In February 1994, Federal Reserve Board Chair Alan Greenspan raised interest rates for the first time in five years, and bond traders were taken completely by surprise.
The yield on the 10-year Treasury note rocketed from roughly 5.6% to nearly 8% within twelve months, a shock Wall Street still calls “the bond market massacre.”
Stocks paid the price, and the S&P 500, fresh off a strong 1993, went nowhere in 1994.
Nothing about the economy was broken, as earnings kept growing and the personal-computer boom that would define the decade was only getting started. Investors simply had to survive the one variable that mattered more than any other that year: interest rates.
Then rates stabilized, and the market launched higher, gaining more than 34% in 1995, the opening act of one of the greatest bull runs in history.
I bring this up because it explains a question that has been driving me a little crazy: Why is not the AI trade breaking out, when I am looking at the best fundamental backdrop I have analyzed in years?
I cover this and more in our latest episode of Being Exponential with Luke Lango. Watch it below:
Stuck, Not Broken
Pull up almost any AI benchmark and you see the same picture. The Invesco QQQ Trust (QQQ) sits almost exactly where it traded in early May. The VanEck Semiconductor ETF (SMH) sits lower than it did in early May. The broader S&P 500 has done a little better, but it, too, has spent four months chopping sideways, every dip bought and every rip sold.
That flatline is not because artificial intelligence stopped paying off. Look at forward operating margins for the S&P 500. From 2007 through the launch of ChatGPT in late 2022, margins bounced in a tight, boring range between roughly 14% and 16% for fifteen straight years. Then ChatGPT launched, and margins broke that range entirely, climbing to roughly 21% today. That breakout is proof, in a single chart, that AI is not a cost center. It is a margin machine, and corporate America is banking the difference.
PwC put a number on how long this runs: $31.6 trillion in cumulative AI infrastructure capital expenditures through 2050, with annual spending climbing from roughly $800 billion today to $1.8 trillion by mid-century. That is not a boom. That is a multi-decade buildout, and we are still in its first few years.
So if the fundamentals are this strong, why are stocks stuck?
The Chart That Explains the Mystery
Here is the one that unlocked it for me. Forward 12-month earnings-per-share estimates for the Philadelphia Semiconductor Index keep climbing. The index’s actual stock price keeps falling. Since mid-June, those two lines have ripped apart.
Nvidia Corp. (NVDA) was supposed to fix this. It delivered the blowout quarter AI bulls wanted, investors bought the news for about a day, and then gave back most of that pop, roughly 60% to 70% of it, within a week.
Something other than earnings is driving this market, and I believe I found it: excess liquidity, a measure of how much money is left over after the real economy soaks up what it needs (money supply growth, minus inflation, minus real economic growth). When that number is positive, extra dollars slosh into stocks and expand valuation multiples even without earnings growth. When it turns negative, multiples compress no matter how good the earnings look.
Right now, excess liquidity has crashed, turning negative for the first time since late 2023, in the sharpest decline since the 2022 bear market.
Two Forces, One Squeeze
The timing here is difficult to ignore. This flatline began almost exactly when Kevin Warsh took over as Fed chair in May, and the 10-year Treasury yield has climbed steadily since, from roughly 4.4% to 4.8% today.
Warsh has run a hawkish playbook since taking the gavel, and the bond market is pricing in that stance for as long as he holds the job, squeezing money supply growth from one side. Renewed tension in Iran, with oil pushing back toward $90 a barrel, is lifting inflation expectations from the other. Slower money growth plus higher inflation is a textbook double squeeze on excess liquidity, and that squeeze, not weak AI demand, is what is holding this market in a range.
Watch the 10-year yield here. We are sitting right at 4.8%, a high last seen in January 2025, and 5.0% is the next real ceiling. Push toward that level, and expect markets to get jittery. Break decisively above it, and things could turn genuinely ugly.
Why I Still Recommend Staying Long
Here is my base case, and it is why I remain bullish. Fed policy and the Iran conflict both feel capped and contained by political incentives that are unusually well aligned right now. Treasury Secretary Scott Bessent hates higher yields, and this administration already showed, during the Liberation Day tariff episode, that it reverses course fast when the 10-year spikes.
My base case has Iran cooling as the administration pushes for rapid de-escalation, oil sliding back toward $70 to $80, and Warsh gradually softening his tone as those pressures fade. Excess liquidity then reverses, and the gap between earnings and stock prices closes the way it always does: prices snap back up to meet earnings, not the other way around. We are one Trump tweet away from Iran de-escalating meaningfully, and missing the ten best days of a rally like that costs you the entire return.
Until that snapback arrives, I recommend leaning on sectors that have held up through this squeeze: biotech, through something like the SPDR S&P Biotech ETF (XBI); energy, through refiners such as Marathon Petroleum Corp. (MPC), a direct hedge on Iran; and software, which grows relatively more attractive as rates rise. I am not making wholesale changes to the Hypergrowth Investing portfolio, because the core AI trade is still where the real earnings momentum lives, and history says the snapback comes fast.
Where the Real Money Gets Made
The 1994 bond scare did not end the PC boom. It paused it for a year, then handed patient investors one of the best entry points of the decade. I believe AI infrastructure is working through the same kind of pause right now, driven by the bond market, not by anything wrong with AI itself.
That buildout does not stop at the chip. Every dollar PwC is forecasting moves through data centers, power systems, connectivity, and, increasingly, physical machines that put AI to work outside a screen. Elon Musk is racing to control more of that stack than anyone else on Earth, and the companies quietly supplying what he cannot build himself are where I am spending most of my research time right now.
I break down which companies pass that test, and Musk’s entire Vertical AI Masterplan, at a free workshop tomorrow at 8 p.m. Eastern, alongside my colleagues Louis Navellier and Eric Fry.
Reserve a seat right here, and for the full breakdown behind everything above, watch the complete episode of Being Exponential right here.
The post The AI Trade Is Not Broken, But It Is Handcuffed by a Single Number appeared first on InvestorPlace.
