When the Bill Came Due

October didn't kill the AI story. It asked it a question it had been avoiding: who pays for all of this, and when do they get paid back?

For months the trade ran on a simple promise: spend now, dominate later. Then the bills started arriving in the same week as the earnings. Meta beat on revenue and still fell 7% in a day, after disclosing roughly $825 billion in future AI commitments. Alphabet's cloud business doubled, and the stock still dropped double digits on negative free cash flow. Amazon held its $220 billion spending plan and reported a $42.4 billion cash outflow. Intel guided down and lost nearly 30% in a session.

The month's numbers tell the story plainly. The Nasdaq composite fell 7.8% from its high and had its worst month since March 2025. Nvidia dropped more than 9%, its worst month since November 2025. CoreWeave fell 37%. The Nasdaq-100 closed October with a 3.5% plunge in a single day, near correction territory.

None of this means AI is a mirage. Demand is real; cloud revenue doubling is not a hallucination. What October repriced was not the technology but the financing. Jensen Huang himself said demand for AI computing is "off the charts" while capital to fund it is scarce. The hyperscalers now account for more than a third of all S&P 500 capital spending. That is not a bubble verdict; it is a concentration fact.

The reflection I keep coming back to: every great buildout in history: railroads, electricity, the internet, had a moment like this, when the market stopped paying for the vision and started charging for the bill. The buildout usually survives. The easy money rarely does.

October was not the end of the AI trade. It was the month the trade grew up or at least started paying adult prices.