Wall Street's telling investors to stop betting on one horse in a two-race track.
The Summary
- JPMorgan's head of alternative investment strategy says AI is running two adoption cycles simultaneously, creating volatility that's actually a feature, not a bug
- The current tech volatility is "healthy" according to Sundar, who advises spreading bets across the entire AI value chain
- Translation: the infrastructure buildout and the application layer are maturing at different speeds, and smart money isn't picking sides
The Signal
Sitara Sundar, head of alternative investment strategy at JPMorgan Private Bank, is watching something unusual. AI isn't following the clean adoption curve of previous tech waves. Instead, two cycles are overlapping. The first is the infrastructure layer: chips, compute, training platforms. The second is the application layer: agents, enterprise tools, consumer products. Normally these phases run sequentially. Right now they're colliding.
This collision explains the whiplash in tech markets. Infrastructure stocks surge when investors bet on foundational buildout. Application stocks pop when a new use case proves commercial viability. The volatility isn't confusion. It's the market trying to price two different timelines at once.
"Diversifying across the AI value chain is the best strategy for investors."
Sundar's advice: don't concentrate. The traditional tech playbook says pick the winner early and ride it. But when two cycles run in tandem, picking winners means picking *which cycle* matters more. Get that wrong and you miss half the story. JPMorgan is telling clients to own the full stack. Chip makers and model builders and application developers. The picks-and-shovels and the gold miners.
What makes this moment different from every other "diversify your tech bets" pitch is timing. We're not in the hype phase where everything AI-adjacent gets funded. We're in the separation phase. Some infrastructure companies are printing cash. Some application startups are proving unit economics. The weak hands in both categories are getting flushed out. That's the "healthy tumult" Sundar described. Markets are doing what they're supposed to do: testing which parts of the stack have durable value.
Key dynamics at play:
- Infrastructure companies face questions about utilization rates and whether demand justifies buildout
- Application companies face questions about whether AI features create defensible moats or commoditize fast
- Both cycles are real, both are early, neither is guaranteed to follow the path investors expect
The Implication
If you're building in AI, this framing matters. The infrastructure players and the application builders aren't competing. They're co-dependent variables in the same equation. A breakthrough in model efficiency changes the economics for everyone downstream. A killer app that proves retention creates demand that justifies upstream investment. The tumult Sundar sees isn't noise to filter out. It's the system finding equilibrium between two feedback loops.
For allocators and operators, the takeaway is the same: thesis diversity beats conviction trades right now. The winners won't all come from one layer. And the market's trying to tell you that in real time.