JPMorgan Says Everything Is an AI Trade: ‘It’s How You Play It’

JPMorgan Asset Management’s David Lebovitz says investors are improving their AI stock-selection process by separating pockets of potential excess supply from areas with robust demand, moving away from a blanket “buy everything AI-linked” approach. The commentary suggests a more selective, less speculative positioning as AI becomes an “everywhere” trade, with likely modest positive read-through for properly valued AI exposures.

Analysis

The bigger market signal is not “buy AI,” it is that the trade is entering a dispersion phase. That usually compresses the multiple of the generic basket and widens the spread between firms with visible monetization versus those leaning on narrative or optionality. In the next 1-3 months, that should hit the most crowded semiconductor and software proxies first, because passive flows have less tolerance for “AI adjacency” once investors start asking what the revenue conversion looks like.

JPMorgan is a relative beneficiary of that shift because a selective AI tape tends to increase demand for financing, treasury, hedging, and capital-markets plumbing around data-center buildouts and M&A among second-tier winners. The less obvious winners are power, networking, and infrastructure names tied to real deployment rather than model-training headlines; the losers are companies where AI is a marketing label rather than a margin driver. If the market starts rewarding cash flow over capex growth, that also supports financials over high-multiple infrastructure plays.

The contrarian risk is that this “discrimination” phase is premature if hyperscaler capex re-accelerates or if a new model cycle renews broad risk appetite. In that case, shorting the crowded AI basket gets dangerous fast because the market will revert to buying duration and revenue optionality. The thesis is falsified if the next round of cloud-provider guidance shows another leg up in AI spend or if AI monetization surprises materially to the upside over the next 1-2 earnings cycles.

Key Decisions for Investors

• Long JPM vs short SMH for 1-3 months: express the idea that capital allocation quality and financing activity will outperform the crowded AI hardware basket if the market keeps pruning weak AI names. Stop-loss if AI capex commentary from hyperscalers re-accelerates materially.
• Trim or hedge broad AI-beta exposure in QQQ/SMH on strength rather than chasing the basket: the setup favors dispersion, not blanket exposure. This is a risk-management trade, not a conviction short.
• Watchlist long in power/infrastructure beneficiaries such as ETN over the next 1-3 earnings cycles: the market may rotate toward picks-and-shovels with visible order books if AI spend becomes more utility-like and less speculative.
• If you need a contrarian hedge, buy downside protection on the most crowded 'AI narrative' names that lack clear revenue proof over the next quarter; the market is likely to punish story stocks first when differentiation rises.