The S&P 500's gain over the past 18 months has been driven roughly half by multiple expansion and half by earnings growth. That ratio is now flipping. With valuations near the top of their historical band, the next leg has to come from earnings. That puts the next set of mega-cap prints — Nvidia above all — into outsize importance, because the market is no longer pricing them as a single-company story but as the proof point for the AI capex thesis.
Key takeaways
- Valuation tailwinds have largely been spent — further multiple expansion looks limited from here.
- Earnings growth must now do the work, and the burden is concentrated in a small handful of names.
- Nvidia's print is both the AI demand signal and the AI capex confirmation for the rest of the market.
The math of where we are
The S&P 500 trades at a forward earnings multiple comfortably above its 30-year median. From here, multiple expansion is unlikely to be a major contributor to returns unless something unusual happens — a sharp drop in long rates, a structural change in equity-risk premium, or both. Most of the work, therefore, has to come from the "E" rather than the "P/E."
| Return component | Last 18 months | Next 12 months (likely) |
|---|---|---|
| Multiple change | Roughly half of total | Small or negative |
| EPS growth | Roughly half | The dominant contributor |
| Dividends | Small | Small |
Why Nvidia is the linchpin
Nvidia is the proxy for two questions at once. The first is operational: are the largest customers — hyperscalers, AI labs, sovereigns — continuing to grow their orders, or are they pausing to digest the existing build-out? The second is strategic: do those buyers believe the unit economics of frontier AI workloads support the trajectory of their own capex plans? Nvidia's commentary on backlog, deferred revenue and customer concentration speaks to both questions simultaneously.
When one quarterly print is doing double duty as the demand signal and the demand explanation, the bar for a clean beat is higher than the headline numbers suggest.
What investors are looking for, in order of importance
- Forward guidance. A reaccelerating top-line outlook would calm fears that the AI capex cycle is mid-rolling-over.
- Customer concentration. Disclosure on top-customer share is a barometer of how broad the demand really is.
- Backlog and shippable supply. Backlog growth versus shipped revenue indicates whether constraints have eased.
- Gross margin trajectory. The product mix into next-generation parts versus mature ones will move margin within a wider band than usual.
What retailers are telling you that Nvidia isn't
The same week's retailer prints are a useful cross-check. If the consumer side weakens noticeably while AI capex stays strong, the market gets a clean rotation story — capex-led growth offsetting consumer softness — that supports a continuation of recent leadership dynamics. If both sides soften, the question of whether earnings can do the work becomes harder.
The retailer signal to watch
- Same-store-sales trajectory, particularly in discretionary categories
- Gross margin compression vs. promotional intensity
- Inventory days relative to last year
- Forward guide for the back half of the year
FAQ
Is the market overly dependent on a few names?
Index concentration is at a multi-decade high. That is not, by itself, a sell signal, but it does mean idiosyncratic prints have unusually large index effects.
What would actually break the rally?
The most likely catalyst is a credible signal that AI capex growth is slowing while consumer demand is also weakening — i.e., both halves of the earnings story turning down together.
Are these dynamics permanent?
No. Earnings concentration has reversed in every prior cycle and will eventually reverse in this one. The question is whether reversal happens via the leaders' growth slowing or via the laggards catching up.
The bottom line
The market has consumed most of the easy fuel from multiple expansion. From here, earnings have to deliver, and one print is doing more work than any single report should reasonably be expected to do.
