultimately, the next 12‑18 months will test whether AI’s profit surge is a sustainable engine or a fragile bubble perched on a narrow base of firms and hardware. Investors should watch chip‑ETF flows, earnings reports from the 42 key players, and any signs of supply‑chain bottlenecks as the most immediate barometers of risk.
Explore related AI topics
