Bank of America Securities chief investment strategist Michael Hartnett is sounding a warning: the current market structure closely mirrors the eve of the dot-com crash in 2000, and he is advising investors to start buying bonds on weakness.

In the latest edition of the "Flow Show" report, Hartnett noted that in the six months before the March 2000 peak of the dot-com bubble, the technology sector gained more than 40% while consumer staples fell 30%, with every sector except tech and telecom declining. He said this pattern is "virtually identical" to current market action. Meanwhile, 400 S&P 500 constituents have fallen below their 50-day moving average and 300 have dropped below their 200-day moving average, with market gains highly concentrated in AI-related sectors represented by the "Mag7."

In the bond market, the 10-year U.S. Treasury yield has risen to 5.33%, the highest since 2002. Hartnett views this as a buying opportunity, championing the slogan "buy humiliation" and recommending clients begin increasing bond allocations. He believes long-term Treasury returns have fallen to a century low, and historically such extreme low returns have often marked the arrival of generational buying opportunities.

Extreme Market Divergence Persists as 1999 Analogy Signals Keep Flashing

Hartnett titled his report "A Tale of Two Cities" to depict the current market's polarized landscape. He wrote that the market is "long AI" (NDX) while simultaneously "shorting everything unrelated to AI" (the S&P 500 equal-weight index SPW), and that Mag7, AI and biotech sectors have shown clear immunity to the high interest rate environment, with "the 1999 analogy still holding."

On the data front, Goldman Sachs' technology trading desk statistics show that even with the index at historic highs, the median S&P 500 constituent has fallen 16% from its peak, and market breadth has dropped to levels seen during the dot-com bubble era. A day later, the percentage of stocks trading above their 200-day moving average fell further below 50%.

The historical template Hartnett cites shows that in the six months before the March 2000 peak, the tech sector alone gained over 40% while all other sectors declined broadly, with consumer staples plunging as much as 30%. He believes the current degree of sector divergence is almost exactly the same as back then.

AI Capex Bubble: Lessons from the Railway Mania

Hartnett characterizes artificial intelligence as "the biggest bubble since railways," and in this week's report provides a detailed review of the bursting of two 19th-century railway bubbles.

The first railway bubble ended in 1873: between 1861 and 1872, railway stock prices tripled, U.S. railway mileage expanded from 35,000 miles to 70,000 miles, and capital expenditure peaked at roughly 5% of GDP. Subsequently, the Franco-Prussian War triggered a U.S. credit contraction, European capital flowed back home, Jay Cooke & Company collapsed in 1873, and approximately 115 railway companies went bankrupt within the following 12 months.

The second railway bubble ended in 1881: railway stocks rose 2.5-fold from 1877 to 1881, at one point accounting for 63% of total U.S. stock market capitalization. Railway construction quadrupled in four years, creating excess capacity, freight rates declined steadily, revenue and profits collapsed, and ultimately triggered a banking crisis and the third-longest economic recession in U.S. history (1882 to 1885).

Drawing a comparison to the present, Hartnett notes that hyperscaler capital expenditure is projected to reach 3.5% to 4% of GDP by 2027, still below the railway era's 5% peak, and semiconductor prices continue to rise — in stark contrast to the railway era when freight rates fell 5% annually.

However, he also warns that both railway bubbles were underpinned by declining government bond yields, "and that is clearly not the case today."

Moreover, both railway bubble bursts were accompanied by credit events, sudden liquidity drops and geopolitically driven capital outflows — and all three signals are now already visible in Oracle credit default swaps (CDS), the Australian dollar versus Japanese yen exchange rate, and the French bond market.

"Buy Humiliation": Hartnett Turns Bullish on Shunned U.S. Treasuries

In the bond market, Hartnett's stance is becoming increasingly pronounced.

He acknowledges that a 100 to 200 basis point decline in yields may require a credit event or recession as a catalyst, but argues that current extreme positioning itself constitutes a reason to buy: asset allocators are broadly overweight stocks and short bonds, betting on one last rally in U.S. tech stocks and one last leg up in Treasury yields.

On the long-end Treasury front, long-duration zero-coupon bonds (ZROZ) have fallen 65% from their March 2020 peak, and the 10-year rolling Treasury return stands at -2%, the lowest in a century. Hartnett cites historical data showing that the last time stock and commodity long-term returns were equally dismal, it precisely marked the emergence of a generational buying opportunity.

He further turns his attention to hyperscaler investment-grade bonds, noting that the U.S. investment-grade technology bond index (Cartica Acquisition Corp, CITE) has fallen 9% in price over the past year, with yields rising from 4.5% to 6.2%. He argues that given the AI industry's backing from the U.S. government, long-term bond yields of hyperscalers such as Oracle (8.4% yield), Meta (7.5%) and Google (6.9%) have approached junk bond levels and may soon attract buying interest.

Investors Begin Following Suit with Bond Purchases

The latest EPFR data (as of the previous Wednesday) shows bonds received net inflows of $18.8 billion, equities net inflows of $15.8 billion, cryptocurrencies net inflows of $900 million, gold net inflows of $700 million, while cash saw outflows of $118 billion due to quarter-end rebalancing.

Specifically, long-duration bonds (government and corporate) with maturities exceeding six years saw weekly net inflows of $7.4 billion, the largest since May 2025; municipal bonds drew net inflows of $4.2 billion, a record high since data began in 2004; European equities attracted net inflows of $1.3 billion, the largest since February 2026; Chinese equities saw net inflows of $3.7 billion, the largest in nearly nine weeks; the technology sector attracted net inflows of $3.3 billion, the largest in nearly five weeks; and utilities drew net inflows of $1 billion, the largest since December 2025.

Notably, Bank of America's private clients now hold equity positions at a record 66.3% of assets under management, while cash allocations have dropped to a historic low of 9.4%, indicating institutional investors' equity positioning is extremely crowded.

Four Key Warning Lines

Hartnett lists four market risk trigger levels to closely monitor: the Global Financials ETF (IXG) falling below $125, the bond market volatility index MOVE rising above 125, the Mid-Cap ETF (MDY) falling below $666, and the Small-Cap ETF (IJR) dropping below $135. He notes that once small-caps follow bank stocks lower, a risk cascade of deleveraging will officially begin, and even policy support will struggle to stop it.

Bank of America's Bull & Bear Indicator fell from 9.3 to 8.8 this week, which Hartnett attributes to widening high-risk bond spreads, outflows from high-yield bonds and rapidly deteriorating global market breadth — a net 27% of constituents in the global equity index have simultaneously fallen below both their 50-day and 200-day moving averages, the worst level since March. Despite the pullback, the 8.8 reading remains in "sell" territory.