Anyone checking their 401(k) this fall could be forgiven for thinking Wall Street has shaken off its worries.
The S&P 500 ended October 22 at 7,623.68, up 11.4% for the year and just 2.5% below its October 6 record. More recently, however, the rally has lost some steam. After three years of AI-fueled gains, the S&P 500 has managed a more modest 3% advance over the past three months.
For investors who stayed in the market through its twists and turns, those are comforting numbers as they look toward 2027.
But in a note shared with me, Bank of America’s Michael Hartnett sees a major problem hiding behind the gains.
As we all know, the AI boom that has helped power stocks depends on extraordinary spending on chips, data centers, and electricity. Much of that expansion touches the bond market, where borrowing costs are rising and lenders are getting choosier.
Hartnett worries investors are overlooking what happens if that money becomes harder to raise. In Bank of America’s October survey, which was shared with me, just 6% of fund managers expected a hard economic landing.
He sees three possible paths for 2027, including one that could leave stock market investors painfully exposed.
Michael M. Santiago / Getty Images
Bank of America warns AI financing could threaten stocks in 2027
Wall Street’s still betting the house on AI heading into 2027, but Michael Hartnett feels it’s overlooking the enormous cost of keeping the boom alive.
More Bank of America:
- BofA splits software into winners and losers as AI agents arrive
- Bank of America reveals hidden moat that makes Morgan Stanley a buy
- Bank of America resets Amazon stock forecast on delivery plan
For perspective, BofA semiconductor analyst Vivek Arya estimates spending by major cloud companies could climb from a whopping $860 billion in 2026 to $1.2 trillion in 2027, nearly 40% growth as reported by Biggo. Moreover, to finance that tremendous investment in chips and energy infrastructure, businesses are turning to debt markets.
Amazon (AMZN), Alphabet (GOOGL), Meta Platforms (META), and Oracle (ORCL) issued $194 billion in bonds through July 7, Reuters reported, compared with $108 billion throughout 2025, and that dependence on borrowing is precisely what worries Hartnett.
He warns that hyperscaler bond spreads widening beyond 200 basis points, or two percentage points above comparable benchmark yields, could trigger forced selling and broader financial stress.
Lenders are already distinguishing between borrowers.
JPMorgan’s October analysis identified credit-default swap spreads of 50 to 120 basis points for financially stronger AI companies, versus roughly 300 basis points for some infrastructure borrowers and 1,100 basis points for a heavily leveraged borrower.
Yet investors appear remarkably confident.
BofA’s October fund manager survey found that 90% expected either a soft economic landing or no landing at all, as reported by Investing. Just 6% anticipated a hard landing, while technology-stock exposure was at its highest overweight since August 2020.
That leaves little room for disappointment.
Consider a 20% shortfall against BofA’s 2027 spending forecast. Investment would still grow 11.6% from 2026, but fall $240 billion short of expectations.
For investors, that’s the danger.
AI spending doesn’t have to collapse to hurt stocks. Even slower growth could undermine earnings forecasts across chipmakers, data-center suppliers, and the broader market.
And Hartnett believes history offers a troubling precedent.
Bank of America’s 2027 AI warning echoes the railroad crash of 1873
The last time America went all-in on infrastructure that promised to completely change everyday life, the technology delivered. Many of its investors didn’t.
Railroads connected cities, opened markets, and moved goods faster than ever before. But laying thousands of miles of track required enormous borrowing, often before the routes earned enough to repay lenders.
By 1873, construction had outpaced demand, and European investors were pulling money out of American railroad bonds as covered by PBS. Financing dried up. That September, Jay Cooke & Co., a major backer of the Northern Pacific Railway, collapsed. The New York Stock Exchange closed for ten days. At least 100 banks failed, according to the Federal Reserve, while PBS reports that 89 of America’s 364 railroads eventually went bankrupt.
The investment collapse was striking.
Federal Reserve Bank of St. Louis data show that builders completed 7,439 miles of new railroad track in 1872, but just 1,606 in 1875. That’s a 78.4% drop in annual construction.
The railroads remained useful. What failed was the assumption that investment could continue to expand without interruption.
Hartnett sees a possible parallel in AI’s trillion-dollar buildout.
If companies spend 20% less than BofA’s projected $1.2 trillion in 2027, that’s $240 billion in anticipated spending that doesn’t materialize. AI investment would continue to grow, but businesses relying on larger orders could face a painful reset.
That said, it’s important to acknowledge that the largest technology companies generate substantial cash, unlike many speculative railroad ventures. Still, revolutionary technology doesn’t guarantee attractive returns for everyone financing it.
History shows how a boom can break. It cannot tell investors whether 2027 brings another AI surge, a broader rally, or a bust.
BofA outlines three paths for stocks in 2027
For Hartnett, 2027 is the year investors may discover whether optimism survives a reality check.
One possibility is that the AI bubble keeps growing. Companies borrow, build data centers and promise profits, while investors bid up valuations. Stocks could climb even as risks accumulate.
A healthier outcome is a Goldilocks economy. Inflation cools, borrowing costs ease and gains spread beyond technology giants. Forecasts for roughly 15% S&P 500 earnings growth as polled by FactSet, and a 40-basis-point decline in Treasury yields support that possibility.
Then comes Hartnett’s concern, which is a potential deflationary bust. Credit gets expensive, weaker AI borrowers struggle, and delayed projects hurt chipmakers, contractors, and other suppliers. Falling earnings expectations could spread the damage across stocks.
BofA questions whether investors appreciate that downside. The clues will emerge in bond spreads, capital-spending guidance, cash flow, earnings revisions, and evidence of broader market participation.
Related: Jim Cramer’s 5% to 8% rule puts Nvidia stock in focus
