Twilight of the AI Bull?

Is the red-hot artificial intelligence investment theme stumbling? Even AI cheerleader Sam Altman, the head of OpenAI, sounded a warning: “Are we in a phase where investors as a whole are overexcited about AI? My opinion is yes.”

Even as the tech-heavy NASDAQ 100 grinds upward to new highs, it’s showing signs of punching below its weight. Its relative performance (middle panel) is exhibiting lower lows and lower highs. Historically, NASDAQ 100 relative performance has been inversely correlated with the 10-year Treasury yield because of its high-duration characteristic, but it hasn’t been able to lead the market despite weakness in the 10-year yield. The weakness in relative breadth (bottom panel) is another warning of stumbling relative strength. All of this is occurring against a backdrop of a rollover in Bitcoin prices, indicating a retreat in the market’s animal spirits.

From a top-down perspective, technology stock valuations appear bubbly. U.S. tech valuations have soared to levels not seen since the dot-com bubble.
 

 

An article in the Economist studied past investment bubbles, from the railroad manias of the 19th Century, to the electric light bubble of the early 20th Century, and the dot-com bubble of the 1990s. Three main factors affect the aftermath of bubbles. The first is the spark. Political sparks, such as easy money and loose lending policies, lead to gargantuan busts like the Japan property bubble of the late 1980s. By contrast, the aftermath of technologically driven bubbles tends to hurt less, though the damage can be serious.

The second factor is the size and durability of the capital investment. Large-scale buildouts of infrastructure can leave white elephants, but they will be absorbed over longer periods.

The third factor is who bears the losses. The dot-com bubble was characterized by individual investors stampeding into tech stocks, but the boom wasn’t accompanied by leverage. The lack of excessive leverage meant that the stability of the banking system wasn’t threatened, which lessened the damage.

Putting it all together, the Economist concluded that, based on its historical study of investment bubbles, “the potential AI bubble lags behind only the three gigantic railway busts of the 19th Century.”
 

 

On the other hand, it could be argued that AI investing is entering its second phase. The gains from large language models are running into the law of diminishing returns. Foundational model buildout is nearing a saturation phase, much like past innovation booms like railroads in the 19th Century and fibre optic capacity in the late 1990s.
 

 

A separate article in the Economist highlighted the gains from small language models (SLMs) which does not require the training capacity of LLMs. SLMs are more useful for use by AI agents focused on specialized tasks.
 

 

Where’s the Leadership?

As NASDAQ stocks gradually give up their leading relative strength status, it’s unclear who the next leaders will be. Breadth hasn’t significantly broadened out. The equal-weighted S&P 500 is lagging and mid- and small-cap stocks have shown limited signs of recovery.
 

 

While SMID relative performance appears mildly reassuring, mid- and small-cap Advance-Decline Lines are exhibiting negative divergences against their respective indices.
 

 

If tech leadership stumbles, a welcome bullish development would be the emergence of cyclical stock leadership. But cyclicals are weakening relative to the S&P 500.
 

 

 

A Time for Caution

These conditions make me nervous about the near-term outlook for U.S. stock prices. Technical analyst Wayne Whaley studied the history of summer returns, defined as the returns of the S&P 500 over the three summer months of June–August, which rose 9.3% in 2025. Past instances of 5% summers saw marginally positive September returns, with a win-loss rate of 14-10, followed by weak Octobers, with a win-loss of 6-18, and rebounds in November and December.
 

 

Looking to the week ahead, the highlight will be the FOMC rate decision on Wednesday. The Fed is highly likely to cut rates in response to employment weakness, but the accompanying chart also highlights the Fed’s inflation dilemma. Core goods inflation has been rising in response to the imposition of tariffs. While the Fed may choose to “look through” the one-time tariff-related price increases, core services CPI, which is not subject to tariff pressures, has stopped falling and stabilized at 3.6% — well above the 2% inflation target. This raises the risk of a hawkish cut, where the Fed cuts rates but signals that the cut was a precautionary move in response to labour market weakness, but it will remain vigilante on inflationary pressures.

Even though inflation pressures may moderate in early 2026, they are likely to rise for the remainder of this year. Equity investors will then be faced with a similar dilemma. Either tariff inflation takes hold, which puts rate cuts on hold, or tariff pass-through is less than expected, which implies margin compression as importers absorb tariff costs. Neither outcome is equity bullish.
 

 

Regardless of how the Fed positions its anticipated rate cut next week, the market is setting itself up for disappointment. The market consensus expects the Fed Funds rate to fall below 3% in 2026, which is an extremely ambitious objective in light of the stubborn reaction of core services CPI, which is not subject to tariff effects and stabilized at 3.6%.
 

 

In conclusion, the leadership of AI-driven stocks is starting to stumble from bubbly valuation levels, which brings up the warning from Bob Farrell’s Rule #4: “Exponential rapidly rising or falling markets usually go further than you think, but they do not correct by going sideways.” While the debate is ongoing as to whether the AI bull is evolving from hyperscaler leadership to the next phase of companies that can better exploit the technology, the lack of cyclical market leadership is concerning from a technical perspective. I am therefore tactically cautious about the short-term outlook for U.S. equities.