Original author: Bu Shuqing
Original source: Wall Street News
Michael Hartnett, chief investment strategist at Bank of America Securities, warned that the current market structure is highly consistent with the dot-com bubble bursting in 2000, and he suggested that investors start buying bonds on dips.
In the latest Flow Show report, Hartnett noted that in the six months leading up to the peak of the dot-com bubble in March 2000, the technology sector was up more than 40%, while the consumer staples sector was down 30%, with all sectors except technology and telecom in decline. He said that this pattern is "exactly the same" as the current market trend. At the same time, 400 constituent stocks in the S&P 500 fell below the 50-day moving average, and 300 fell below the 200-day moving average. The market gains were highly concentrated in the artificial intelligence-related sectors represented by "Mag7".
In the bond market, the 10-year U.S. Treasury yield has risen to 5.33%, the highest level since 2002. Hartnett saw this as a buying opportunity and called out "buy humiliation," advising clients to start adding to their bond positions. He believes that long-term U.S. bond returns have fallen to the lowest level in a century, and similar extreme low returns in history often mark the arrival of generational buying opportunities.
Market gains are extremely divergent, and analog signals continue to shine in 1999
Hartnett used the title "A Tale of Two Cities" in the report to describe the polarization pattern of the current market. He writes that the market is "long AI" (NDX) while "short AI-agnostic assets" (S&P 500 Equal Weight Index SPW), while the Mag7, AI and biotech sectors have shown apparent immunity to a high interest rate environment and "the 1999 analogy still holds."
At the data level, statistics from the Goldman Sachs Technology Trading Desk show that even though the index is at a historical high, the median S&P 500 component stock has fallen 16% from its high point, and the market breadth has fallen to the low level during the dot-com bubble. The day after, the share of stocks trading above the 200-day moving average fell further below 50%.
The historical template cited by Hartnett shows that in the six months before the peak in March 2000, the technology sector alone rose more than 40%, while other sectors fell across the board, with the consumer staples sector falling as much as 30%. He believes that the degree of sector differentiation in the current market is almost the same as that then.
The AI CapEx Bubble: A Historical Mirror from the Railroad Bubble
Hartnett characterizes artificial intelligence as "the biggest bubble since railroads," and in this week's report provides a detailed review of the bursting of two railroad bubbles in the 19th century.
The first railroad bubble ended in 1873: railroad stock prices tripled between 1861 and 1872, U.S. rail mileage expanded from 35,000 to 70,000 miles, and capital spending peaked at about 5 percent of GDP. Subsequently, the Franco-Prussian War triggered a contraction of credit in the United States and the return of European capital. Jay Cooke & Company collapsed in 1873, and approximately 115 railroad companies went bankrupt within the next 12 months.
The second railroad bubble ended in 1881: Railroad stocks rose 2.5-fold from 1877 to 1881, and at one point accounted for 63% of the total market capitalization of the U.S. stock market. Railroad construction quadrupled in four years, creating overcapacity, causing freight rates to fall, revenue and profits to collapse, and ultimately to trigger a banking crisis and the third-longest recession in U.S. history (1882-1885).
Comparing the current situation, Hartnett pointed out that capital expenditures for hyperscalers are expected to reach 3.5% to 4% of GDP in 2027, which is still lower than the peak level of 5% in the railway era, and semiconductor prices continue to rise, which is completely different from the 5% annual decline in freight rates in the railway era.
However, he also warned that both railroad bubbles were supported by falling Treasury yields, "and this is obviously not the current situation."
In addition, the end of both rail bubbles was accompanied by credit events, liquidity collapse and geopolitically driven capital outflows - all three signals are now reflected in Oracle credit default swaps (CDS), the AUD/JPY exchange rate and the French bond market.
'Buy Shame': Hartnett bullish on U.S. Treasuries that have been ignored by the market
In the bond market, Hartnett's stance has become increasingly clear.
He acknowledged that a 100- to 200-basis-point decline in yields would likely require a credit event or recession as a catalyst, but he argued that the current extreme positioning itself constitutes a reason to buy: asset allocators are generally heavy on stocks and short on bonds, betting on a final leg up in U.S. tech stocks and a final leg up in U.S. Treasury yields.
In terms of long-term U.S. bonds, long-term zero-coupon bonds (ZROZ) have fallen by 65% from their peak in March 2020, and the 10-year rolling U.S. bond return is -2%, the lowest in a century. Citing historical data, Hartnett points out that the last time long-term returns for stocks and commodities were equally dismal, it marked the emergence of a generational buying opportunity.
He further set his sights on investment-grade bonds of hyperscale cloud computing companies, pointing out that the price of the U.S. Investment-Grade Technology Bond Index (CITE) fell 9% in the past year, and the yield rose from 4.5% to 6.2%. He believes that given that the AI industry has been endorsed by the U.S. government, the long-term bond yields of ultra-large companies such as Oracle (yield 8.4%), Meta (7.5%), and Google (6.9%) are close to junk bond levels and may soon attract buyers.
Investors began to follow up and increase their bond holdings
The latest EPFR data (as of Wednesday) showed that bonds had a net inflow of $18.8 billion, stocks had a net inflow of $15.8 billion, cryptocurrencies had a net inflow of $900 million, gold had a net inflow of $700 million, and cash outflows were $118 billion due to quarter-end rebalancing.
Specifically, long-term bonds (government bonds and corporate bonds) with a maturity of more than 6 years saw a net inflow of US$7.4 billion in a single week, the largest since May 2025; municipal bonds saw a net inflow of US$4.2 billion, a record high since data was available in 2004; European stocks saw a net inflow of US$1.3 billion, the largest since February 2026; Chinese stocks saw a net inflow of US$3.7 billion, the largest in the past nine weeks; the technology sector saw a net inflow of US$33. billion, the largest in the past five weeks; utility net inflows were US$1 billion, the largest since December 2025.
It is worth noting that the proportion of stock positions currently held by Bank of America’s private clients to assets under management has risen to a record high of 66.3%, while the proportion of cash has dropped to a record low of 9.4%, showing that institutional investors’ stock positions are extremely crowded.
Four key warning lines
In the report, Hartnett listed four market risk trigger levels to track closely: the Global Financials ETF (IXG) fell below $125, the Bond Market Volatility Index MOVE rose above 125, the mid-cap ETF (MDY) fell below $666, and the small-cap ETF (IJR) fell below $135. He pointed out that once small-cap stocks follow the decline of bank stocks, the risk cascade effect of deleveraging will officially start, and policy support will be difficult to stop by then.
BofA's bull/bear indicator fell back to 8.8 from 9.3 this week, which Hartnett attributed to wider risk bond spreads, outflows from high-yield bonds and a rapid deterioration in global market breadth - with a net 27% of global equity index components falling below both their 50-day and 200-day moving averages, their worst levels since March. Although the indicator has retreated, the reading of 8.8 remains in "sell" territory.
Risk warning and disclaimer