Making Connections

Writing a quarterly newsletter for clients is a terrific exercise because it forces us to organize our thoughts on everything we’ve been reading about the economy and financial markets. And in this process of synthesizing all the various data points and financial news with our decades of experience and historical perspective, we often make connections that aren’t immediately obvious.

In my Winter 2025 PASI News article (to be published in the coming weeks) I look at U.S. economic history in the post WWII period and assess whether any of the conditions which previously led to recessions are present today. There have been 13 distinct recessions in the last 80 years. Two were caused by lower government spending following major wars (WW2 and Korea), one was caused by a global pandemic (2020), and the rest were due to either Federal Reserve interest rate hikes in response to elevated inflation or the bursting of major investment bubbles (most notably the internet bubble in 2000 and housing bubble in 2008).

While inflation still remains a bit above the Fed’s 2.0% target – averaging around 2.7% in 2025 – about 0.5 percentage points of that difference reflected a one-time adjustment from President Trump’s tariffs, which are likely to be overturned by the Supreme Court in the near future (70% odds according to betting markets). In addition, oil prices are low (currently $58 per barrel), the housing market is weakening, and unemployment is rising modestly, easing pressure on wage growth. The result is that the Fed is currently lowering, rather than raising, interest rates, which means the threat of a Fed-induced recession looks unlikely in 2026.

There is, in our view, a lot of speculative activity surrounding artificial intelligence in the financial markets reminiscent of the 2000 internet bubble, which we don’t think is justified or sustainable. Specifically, hundreds of billions of dollars are being expended – and trillions more is projected – for the construction of exorbitantly expensive AI data centers. But AI technologies, particularly generative AIs like ChatGPT, don’t easily lend themselves to profitable business models because 1) they consistently make mistakes called “hallucinations” that reflect the inherent limitations of probabilistic systems (i.e., systems that generate responses based on statistical likelihoods rather than deterministic rules), and 2) the models can’t achieve profitable operating leverage because they generate responses to the same questions repeatedly, causing costs to increase roughly in line with usage. Conversely, businesses with strong economies of scale will see costs decline as utilization increases.

So is the wave of enthusiasm over AI going to crash like the internet and housing bubbles and bring the economy down with it? Anything is possible, but if history is a guide, probably not anytime soon. What we realized when looking at the unwinding of previous bubbles – the Roaring Twenties, the Nifty-Fifty stocks of the late 1960s/early 1970s, the Japanese asset bubble of the late 1980s/early 1990s, and more recently internet and housing bubbles – is that the precipitating factor in each case was central bank interest rate hikes (by the Federal Reserve in the U.S. and the Bank of Japan in Japan).

Speculative manias are always facilitated by easy money. When central banks raise rates they tighten monetary policy, starving investment bubbles of the liquidity needed to keep inflating. And once asset prices start to fall, a self-reinforcing downward spiral can take hold, as leveraged participants are forced to sell, liquidity dries up, and selling begets more selling.

Fortunately, inflation is currently trending in the right direction, and the Federal Reserve is likely to continue modestly lowering interest rates in 2026. This isn’t to say that the speculative fervor around AI won’t recede or that stocks in general can’t decline. But it does suggest that the kind of abrupt, liquidity-driven collapse that’s historically turned asset bubbles into full-blown bear markets and recessions is unlikely in the near term.