The Math Behind the Multiple
The S&P 500 closed Friday at roughly 7,475. That is its eighth consecutive winning week, the longest streak since December 2023.
The Dow hit a record above 50,500. The Nikkei breached 65,000 for the first time on Monday, surging nearly 3% on Iran deal optimism. Taiwan's Taiex hit its own all-time high. Global equities are moving in one direction, and that direction has a price tag.
The trailing price-to-earnings ratio on the S&P 500 sits above 27 times operating earnings, depending on the calculation method. Some measures using GAAP earnings push it above 31. The 10-year average is roughly 23. The long-term average going back to the 1950s is around 17.6.
But the number that matters most right now is the Shiller CAPE ratio. It sits near 41. In 140 years of U.S. market history, it has only been higher once: December 1999, at 44.19. That month marked the exact top of the dot-com bubble.
This is not a prediction. High CAPE readings have persisted for years before mean-reverting. But it does set the terms of the debate. At 41, the earnings yield is roughly 2.4%. That is what the index offers in cyclically adjusted real return if profits simply hold steady. The 10-year Treasury yield is around 4.56%. The 30-year is above 5%.
For the first time in over a decade, bonds are offering competitive real returns. Stocks need to earn their premium.
What the Forward P/E Actually Requires
The forward P/E near 25 times looks more reasonable than the trailing number. But it contains an aggressive assumption.
JPMorgan recently raised its 2026 S&P 500 EPS estimate to $330, which represents 22% year-over-year growth. Their 2027 estimate jumped to $385, implying another 17% on top of that. Goldman Sachs is more conservative, projecting 12% growth for 2026 and 10% for 2027. FactSet's bottom-up consensus falls closer to JPMorgan's range.
The gap between Goldman and JPMorgan is not trivial. At Goldman's 12% growth rate, the forward P/E closer to the current level suggests the market is fully valued. At JPMorgan's 22%, there is still room. The difference is almost entirely explained by Nvidia and the AI capex cycle.
Nvidia alone contributed more to S&P 500 earnings growth in Q1 than the entire energy sector. Apple and Microsoft together account for roughly 14% of the index weight. When you strip out the top 10 names, the remaining 490 stocks trade at a far more modest multiple. The equal-weight S&P 500 has been outperforming the cap-weighted version recently, which tells you the market is starting to broaden. But the index-level valuation still depends heavily on a handful of names delivering.
Three Tests This Week
The Early Start email this morning laid out the individual catalysts. Here is why they matter for the valuation math specifically.
Salesforce reports Wednesday after the close. Consensus expects $11.06 billion in revenue and roughly $3.13 in adjusted EPS, both representing double-digit growth. The options market is pricing an 8.7% move in either direction, more than double the four-quarter average. Agentforce, the company's AI agent platform, has closed over 29,000 deals and generates $800 million in annualized recurring revenue.
CRM is the test case for whether AI is converting enterprise software spending into actual profit growth. Workday already set a positive tone with a 12% pop after its beat. If Salesforce confirms the pattern, it validates the forward earnings estimates for the entire software sector. If it misses, the "AI saves SaaS" thesis cracks.
Dell reports Thursday after the close. Consensus expects roughly $35.5 billion in revenue, up 52% year over year, with $13 billion in AI server revenue and an update on the $43 billion AI server backlog. Dell has rallied roughly 131% this year on the AI infrastructure buildout.
Dell is the stress test for whether AI capex is translating past Nvidia into the broader supply chain. The single most important number is not revenue. It is the ISG operating margin. Selling $13 billion in AI servers at thin margins does not support the earnings growth the market is pricing. If margins compress, the AI hardware trade narrows back to Nvidia alone.
The PCE inflation print lands Thursday morning at 8:30 AM. March core PCE came in at 3.2% year over year, the highest since November 2023. The April reading is expected to hold near that level or move higher, driven by energy costs that have not fully retreated.
PCE is the Fed's preferred gauge. It is the first reading Kevin Warsh will face as chair. If it surprises above 3.4%, the rate hike debate returns in force. The 30-year yield above 5% would look less like an overshoot and more like a new normal. Rate-sensitive sectors from homebuilders to REITs would reprice. If it comes in soft, Warsh gets room to maneuver, and the eight-week winning streak has a shot at nine.
What Breaks If They Miss
The risk is not that any single data point disappoints. The risk is the compounding effect.
If CRM misses and signals that AI is not yet flowing through enterprise earnings, the forward growth assumptions embedded in the software sector deflate. If Dell reports strong revenue but weak margins, the "AI is real" thesis survives but the "AI is profitable beyond Nvidia" thesis does not. If PCE runs hot, the bond market reprices again, and the equity risk premium that was already thin gets thinner.
At 27 times trailing operating earnings and a CAPE near 41, the market has limited room to absorb negative surprises. Corrections from these valuation levels have historically been triggered not by recessions but by disappointments, by the moment when realized earnings fail to match the growth rate the multiple was pricing.
The S&P 500 does not need a crisis to pull back 5% to 10% from here. It needs one bad week of data.
What the Bull Case Requires
The bull case is simple and it is not irrational. AI capex started 2026 at $515 billion across the major hyperscalers and has been revised to $740 billion at mid-year. Penciled-in 2027 spending is $889 billion, which would represent 2.7% of U.S. GDP going into AI infrastructure alone. In 2019 that figure was 0.3%.
The critical difference between this cycle and the dot-com comparison is that earnings are tracking the spending. Tech EPS estimates for 2027 are up roughly 20% year to date. In 1999, profits lagged capex by a mile. That is not happening here.
But a cycle where earnings follow capex is still a cycle. It has a peak. The question is not whether AI is real. It is whether the growth rates baked into a 41 CAPE can be sustained for the three to five years the multiple implicitly assumes.
What to Watch From Here
This is not a week for conviction in either direction. It is a week for data.
If CRM beats, Dell margins hold, and PCE comes in at or below 3.2%, the S&P 500 at current levels looks defensible. The earnings growth story is intact. The winning streak extends.
If two of those three disappoint, the market is sitting at a valuation level where the response is fast. Eight consecutive winning weeks create positioning asymmetry. When everyone is leaning the same way, the reversal is not gradual.
The Iran deal adds another variable. Brent crude fell below $99 on Sunday, down over 5%, on signals that a framework is taking shape. But Iran's foreign ministry said a deal is not imminent. If the deal stalls and oil reverses toward $105, the PCE print gets uglier and the rate calculus shifts again.
The math at these multiples is not broken. But it is demanding. And this week will tell you whether the market can meet its own expectations.
