Bank of America Compares AI Chip Bubble to 1720 Mississippi Collapse — And the Numbers Back It Up

Bank of America Compares AI Chip Bubble to 1720 Mississippi Collapse

Bank of America strategist Michael Hartnett published a note on Thursday that examined semiconductor valuations using historical comparison points.

The SOX semiconductor index is trading at a price 62% above its 200-day moving average. According to Hartnett's analysis reported by CNBC, the Dow was 30% above its 200-day moving average before Black Tuesday in 1929. It was 28% ahead before Black Monday in 1987. The Nasdaq hit 55% before the dot-com implosion in 2000.

The AI chip sector exceeds all of them.

The Mississippi Bubble Comparison

Hartnett also flagged the French CAC All Tradable index's 73% spread before the Mississippi Bubble burst in 1720. The AI chip sector is now in that range.

That bubble — where shares of the Mississippi Company were literally used as legal tender and doubled France's money supply — destroyed fortunes, destabilized the French economy, and contributed to social conditions that eventually led to revolution.

The price distortion is in the same ballpark.

Nvidia: $4 Trillion

According to Wikipedia's AI bubble tracking, Nvidia became the first company in history to reach a $4 trillion market cap in July 2025 — having quadrupled from $1 trillion in 2023. At that peak, Nvidia alone represented 7.3% of the entire S&P 500.

The company kept climbing past $4 trillion into October 2025.

One chipmaker representing nearly one-tenth of the entire U.S. large-cap index represents a significant concentration risk.

The Passive Fund Exposure

Millions of ordinary Americans hold semiconductor exposure through passive investment funds that automatically track index weightings. Retirement accounts and ETFs are now heavily loaded with semiconductor exposure according to Domain-b's May 2026 reporting. No explicit opt-in occurred; it happened automatically as chip stocks rose.

If AI growth expectations soften — even slightly — the index weighting reverses, passive funds sell automatically, and the contagion spreads fast. This is a 401(k) problem, not just a hedge fund problem.

Wall Street Is Already Hedging

The Los Angeles Times reported in December 2025 that Nvidia and Oracle both saw significant sell-offs as investors started questioning whether the capital expenditure tsunami would ever produce returns. Oracle's stock dropped after it reported mounting AI spending without corresponding revenue clarity.

Sameer Bhasin of Value Point Capital told the LA Times: "These stocks correct when the growth rate doesn't accelerate any further." Not when it goes negative. When it merely stops accelerating. That's a razor-thin margin for error.

Jim Morrow, CEO of Callodine Capital Management, said: "We're anteing up at this point to see whether the returns on investment are going to be good." Translation: nobody actually knows yet.

The Bull Case and Its Limits

US mega-cap spending on AI infrastructure is projected to hit $1.1 trillion between 2026 and 2029, with total AI spending expected to surpass $1.6 trillion, according to Wikipedia's sourcing on late-2025 forecasts.

Economist Ann Pettifor, director of Policy Research in Macroeconomics, told CNBC: "Having to amass more than a trillion dollars in cash to support the investment has led to what everybody talks about as a bubble."

The FT's Robin Wigglesworth offered a counterpoint — telling the Unhedged podcast that compared to the 1860s railroad boom (which scaled to the equivalent of $10 trillion in today's GDP), AI investment is still relatively modest. That's a fair point, though it's also not an argument that valuations are justified.

Coverage and Implications

Left-leaning outlets are covering the bubble angle but focusing heavily on worker displacement and inequality narratives. Center-left coverage like CNBC is reporting the Hartnett note accurately but soft-pedaling the passive fund exposure angle — the part that affects regular people directly.

The AI buildout may be real. The technology may be transformative. Both of those things can be true and the current valuations can still be disconnected from any near-term reality.

Hartnett's data is arithmetic. When a sector trades 62% above its 200-day moving average, historical precedent doesn't offer a single example of that ending quietly.

The question is whether your retirement account is positioned for what happens when semiconductor valuations correct.