AI forges the strongest moat in history? JPMorgan Asset Management: Two major mid-term risks conceal a liquidation crisis

If the current US stock market could be summarized in one sentence, it would be: **AI is “smotheringly” dominating everything.** According to Chasing Wind Trading Desk, JPMorgan Asset Management (JPMAM) released its 2026 outlook report "Smothering Heights" on January 1st, pointing out that since ChatGPT was released in 2022, 65% to 75% of the S&P 500 index’s return, profit growth, and capital expenditure have come solely from 42 companies related to generative AI. Stripping out these 42 companies, the performance of US stocks trails that of Europe, Japan, and China. Even more astonishing, capital expenditure by the tech sector contributed 40%-45% of US GDP growth over the past three quarters, whereas in the first three quarters of 2023, this figure was less than 5%. This is an unprecedented gamble. The “moat” formed by Nvidia’s chip design, TSMC’s manufacturing, and ASML’s lithography seems unbreakable. The market capitalization of just four semiconductor companies and four hyperscale cloud providers has ballooned from $3 trillion seven years ago to $18 trillion today, occupying 16% of the global stock market’s total value. However, when the market is extremely concentrated and at historic highs, the question investors must ask is no longer “what’s the next catalyst,” but “what could go wrong.” JPMorgan Asset Management’s report clearly identifies two major medium-term risks that could trigger market correction: **first, the risk of not realizing profits after massive capital expenditure (similar to a “metaverse-style” collapse); second, the physical bottleneck of US power infrastructure.** For investors, the script for 2026 may repeat 2025: after a 10%-15% correction triggered by profit pullbacks and growth panic, the market may eventually close higher at year-end, but before that, **questions about AI profitability and energy supply will hang overhead like the sword of Damocles.** ## Risk 1: The Trillion-Dollar Capital Expenditure’s “Metaverse Moment”—Where is the Return? **This is the most urgent threat to the current “moat.”** Since Q4 2022, the four major cloud providers (Microsoft, Google, Amazon, Meta) have spent $1.3 trillion on capital expenditure and R&D, much of it related to generative AI. The core question is: can these massive investments be converted into corresponding profits? If not, these tech giants might face a “metaverse liquidation” similar to 2022, when the "Magnificent 7" stocks were slashed in half. The report notes that **although corporate adoption of AI is rising, debate around return on investment (ROI) is extremely fierce.** MIT research shows that **despite $30-40 billion being invested, 95% of projects have zero return,** and CEOs’ confidence in their AI strategies has plummeted. While Goldman Sachs expects AI to boost productivity, aside from infrastructure providers, very few companies have achieved significant excess returns through AI yet. Moreover, there are hidden concerns in cloud vendors’ finances. Aside from Microsoft, which has published clear AI revenues, other giants’ paths to profit remain unclear. As free cash flow profit margins gradually decline and cash reserves shrink, the market is urgently awaiting clearer profit models. Meanwhile, some cloud vendors are extending the depreciation period for GPUs and network devices—from three to four years, now five to six years—to optimize financial reports. Should accounting standards or chip iterations return depreciation periods to normal, related companies’ EPS and profit margins may face downward revisions of 6%-8%. As a key driver of the AI narrative, OpenAI also harbors risks. Though its revenue growth is rapid, achieving its 2030 goal requires an additional 30 gigawatts of power supply, while 72% of GPT queries are currently non-commercial. These factors make OpenAI a potentially greater single-enterprise risk than Nvidia, further highlighting the sustainability concerns lurking behind the AI investment frenzy. ## Risk 2: Hard Constraints of the Physical World—Imminent US Power Shortage As OpenAI CEO Sam Altman said: “AI’s costs will ultimately converge on energy costs.” The report warns that physical world limitations are becoming the greatest bottleneck to AI development. Currently, data centers account for only 4%-8% of US power demand, but they are expected to drive two-thirds of future load growth. Because improvements in chip efficiency are offset by exponential increases in computing needs, power supply is facing severe challenges. This challenge is first reflected in **a huge supply-demand gap.** Just four OpenAI partner agreements alone will require an additional 30.5 gigawatts of power—75% of the peak added US nuclear capacity over the past five years. Meanwhile, key infrastructure construction badly lags. Core equipment like gas turbines and transformers now require delivery times of three to seven years, with costs continually rising, while the median wait time to connect to the grid is over 70 months. **Regional power crises are already apparent.** In the PJM grid area, which is dense with data centers, rapid retirement of fossil-fired plants and soaring data center loads have driven up power capacity prices. California has even seen data centers built but forced to sit idle due to lack of electricity. Although solar and battery storage are seen as partial solutions, considering supply reliability and total costs, it is expected that 60% of new data centers will still rely on natural gas generation. This means AI’s expansion speed will depend not just on algorithms and code efficiency, but increasingly on the carrying capacity of America’s aging grid and the buildout speed of traditional energy infrastructure such as natural gas pipelines. While solar and batteries are one solution, considering stability and cost, 60% of new data centers will still rely on natural gas generation. This means, **AI’s growth rate will be determined less by code efficiency and more by the carrying capacity of America’s old grids and the speed of natural gas pipelines construction.** ## Market Valuation and Debt Risks: It’s More Than Just P/E Ratio Despite these risks, the report believes that current tech stock valuations, though high, aren’t as crazy as the 2000 dot-com bubble. Today’s PEG (price/earnings to growth ratio) is just 1-3 times, far below the 4-8 times back then. Also, today's tech titans have extremely high profit margins, a fundamental difference from “young, unprofitable companies” (YUCs) that filled the market then. However, subtle shifts in financing structure are worth noting. Although AI giants have very low net debt, some firms are starting to borrow to fund data centers. For example, Oracle, whose cash flow is weaker than other big names, has taken on substantial debt to meet OpenAI’s computing needs; Meta is using off-balance-sheet financing tools via Blue Owl to build data centers. Although S&P maintains its rating, if these hidden debts are consolidated, Meta’s leverage would rise sharply. This shows that the AI arms race is quietly moving from "cash flow-driven" to "debt-driven," undeniably increasing systemic fragility.Risk DisclaimerThe market has risks, investments must be made cautiously. This article does not constitute personal investment advice and does not take into account individual users’ specific investment objectives, financial conditions, or needs. Users should consider whether any opinions, views, or conclusions herein fit their particular circumstances. Invest accordingly at your own risk.