Analysis
About 30% of every dollar in the S&P 500 now sits in five companies -- Nvidia, Apple, Microsoft, Alphabet and Amazon -- and most of that concentration traces back to the market's bet on AI, according to Fortune. With roughly $10 trillion of 401(k) money sitting in the stock market, mostly through target-date funds and S&P 500 index funds, the overwhelming majority of American retirement savers now hold a concentrated AI bet they never actively chose.
A Bubble Warning From Inside Treasury
The concentration numbers land alongside a leaked Treasury Department draft report, first reported by NOTUS, that warns an AI-driven downturn could ripple far beyond Silicon Valley -- hitting stock markets, private credit, banks, utilities, chipmakers and cloud providers, the same sectors that dominate many retirement portfolios. Treasury's career analysts reportedly concluded that AI firms are now more deeply entrenched in the U.S. economy than dot-com-era companies were in 2000, and that a sharp contraction in AI valuations or missed productivity targets could stress the financial system in ways that are hard to model in advance. Treasury itself has pushed back, telling reporters the findings are unvetted and that the department remains bullish on AI-led productivity gains -- an internal disagreement that is itself notable, since it means the U.S. government does not yet have a settled view on its own systemic exposure.
“Treasury's career analysts reportedly concluded that AI firms are now more deeply entrenched in the U.S.”
How This Differs From 2000
Pulse has tracked the mechanics of this cycle -- AI infrastructure IPOs racing ahead of the labs themselves -- and the difference from the dot-com bubble is capital intensity, not just valuation. Dot-com companies burned cash on marketing and user acquisition; today's AI leaders are burning it on data centers, chips and power contracts financed increasingly through debt, not just equity. Nscale's $3.36 billion pre-IPO convertible financing and CleanSpark's $2.276 billion in senior secured notes this month alone show how much of the AI buildout now runs through credit markets that don't show up in a stock-index concentration figure, which likely understates total systemic exposure rather than overstating it.
What The Concentration Number Overstates
None of this means the AI trade is about to collapse. Nvidia, Microsoft, Alphabet and Amazon are, unlike most dot-com-era leaders, genuinely profitable at scale, with real revenue growth behind the multiples -- the bear case here is about concentration risk and sequence-of-returns exposure for near-retirees, not a claim that the underlying businesses are fake. A retiree drawing down savings during a sharp AI-stock correction faces a very different outcome than one still decades from retirement, and target-date funds are not uniformly built to account for that asymmetry. Critics of the Treasury draft also note it remains a leaked, internal, non-final document -- not policy -- and Wall Street's own risk desks have so far kept underwriting the buildout rather than pulling back.
What GPs And LPs Should Actually Watch
For venture investors, the practical takeaway isn't about public-market timing -- it's about correlation. A fund whose LP base is heavily exposed to AI-concentrated public equities is more likely to face capital-call pressure in a drawdown precisely when private AI companies need follow-on capital most, compressing exactly the deals GPs would want to lean into. The four bubble signals researchers point to -- top-ten concentration, tech's share of the index, the Buffett Indicator, and how much of the market index funds themselves own -- are all now running hotter than they did ahead of the 2000 crash, which is a reason to model correlated LP behavior explicitly rather than treat public and private AI exposure as separate books.