Safety is the only thing that’s worked in the stock market

For the better part of the last three years, AI has been the jet fuel propelling the stock market to ever-greater heights.

2026... not so much. And that’s created a unique setup where the stock market is still within a stone’s throw of all-time highs, yet appears very vulnerable under the surface. Safety is really the only thing that’s worked this year. And when confronted with the arrival of new AI tools that may alter the long-term outlook for various stocks and sectors, investors have taken a shoot first, ask questions later approach.

Call it agor-AI-phobia: the threat of AI disruption has been a rolling thunder sweeping across swaths of industries. Most notably, software stocks have come under the knife, but other seemingly more insulated sectors like commercial real estate and even trucking stocks have tumbled with AI cited as the proximate cause, or, at least, the excuse.

Safety first

Capital-light stocks (which describes most of the software cohort) have seen their valuations come in sharply relative to firms with elevated capital outlays:

But we also know that the biggest spenders — the Magnificent 7 hyperscalers — by and large aren’t getting rewarded for their massive capex budgets either. On the contrary, Microsoft and Amazon are the biggest drags on the SPY year to date. De-rating and selling hyperscalers implies doubt as to whether this capital spending will be worth it.

The biggest line item in their data center build-outs is the IT infrastructure — in particular, chips. And the company that’s nearly synonymous with the AI boom, Nvidia, isn’t benefiting either.

And yet, the S&P 500 is less than 2% from its record closing high, despite Nvidia and these aforementioned hyperscalers being its largest components.

How does this happen? Well, to oversimplify, the flip side of this is that investors have bid up safety and high earnings visibility (which is, in itself, kind of a derivative of safety, if you think about it!).

Look at the two biggest components of the Consumer Staples Select Sector SPDR Fund, a notoriously defensive sector:

Costco — which unlike software, boasts a recurring revenue model that AI can’t disrupt — trades at a forward price-to-earnings ratio of nearly 48x, up from 41x at the end of 2025. For Walmart, that’s risen to nearly 45x from 38x. These companies trade at nearly double the multiple of the average Mag 7 hyperscaler or Nvidia!

Memory stocks represent the other key source of market support, thanks to intense shortages that have given major suppliers immense pricing power.

To a lesser extent, semicap companies like Applied Materials, which just reported a “narrative-changing quarter,” and industrials levered to the data center build-out, such as Caterpillar, are a part of the same theme.

If there’s an AI bubble, it is arguably much more in real economic activity than it is in financial markets. Investors are willing to bet narrowly on the profits provided by the AI capex, not the potential returns from this spending. That’s the opposite of the “extrapolative expectations” that defines investor behavior during bubbles.

*Screams internally*

The price action in individual stocks has been anything but normal even as the benchmark US stock has gone nowhere in 2026. Large-cap stocks are behaving more like the stock market is deep in a bruising bear market rather than close to all-time highs:

For portfolio managers, a world where their up days are immense and their down days are terrible is not a world where you want to be running with higher leverage or gross exposure. Volatility is not just an output of price action, but an input for positioning.

And by all accounts, positioning coming into this year was so elevated that there was little in the way of dry powder to put to work.

A market in which the individual components are going haywire becomes much more vulnerable to a more significant decline in the event that there’s a common cause for them to move together.

The bad news is the good news

That being said… if you squint, all of the above also helps inform the bull case.

Since the start of 2020, the only events that have sparked a meaningful, sustained pickup in cross-asset correlations have been seismic macroeconomic events: the onset of the pandemic, generationally high inflation, and the announcement of a tariff regime that threatened to redefine the nature of cross-border commerce.

Again, we’re still less than 2% from all-time highs in a world where all of the Magnificent 7 are down on the year.

Profits are growing. AI disruption is still more of a threat than a reality for most major incumbents, and slower inflation, if sustained, may provide a window for rate cuts without requiring economic weakness,

If a desire to seek hidey-holes has left us here, imagine what could happen if traders arrive at the same calculation as tech CEOs: that the risk of underinvesting in AI is greater than the risk of being too exposed.