The Coming Handoff in AI Investing
The market faces a shift that investors are not pricing in. It is moving from "buy everything" - semiconductors, optical stocks, data center stocks - to a pickier stage where the real question becomes whether these investments will pay off. History shows every such transition carries a risk that investors ignore.
Interest rates make this worse. Focus mostly on the 10-year Treasury yield. Right now it sits around 4.78%. A higher yield lowers the intrinsic value of long streams of future cash flow. The returns on today's AI investments come 2, 3, or 4 years down the road, so the discount rate matters a lot. If yields climb, the value of all these company investments drops.
The Global Debt Problem
There is too much debt in the world, both government and corporate. Total US debt is over $40 trillion. Bond selloffs are already happening. Short term it can improve or worsen, but the long-term problem stands. A "day of reckoning" will come - maybe next month, maybe a year or two out.
Debt keeps growing with no consequences, so it keeps growing more. Now the government competes for borrowing against the hyperscalers and the data center buildout, which adds more debt to the system. Unless the economy grows a lot, that debt becomes a problem.
What does a day of reckoning look like for the market? A Treasury auction partly fails, meaning a very long tail on the auction. Corporate bond yields spike. Spreads widen. Right now investors are not worried - they are buying corporate bonds, which is giving Treasuries a run for their money - but that calm can break.
How Far Could Stocks Fall?
A 30% to 50% drop is easy to picture. US stock market capitalization stands near 238% of GDP (some say 245% or 250%). That is higher than at the dot-com peak, the 2008 financial crisis, the 1987 crash, and even 1929. Because the stock market is so large relative to GDP, even a 30% drop wipes out a big chunk of GDP.
The macro data looks fine on the surface. Strip out all AI-related spending and the economy grows only about 1% - not gangbusters. That is the real issue.
Spending drives the danger. In the K-shaped economy, the bottom 50% do not spend, and the next 20% do not spend either. The top 10% do the spending, and they spend because they generate huge wealth from the stock market. If the market corrects, that spending stops. AI is already causing some job losses at the lower end. So the economy can slow very fast, and markets overshoot on the way up and undershoot on the way down. Stay alert over the next 3 years or so.
Cyclical or Secular Earnings?
Earnings are running hot. Consensus for the S&P 500 is $412 for 2027, an amazing level. Earnings have never been this high. Against the earnings trend line over the last 50 years, current earnings sit 60% above trend.
The key question: is this earnings growth cyclical or secular? Maybe AI is a magical bullet that keeps it going. But if it proves cyclical, earnings can drop hard - a 20% to 30% fall in earnings on top of the debt and valuation problems would not be a great day.
The Policy Trap
The current plan, as the trade secretary frames it, is to run the economy hard and grow out of the debt. That beats cutting spending and raising taxes. But the US economy is already running hot, and modeling shows it is very hard even to shrink the deficits, let alone pay down debt. So debt keeps rising.
Refinancing adds pressure that few discuss. Treasuries coming due over the next year carry an average interest rate near 3.23%. Depending on the duration the Treasury issues, replacements will cost 4%, 4.75%, maybe 5%. The cost keeps climbing.
The real solution has to be cutting government spending. Whether politicians will find the courage to do it is unknown.
