Top analyst on Nvidia fueling bubble: We've seen this before. It was Enron, Tyco (fortune.com)

Morgan Stanley Wealth Management CIO Lisa Shalett warned that the current U.S. equity rally is dangerously concentrated on one narrative—massive capital spending for generative AI infrastructure—and likened the pattern to past market manias (Cisco/dotcom analogies). She says roughly 40 stocks (the Magnificent Seven plus ~34 AI data‑center ecosystem companies) account for about 75% of S&P 500 returns since 2022; the index is up ~90% while the other 493 names gained ~25%. Hyperscalers are spending nearly $400 billion a year on capex, which Morgan Stanley estimates added ~1 percentage point to Q2 GDP, but free cash flow for these firms has turned negative and Strategas forecasts FCF could fall >16% over the next 12 months—signs Shalett calls “seventh inning” risks that could precipitate a correction within 24 months. Her key technical concerns center on “circular” financing and interwoven ownerships that increase systemic risk: Nvidia’s headline investments (a reported $100B against OpenAI, $5B to Intel) and complex vendor/offtake deals, Oracle’s ~$300B data‑center pact (an estimated $100B of borrowing), and cross‑holdings between chipmakers, cloud providers and AI startups. That could amplify contagion if capex slows or credit strains emerge (CDS spreads are being watched). Bank of America’s analysts are more sanguine, arguing many deals are performance‑linked and that Nvidia’s disclosed direct funding is far smaller; both camps agree, however, that a period of digestion or repricing is likely—meaning compute availability, pricing and project timelines could tighten, and ML practitioners should plan for higher infrastructure risk and potential cost volatility.