Santoli: The youthful phase of AI is over. What it means for investors

"We'll laugh again," answered the JFK aide Daniel Patrick Moynihan. "It's just that we'll never be young again."

At the risk of seeming to conflate a president's assassination with an investment motif, this is where the AI theme sits right now. 

There will be cheer and thrills and profits generated by the burgeoning technology in the years to come. But the youthful phase of AI - when promise was unlimited and winners were easy to spot and vastly outnumbered losers - is over. 

This reality predates the past week's litany of ominous pronouncements about the destructive potential of AI models and the industry's professed desire for some restraint on the velocity of development. 

Semiconductor shares were already 20% off their June highs prior to Monday's 4% slide. The tech sector of the S&P 500 has seen its forward price/earnings multiple contract from 29 to 21 in the past year with the market unwilling to extrapolate the profit surge in the absence of a clear path back to reaping oodles of free cash flow. Data centers have a lower approval rating than Congress – to the point where it's hard not to assume the overheated ire will burn itself out after the election.  

Perhaps not much will change in terms of AI policy, model training or the demand for computing capacity. But in markets, narratives matter quite a bit in the near term. 

And for the moment, the AI narrative has become more fixated on risks than opportunities. Right or wrong, the breathless talk of throttling the buildout makes it marginally tougher to argue that 2027 earnings projections – enormously dependent on semiconductors – are too low. 

And anyone already worried that the eyelash-singeing profit growth of the second quarter represented plenty of pulled-forward demand and temporary margin expansion will find only their bias confirmed. 

AI is unpopular with the public, projections of order backlogs beyond a few quarters are now suspect and investors might be dialing back their assessment of how quickly a whole class of entrenched businesses will be displaced by the automated hive mind. 

There are positives here. If a year ago the most-cited concern was that an AI bubble was forming and speculative exuberance was frothing over, that's been taken care of. The savage purge since June of high-momentum stocks – dominated by AI hardware and related industrials – has sapped the bullish aggression of traders, helping to reset tactical sentiment toward a more neutral level. 

Monday's perky action in shares of the huge tech platforms doing most of the AI capex – Microsoft, Alphabet, Meta – reveals a silver lining from a possible slowdown, allowing these companies to take a beat before accelerating their capital raising and deployment further still. 

The market has already shown a voracious appetite for companies demonstrating copious free cash flows at a time when it's become scarce. Here's the VictoryShares Free Cash Flow ETF, which for two years tracked the Mag 7-dominated S&P 500 until leaving it in the dust the past few months. This ETF is jammed with software, healthcare and energy stocks that look cheap, at least cosmetically, on FCF yield metrics.

Any dialing-down of "AI maximalist" speculative fervor also prevents the current cycle from building excesses that would merit straight comparison with those of the end of the 1990s tech boom. 

Which helps, because from a macro perspective, the echoes of 1999 continue to redound in the background. Consider:

The Federal Reserve is now poised this week to begin lifting interest rates as 10-year yields top a boldfaced threshold (5%) and a tech-capex boom threatens to overheat parts of the economy, having cut rates three times late last year to counter an economic-growth scare as inflation ebbed. In June 1999, the Fed began hiking rates as 10-year Treasurys surged to a key round number (6%) and a tech-capex boom raged, having cut rates three times late the prior year in response to a deflationary growth scare. 

The Fed would hike a full percentage point into early 2000, when Fed Chair Greenspan testified that he was surprised the economy hadn't slowed much. He would jack rates another 75 basis points as stocks would peak before a lost decade and the economy slid toward recession. 

I prefer to avoid suggesting this cycle will hew too closely to that one, doubting that the exuberance will carry so far in quite the same manner. But that doesn't mean a "lower extreme" won't be reached this cycle, ahead of a payback phase. 

Many observers invoke the '90s pattern to suggest it's still "early" in the AI bull trend. Bespoke Investment Group has plotted the Nasdaq Composite since ChatGPT was released against the index following the introduction of Netscape in 1994. It's tracked pretty well, coincidence or not. I've always thought that was too forgiving, given that tech was already ascendant to a degree it wasn't in 1994 and the market enthusiasm for AI kicked in instantly, while it was a slower boil after Netscape's 1995 IPO.

Evercore ISI strategist Julian Emanuel has gone so far as to cast the June 2026 IPO of SpaceX as akin to the Netscape IPO rather than as a culmination of AI enthusiasm that coincided with a severe tech hangover since.

One bullish argument for the market's ability to shake off 5% 10-year Treasury yields is that the 1999 bubble ramp happened with yields between 5-6%.

This is true, but not a one-for-one match. Yields then were down from 7%-8% five years earlier, rather than up from 1% over the past five years. And the '99 bubble also happened with corporate tax rates at 35% against 21% now, and with retail-brokerage firms charging $15 commissions per trade. 

Would today's market avoid a hiccup if we quickly reverted to those conditions?