JPMorgan: The Key Question in AI Investment Has Shifted from "Whether to Invest" to "How to Invest"
0xBroomberg
JPMorgan strategist David Lebovitz says AI now touches nearly every asset class, shifting the core investor question from whether to participate to which link in the value chain to own — a call the recent chip-sector selloff is validating in real time.
"Everything is becoming an AI trade" — what does that actually mean?
Lebovitz's own words: "AI is everywhere — the question is how you participate, not whether you participate."
This means → AI is no longer an optional theme. It is a baseline variable running through virtually every asset class.
In plain terms = the question is not "should I buy AI?" but "which segment of the AI chain do I buy?" Pick the wrong link and the rally passes you by.
Same value chain — why does risk diverge so sharply?
Lebovitz draws a clear line: data-center buildout and operations have structural demand support; chip and hardware production face greater oversupply risk.
His words: "My supply-side concerns are mainly in chips and hardware — which is also where investor enthusiasm is most concentrated."
This means → the most crowded trade sits right on top of the highest supply risk. History shows that when enthusiasm overshoots, people tend to overshoot with it.
Why did chips crash on record-breaking earnings?
SK Hynix has fallen more than 20% from its June all-time high, despite cumulative gains exceeding 200% since the start of 2026. A planned Nasdaq listing this week was overshadowed by the AI-chip selloff.
Samsung Electronics on July 7 issued record Q2 guidance — revenue of roughly KRW 171 trillion, operating profit of roughly KRW 89.4 trillion — yet the stock dropped, dragging Korea's KOSPI index down over 8% intraday and triggering a circuit breaker.
In plain terms = classic "buy the rumor, sell the fact." Months of rallying had already priced in the full upcycle; the earnings release became a window for profit-taking.
Adding pressure, Meta's reported plan to sell surplus AI compute capacity fueled concerns that AI infrastructure demand may already be oversupplied.
What is the "earnings rift," and why does it matter?
A separate JPMorgan report notes that AI chip and memory stocks have sharply outperformed hyperscalers such as Amazon, Microsoft, and Google over the past year, widening an "earnings rift" to unsustainable levels.
This means → whether the AI rally continues now hinges less on capex size and more on whether commercial monetization can actually deliver.
In plain terms = chip stocks have risen faster than their own customers. That gap must close — either the customers catch up, or the chips pull back.
How can long-term optimism and short-term pressure coexist?
JPMorgan projects the four major U.S. cloud providers will increase 2026 capex by 80% year-on-year, topping $575 billion.
The bank has raised its 2030 global AI infrastructure spending forecast to $5.5 trillion.
This reflects a core tension: the long-term investment direction is clear, but in the short term, only investors who identify links where demand is real rather than expectation-driven will capture differentiated returns.
Content is for reference only, not financial advice.