HOLY SH*T! Microsoft JUST Started DUMPING AI Data Centers

Summary

Steven Van Metre argues that Microsoft’s decision not to renew some rented Oracle computing capacity signals weakening economics in the AI data-center buildout. He says older Nvidia GPUs may lose value faster than expected, while delays in securing power leave facilities idle and interest costs accumulating. In his view, this could pressure smaller data-center landlords, their lenders, AI-related credit, and ultimately technology stocks.

These are the presenter’s interpretations and forecasts; the subtitles do not independently verify them.

Main Claims and Market Implications

• Microsoft and Oracle: Microsoft reportedly stopped using some Oracle capacity and declined to renew a contract, with older-generation Nvidia chips cited as no longer meeting Microsoft’s requirements. The presenter contrasts this with Oracle’s claim that demand remains strong, including for earlier-generation GPUs.
• Data-center landlords and debt: According to the presenter, Microsoft has committed more than $80 billion to rent capacity from other cloud providers. He says it is also scrutinizing projects and reconsidering new deals. Smaller “neocloud” operators, some of which borrowed at 9–12%, may be vulnerable if expected rental income does not materialize.
• Credit-market stress: The presenter describes a feedback loop in which weaker customer funding or canceled rentals leave data-center operators struggling to service debt, potentially pushing losses onto lenders and weighing on equities. Examples and figures he cites include:
  • CoreWeave: Unsecured notes were said to trade at 78.5–83 cents on the dollar, yielding about 15.5%.
  • Project financing: One example used a 9.875% coupon and an estimated $87 million in quarterly interest before the facility begins earning rent.
  • ING: Reportedly considering selling roughly $500 million in data-center loans, including a $25 million portion tied to a New Mexico site associated with the Stargate project.
  • OpenAI: Said to have secured $11 billion of a targeted $30 billion.
  • AI-related credit spreads: Described as about 44 basis points wider than the investment-grade index, versus roughly 10 basis points tighter a year earlier.
• Stocks and macro outlook: The presenter says credit spreads tend to widen before equities fall and warns that further spread widening could signal more downside. He characterizes the AI boom as debt-dependent and argues that reduced capital spending could weaken growth and inflation expectations, potentially supporting longer-dated Treasuries.
• Recent market move: The Nasdaq reportedly fell 6.4%, erasing $2.3 trillion in market value in a record session. The presenter frames the debate as either a 1998-style liquidity event—potentially a tactical buying opportunity—or a 2000-style structural technology unwind. He says he leans toward the former tactically but advises against buying for the long term.
• Technical and systematic-trading levels: He says Nasdaq-100 levels near 27,600 could trigger CTAs to turn outright short, while a clean break below 26,800 could sharply reduce CTA exposure. For the Nasdaq-100 ETF (QQQ), he sees a topping pattern and a downside path toward an area with heavier three-year volume support, while allowing for a possible bounce.
• Trade example and promotion: Van Metre says subscribers were advised to buy TLT after it moved above its 9-day moving average, add during the rise, and exit at a stop a little over two weeks later, for a claimed net return of about 12%. He promotes paid trading and portfolio-model subscriptions.

Methodologies, Signals, and Cautions

• Credit-market monitoring: Watch AI-company credit spreads as an early warning signal; the presenter expects further widening to pressure equities.
• CTA/systematic levels: Monitor Nasdaq-100 breaks at 27,600 and 26,800 for potential systematic selling.
• QQQ technical approach: Use volume profiles across six-month, one-year, and three-year windows to identify low-volume zones, potential downside movement, and heavier support.
• TLT trade framework described: Buy on a move above the 9-day moving average, add during the upward move, and use a stop-loss. The reported result was about 12% over just over two weeks.
• Explicit caution: The presenter says long-term investors should not buy the dip, while distinguishing that from possible short-term tactical trades. He urges investors with tech exposure to monitor credit spreads and the bond market.

Assets, Companies, and Instruments Mentioned

• Equities, indices, and ETFs: Nasdaq, Nasdaq-100, QQQ, TLT
• Companies and projects: Microsoft, Oracle, Nvidia, CoreWeave, OpenAI, SoftBank, ING, Stargate
• Markets and instruments: Corporate bonds, unsecured notes, data-center loans, high-yield and investment-grade credit spreads, long-dated U.S. Treasuries, CTA/systematic positioning
• Sectors: AI infrastructure, cloud computing, and data centers

Disclosure: No explicit “not financial advice” disclaimer appears in the subtitles. The presenter promotes paid trading and portfolio services.

Presenter and sources mentioned: Presenter: Steve Van Metre. Organizations and companies referenced as sources or subjects include Microsoft, Oracle, Nvidia, OpenAI, SoftBank, ING, Barclays, Morgan Stanley, Bank of America, and Goldman Sachs.