The consensus setup for this earnings season is genuinely strong on the surface: Wall Street expects roughly 22% year-over-year EPS growth for the S&P 500. In a year that opened with recession chatter, double-digit index earnings growth is a real achievement. But the number underneath the number is the one investors should sit with — because by one estimate, AI infrastructure stocks are expected to contribute nearly 60% of that growth, and Micron and Nvidia alone account for more than 40% of it.
That reframes the whole picture. This is not a broad, healthy expansion where a rising tide lifts everything. It's a narrow one, carried by a handful of names tied to a single theme.
The math of concentration
Let me lay out what those percentages actually mean.
| Claim | Implication |
|---|---|
| S&P 500 EPS growth | ~22% YoY expected |
| AI infrastructure share of that growth | ~60% |
| Micron + Nvidia share of that growth | >40% |
| The rest of the 500 | Carrying far less of the load |
If AI infrastructure is ~60% of the growth, then the other 490-plus companies — every bank, retailer, industrial, healthcare and consumer name combined — are contributing the minority of the index's earnings expansion. And within the AI slice, two companies do most of the lifting. The market's earnings story has rarely rested on so few shoulders.
Why this is both bullish and fragile
The bullish read is straightforward and I don't dismiss it. The AI buildout is producing real revenue — TSMC's sales are accelerating, Morgan Stanley's markets desks are busy, the capex is showing up in actual foundry and memory numbers. If that continues, these companies genuinely earn the growth, and 22% is achievable and repeatable.
The fragile read is the mirror image. When a majority of index earnings growth depends on one capex cycle, the index inherits that cycle's risk. A single disappointing guide from a hyperscaler, a pause in AI spending, or a stumble at Nvidia or Micron doesn't just dent two stocks — it undermines the earnings thesis for the whole market. Diversification at the index level is partly an illusion right now: you can own 500 companies and still be making one concentrated bet.
What actually reduces the risk
The healthy development to watch for is breadth — evidence that earnings growth is spreading beyond the AI complex. This week's bank results help: a trading boom and a 50%-plus recovery in investment banking is real, AI-independent earnings growth. If financials, industrials and consumer names start carrying more of the load, the 22% becomes far more robust because it stops depending on one theme.
That's the tell I'm watching this season. Not whether Nvidia beats — it probably will — but whether the other 490 companies show enough growth that the index isn't a single-theme trade wearing a diversified costume.
My take
I hold two ideas at once here. The AI-driven earnings growth is real, not imagined — the revenue is landing in actual results. And the market's dependence on it is a genuine concentration risk that a lot of investors are underpricing precisely because the index looks diversified.
The practical conclusion isn't "sell everything." It's to know what you actually own. If your portfolio is the index, understand that you are heavily long one capex cycle. Balancing that with earnings streams that don't depend on AI spending — the reopening capital-markets trade, quality cyclicals, cash-generative businesses at reasonable multiples — is how you keep the upside without betting the whole thing on a single theme holding.
Bottom line: 22% expected EPS growth is genuinely strong, but with ~60% of it riding on AI infrastructure and 40% on just two stocks, the market's earnings story is narrow. The bullish and fragile cases are the same fact — respect both.
Not investment advice.