Oracle Corp. (NYSE:ORCL) stock tanked roughly 12% Thursday, its worst day in more than a year, after the company laid out an artificial intelligence spending plan far bigger than Wall Street had modeled.

The figure that spooked investors was gross capital expenditure of $90 billion to $95 billion for fiscal 2027, against analyst estimates closer to $61 billion. A clean top- and bottom-line beat barely registered.

The $300 Billion OpenAI Problem

Oracle’s data-center spending will hit $90 billion to $95 billion next year, the figure that drove Thursday’s selloff, and Benzinga reported the bill could climb toward $100 billion. Oracle’s counter is that it only funds about $70 billion of that itself. Customers cover the rest by paying upfront or bringing their own hardware.

That help only goes so far. Oracle burned through $23.7 billion in cash last year, and still plans to raise about $40 billion in debt and equity to keep building.

The bigger worry is who the customers are. Oracle’s backlog hit a record $638 billion, but Bank of America estimates more than half, roughly $300 billion, comes from one company: OpenAI. If OpenAI can’t pay its bills as those data centers come online, Oracle is left holding them, and the debt that built them.

What Prediction Markets Say About The AI Bubble

On Polymarket, traders currently put a 26% chance on an AI bubble bursting by the end of 2026. The bar is deliberately extreme. The contract resolves yes only if, inside a 90-day window, something breaks hard, such as Nvidia Corp. (NASDAQ:NVDA) falling 50%, OpenAI or Anthropic going bankrupt, or H100 rental rates collapsing toward $1 an hour.

Kalshi prices the risk of a recession this year at 17%.

The Nvidia Paradox

On paper, Oracle’s buildout is a tailwind for Nvidia, its biggest chip supplier. The contagion fear flips it. If the buyers stretch their balance sheets to breaking and the debt runway runs dry, the chip orders dry up with them.

If a legacy giant must torch free cash flow and add $40 billion in debt just to serve AI workloads, the promised golden age of high-margin AI may be a capital-intensive slog, though some analysts still see a cash flow waterfall once those centers fill.

The bull case and the bear case on ORCL now collapse into one question: whether OpenAI’s revenue ever catches up to the compute it has ordered.

Image: Shutterstock