J.P. Morgan: July's AI Sell-Off Was a Valuation Reset, Not a Story Collapse. Earnings Grew 45%.

The bigger question hanging over markets in July was whether the AI spending boom actually pays off. J.P. Morgan Asset Management's July market review, published August 6 by global strategist Kerry Craig, says the answer so far is yes, even after a rough month for chip and memory stocks.

Craig's team frames July as two separate stories that got tangled together. One was a fast reassessment of AI capital spending that rattled the whole supply chain, as investors started asking hard questions about return on investment. Open-weight models are getting cheaper and more competitive, token costs are rising for enterprise customers, and Wall Street wants proof that all this spending translates into productivity gains, not just bigger data centers.

The second story was old-fashioned geopolitics. Renewed fighting in the Middle East pushed oil prices higher during July, according to the report, and that's what actually spooked central banks. Developed-market policymakers had been sitting on soft June inflation readings, but energy-driven price pressure changed the calculus. Markets are pricing in more policy tightening by year-end as a result.

That separates two different risks that have been getting mashed together in market commentary lately. A reassessment of AI capital spending and crowded chip-stock positioning is a valuation story about the tech supply chain specifically. Oil-driven inflation forcing central banks to stay tighter for longer is a completely different mechanism, and J.P. Morgan's report treats it as the more immediate threat to the rate outlook, given the risk it poses to the broader disinflation trend.

On the AI valuation question specifically, Craig makes a case worth taking seriously: the semiconductor and memory names that ran hardest earlier in the year were also the most crowded trades on the Street. A reset in performance after a big run isn't automatically evidence the underlying business is broken. It can just as easily be a move away from crowded positioning and a resetting of expectations rather than a structural shift in the AI story. J.P. Morgan's framing is that this looks more like the latter than the former, at least based on what showed up in second-quarter numbers.

And the numbers are the strongest part of the argument. U.S. second-quarter earnings growth came in at roughly 45% year-over-year, according to the report, more than double the 22% consensus that Wall Street was working with when earnings season kicked off. That's a massive beat. If AI capex were actually failing to generate returns, you'd expect earnings to be disappointing across the board, not blowing past expectations by more than double.

But the report doesn't wave away the skeptics either, and it shouldn't. Companies that missed on revenue experienced sharper drawdowns on the day following their earnings announcement than those that missed on earnings. That's a market that has gotten far less patient with any hyperscaler or chipmaker that can't show the growth is real and durable. J.P. Morgan's own language is blunt on this: capex plans "will need to be justified not only by future demand, but also by credible evidence of monetisation." In plain terms, spend the money, but prove it's working, or the stock gets punished.

That's the fair skeptic's case, and it's not a fringe view. Investors who watched hyperscalers commit huge sums to AI infrastructure, only to face rising competition from cheaper open-weight models and climbing per-token costs for customers, have a legitimate reason to demand proof of payoff before extending more credit to the story. The market's harsher treatment of revenue misses last month backs that up. This isn't blind AI doubt, it's investors pricing in execution risk on a genuinely enormous capital bet.

Australia was the quiet outperformer in July, according to Craig's report, with the Reserve Bank of Australia holding rates steady and easing local inflation supporting Australian equities, while investors globally became more selective and broadened their focus beyond concentrated tech bets. That diversification trend is itself a signal that some of the market's confidence in the AI trade specifically, as opposed to equities generally, is being hedged.

None of this settles the argument. J.P. Morgan's own report states that the fundamentals "have not deteriorated to the same degree as sentiment," which is a careful way of saying the gap between what investors fear and what companies are actually delivering hasn't closed. The next test comes with third-quarter earnings, when investors will find out whether 45% growth was a peak or a new baseline, and whether the companies that missed revenue in July can turn that around before markets lose patience for good.