OpenAI expects to lose hundreds of billions of dollars through 2030 even as revenue grows tenfold, a warning sign for the AI boom but still a win for Nvidia and the chip makers who get paid first.

• OpenAI projects revenue of $350 billion a year by 2030 but plans to spend over a trillion dollars getting there, with most of that money flowing to Nvidia.
• The AI financing loop is fragile: data center builders like NScale are promising compute to Anthropic before they have the money, so a weak NScale or Anthropic IPO would be the canary in the coal mine.
• Model compression is quietly extending the bubble because smaller, cheaper Chinese-style models run well on older chips, so old and new Nvidia hardware are both rising in value.
• That same trend is less good for memory-chip stocks, since faster GPUs need less memory headroom to do the same work.
• Figure's humanoid robots are progressing much faster than expected, and Nvidia owns nearly the entire robot stack from training chips to on-device brains, lifting AMD, Broadcom, Marvell, ARM, and Qualcomm along the way.
• Buffett's Berkshire buying builder Taylor Morrison signals a bet on housing: whether AI causes deflation or a crash, rates fall, mortgages could drop below 2%, and a worsening home shortage pushes prices up.

Outlook: Stay bullish on chips and software near-term, but watch for an AI spending collapse or a weakening job market as the two paths to recession, and use the boom years to pay down debt and build home equity.