JPMorgan: AI Capex Spiraling Out of Control — META and Microsoft Could Emerge as Winners
Taylor Wilson
JPMorgan TMT analyst Schilsky argues that semiconductors and internet/software stocks now move in opposite directions almost daily, yet META and Microsoft may be the first to flip from 'AI losers' to winners — if their capital spending can deliver visible revenue returns.
What is the market arguing about? Why are semis and internet stocks moving in opposite directions?
The tech market is sharply divided: semiconductor companies are priced as AI winners, while internet and software firms are tagged as AI losers. The two groups move inversely almost every day.
The core investor question has shifted — no longer "how far can AI go?" but "when does the party end?"
This means → The market is not bearish on AI itself. It is pricing who gets paid back first. Semis sit close to hardware and have clearer monetization, so they get bought first; software and internet companies are still spending without proven returns, so they get sold.
How do you tell if the party is ending? Two leading indicators
Schilsky flags two signals: ① whether the second derivative of annualized revenue growth at major AI labs — the acceleration of the acceleration — hits an inflection point; ② whether Nvidia's Blackwell architecture can truly validate AI scaling laws (the empirical rule that bigger models trained on more data keep getting more capable).
In plain terms = The first indicator asks "is AI companies' revenue growth still speeding up, or starting to slow down?" The second asks "does spending more on bigger chips still produce a step-change in model capability?"
He believes Blackwell can deliver a step-function leap in large-language-model capability. Multiple AI labs are expected to release major model upgrades before year-end — "these upgrades must be impressive enough."
META is already making money from AI — so why is it labeled a loser?
The irony is hard to miss: META's core ad business is deeply AI-driven. Q1 revenue grew 33% year-over-year — a pace difficult to achieve without AI.
What truly unsettles investors is the Meta Superintelligence Lab (MSL), which keeps consuming massive opex and capex while generating negative returns. Zuckerberg currently has "very little to show for it."
This means → The market labels META a loser not because AI failed to boost its revenue, but because its new AI spending has no visible payoff. With the stock below $600, investors increasingly fear MSL is turning into another FRL — Reality Labs, Meta's long-loss-making metaverse division.
What does META need to show to flip the narrative?
Schilsky identifies four breakthrough directions: consumer agents, AI tools for advertisers and creators, entry into the search-ad market, and enhanced Meta Ray-Ban features.
He views the September Meta Connect conference as a potential product-launch window.
In plain terms = META has several cards to play, but it must show something that makes money in at least one of these lanes — otherwise the market will keep pricing MSL as a bottomless pit.
Why might Microsoft flip first?
Schilsky sees Microsoft as the most likely first application-software company to complete the "AI loser → winner" transition. The key shift: after Q3 earnings, investors now weigh M365 commercial-cloud revenue growth as heavily as Azure.
M365 commercial-cloud growth accelerated from 14% in Q2 to 15% in Q3 (constant currency). If subsequent quarters keep accelerating and are clearly Copilot-driven, Schilsky expects "a meaningful re-rating of the stock."
On Azure, he forecasts revenue growth will mechanically accelerate into the mid-40% range as data-center capacity ramps. This reflects a clearer AI monetization path than META's: one track is enterprise Copilot subscriptions, the other is Azure compute rental — and both are already accelerating.
So what is the core verification checkpoint?
For META: whether MSL can deliver a monetizable product at the September Meta Connect conference determines if the market keeps treating it as "another FRL."
For Microsoft: whether Copilot achieves enterprise-scale monetization is the make-or-break node for its valuation thesis.
This means → The flip window for both companies falls within the next two quarters — what matters is not the story, but the numbers.
Content is for reference only, not financial advice.