31 October 2025 — Peak AI Concentration: Nvidia at $5 Trillion and the Bubble Question

On Wednesday 29 October, shares of Nvidia rose 4.5%. When the closing bell rang, Nvidia was worth $5.04 trillion — the first publicly listed company ever to cross that threshold.¹ Nvidia was worth more than the combined value of every listed bank in the U.S. and Canada. More than the entire U.K. stock market. Larger than the annual economic output of every country except the U.S. and China.² Four months ago it passed $4 trillion.

Nvidia’s rise is a business story. Its specialised chips are the single most efficient way to train and run modern AI. In the three months to September, it posted quarterly revenue of $57 billion, up 66% year-on-year.³ Above all, it benefits from an extraordinary dynamic: almost all its biggest customers are customers of each other. Microsoft is spending $80 billion on AI infrastructure this year, most with Nvidia, partly to run services OpenAI sells back to users. OpenAI has signed a separate $100 billion deal with Nvidia. Oracle has a $300 billion contract with OpenAI.

The bubble question has become unavoidable. Nvidia accounts for roughly 8% of the S&P 500 by itself. The 10 largest companies — nine AI-linked — account for about 30% of the entire index. An ordinary investor buying a “diversified” S&P 500 fund is putting 8 cents of every dollar into a single AI chipmaker.

The most influential worried voice has been the Bank of England. Its Financial Stability Report this month warned of “the growing risks of a global market correction due to a possible overvaluation of leading AI tech firms.”² Governor Andrew Bailey has said dot-com comparisons are “worth taking seriously.” Jeremy Grantham, 87, has called the AI boom “the most expensive bubble in American history by many measures.”⁴ Grantham has been warning for four years while the market has kept rising.

A more awkward concern is visible in filings. OpenAI is understood to be losing billions a year and expects significant losses through 2028. AI-related companies have committed to roughly $3 trillion of data-centre investment by 2028, about half financed by private credit.² The resemblance to the late 1990s is not lost on older investors: then, an infrastructure boom produced valuations built on customers built on debt. When the circular spending stopped, valuations collapsed 78%.

For an ordinary investor today, the question is not “Is this a bubble?” but “Am I comfortable with how much of my savings rests on the AI story?” If you hold a typical globally diversified portfolio, roughly 15–20% of your equity exposure is in the Magnificent Seven. On the way up, concentration delivers outsized gains; on the way down, it works in reverse. History suggests investors love concentration until it hurts them, and then remember, belatedly, that diversification exists for a reason.