The AI Divide: How the AI Economy Is Splitting In Two
Listen to the audio version of this article (generated by AI).
Last Sunday night, markets were bracing for the worst.
Oil spiked while futures dropped amid headlines that screamed “escalation in the Middle East.” In a nut shell, the macro narrative looked fragile, possibly even broken…
And then earnings week started.
What we got wasn’t confirmation of a weakening economy.
We got proof that there are two economies now… and they’re moving in completely different directions.
For months, we’ve argued that the global economy is quietly splitting in two.
On one side: companies building and powering the artificial intelligence revolution.
On the other: everyone else.
Last week, that idea started showing up consistently and overwhelmingly in earnings.
To understand what’s happening, we need to zoom out…
The Winners: A Stack-by-Stack Victory Lap
The first thing you need to understand is what “AI infrastructure” actually means.
Building an AI data center is not just buying a bunch of Nvidia (NVDA) GPUs and plugging them in. It requires a complete physical stack: specialized chip-making equipment, memory chips, processors, power management semiconductors, optical fiber connections, and — at the very foundation — physical buildings filled with HVAC systems, electrical wiring, and cooling infrastructure.
Each layer of that stack reported earnings this past week and every single one broke records.
- Chip equipment firms like ASM International (ASMIY) and Lam Research (LCRX) reported surging demand and backlogs stretching years into the future.
- SK Hynix — the dominant supplier of high-bandwidth memory — posted revenue up nearly 200% year-over-year, with margins that more closely resemble a luxury goods stock than semiconductors.
- Texas Instruments (TXN), the most diversified chip company in the world, saw data center revenue jump 90%.
- Intel (INTC) revealed that AI now drives 60% of its business and is accelerating.
- Optical and connectivity players like MaxLinear are seeing triple-digit growth tied directly to expanding AI clusters.
- Power and thermal infrastructure providers like Vertiv (VRT) and GE Vernova are raising guidance as hyperscalers build out capacity at unprecedented scale.
- And at the ground level, Comfort Systems (FIX) — the company physically constructing data centers — is growing revenue more than 50% with a record backlog that keeps rising.
Translation: buyers are no longer negotiating price. They’re just asking for volume. That is a strategic reorientation.
“Customers are prioritizing securing volume over pricing, which is sustaining the current strength.” — SK Hynix CFO, Q1 2026 Earnings Call
This is synchronized strength across an entire industrial ecosystem.
And most importantly, demand is outrunning supply… everywhere.
Customers aren’t negotiating price anymore.
They’re fighting for volume…
The Losers: A Guided Tour of the Other Side
Now let’s cross the divide.
Consumer-facing and non-AI-exposed companies told a completely different story.
- Packaging giant Sonoco (SON) cut guidance as energy and supply chain costs rise.
- United Airlines (UAL) warned that higher oil prices are squeezing margins and weakening demand.
- ServiceNow (NOW) flagged longer sales cycles as enterprise buyers hesitate.
- Pegasystems (PEGA) dropped to new lows after weak results.
- Lululemon (LULU) replaced its CEO amid slowing growth.
These businesses are tied to consumer demand, input costs, and discretionary spending cycles.
And right now, those forces are soft.
Why This Is Structural
Skeptics will argue that this is just a boom cycle… that AI infrastructure spending will eventually slow, that the winners will give back their gains, and that the bifurcation will narrow as the economy normalizes. History suggests otherwise, and the earnings data makes the case explicitly.
Consider what we heard from management teams across the winning cohort: SK Hynix said customers’ demand for memory over the next three years “far exceeds current supply capacity.”
Intel said server CPU demand has “improved over the last 90 days” and momentum extends into 2027.
MaxLinear said backlog is already building into 2027.
Comfort Systems said their backlog rose sequentially even as their burn rate accelerated — meaning they are signing new contracts faster than they can complete existing ones.
These are not companies talking about a hot quarter. These are companies describing structural supply constraints that are measured in years.
The reason is simple: the physical infrastructure required to run AI at scale — fabs, fabs equipment, memory capacity, power plants, fiber networks, data center buildings — takes two to five years from initial investment decision to operational capacity. The demand is here today.
The supply is not.
That gap does not close quickly, which means the pricing power, the margin expansion, and the earnings growth will persist far longer than a traditional cyclical recovery would suggest.
There is also a demand-side structural argument that emerged clearly from the Intel and SK Hynix transcripts this week.
AI is evolving from the training phase — where massive models are built in specialized data centers — into the inference and agentic phase, where AI actively processes real-world requests at scale, continuously, across millions of concurrent users.
Every agentic AI task generates intermediate data that must be stored and processed. Every inference call requires a CPU to orchestrate it alongside a GPU to compute it. The more AI is deployed, the more of every layer of the infrastructure stack is required. The demand loop is self-reinforcing.
What This Means for Your Portfolio
The bifurcation thesis has now graduated from analytical framework to documented business reality.
It is showing up in quarterly revenue lines, gross margin percentages, backlog figures, and earnings-per-share numbers across the entire AI infrastructure stack simultaneously. That is not a coincidence. It is a structural shift in where economic value is being created.
The companies building, powering, connecting, and computing for the AI era are operating in an environment of structurally constrained supply against structurally growing demand… the most favorable economic condition that exists for sustained earnings growth.
The companies outside that world are operating in an environment where consumer confidence is shaky, input costs are elevated, and discretionary spending is vulnerable.
The data has spoken. The AI divide isn’t coming….
It’s already here.
If you want the picks-and-shovels play on the most ambitious financial product launch of our generation, you need to see our latest research.
I just launched a full presentation on X Money – Elon Musk’s long-awaited attempt to turn X into the “Bank of Elon” – and the public companies powering it behind the scenes.
In it, I break down the full story on how you could benefit as X Money rolls out.
Click here to watch it now.
