The S&P 500's Earnings Boom Is Built on Paper Profits — Here's What's Underneath

The S&P 500's Earnings Boom Is Built on Paper Profits — Here's What's Underneath

The S&P 500 is wrapping a second-quarter earnings season that looks, on paper, like the best since 2021. Blended earnings growth is at 50.4%. Eighty-six percent of reporting companies beat estimates — 86% beat estimates. The headline number alone would be enough to make anyone feel comfortable about the broad market.

The problem is that the headline number is not the earnings number.

About a fifth of that reported growth — roughly 19 percentage points — came from two companies writing up the value of stakes in private AI startups at quarter-end. Alphabet added nearly $99 billion in net gains on equity securities, mostly from SpaceX. Amazon added $53.4 billion, mostly from Anthropic. Those are not cash collections. They are mark-to-market snapshots that reverse just as fast as they appear. SpaceX has already fallen roughly 34% from its June 30 closing price. When Alphabet closes its third-quarter books, that gain becomes a loss on the same earnings line.

Strip out Alphabet and Amazon's investment gains and the index still grew 32% in the quarter. That's strong — the seventh consecutive quarter of double-digit growth. But it's not the story the 50.4% figure tells you. And confusing the two is the kind of error that leads to wrong positioning when the marks flip.

The broader picture is even more telling. The seven largest tech companies together reported $315.6 billion in Q2 profit. $134.6 billion of that came from unrealized investment gains. Without those marks, the group's core earnings were roughly flat. Goldman Sachs estimates that AI infrastructure stocks account for roughly one-third of the index's EPS growth. Those are real businesses with real revenue — Nvidia, Broadcom, Micron — but the accounting picture blends their operating results with paper swings from companies like Anthropic and SpaceX that aren't publicly traded in any meaningful sense.

So what's the actual takeaway? Not that the S&P 500 is broken. Not that the earnings season was a fraud. The takeaway is that you have to separate the signal from the noise before you position against either. The factor stack still supports broad index exposure — 10 of 11 sectors posted growth, revenue is expanding across all 11, and the beat rate is above its five- and ten-year averages. But concentration risk has shifted into a new gear. The top 10 S&P 500 companies generate 34% of all index profits, roughly double their share in the mid-1990s. Alphabet alone accounted for 92% of the net dollar-level increase in index earnings during its reporting week.

Here's how the major players look when you peel back the accounting layer and read the actual factor data.

Alphabet (GOOGL) — Cheap on the surface, volatile underneath

Alphabet closed at $341.31 today, down roughly 1%. On a trailing PE it sits at 17.1x — the cheapest of any major tech name in the index. The forward PE of 33.3x is still below Microsoft's 26.8x trailing multiple. If you only look at the multiple, Alphabet looks like the value pick.

But the 17.1x trailing PE is propped up by that $77.1 billion net income boost from equity gains — or $6.26 per share, nearly 70% of GAAP earnings for the quarter. Without it, Alphabet's EPS for the quarter was $2.85, just below the $2.88 consensus estimate. The underlying business — search, cloud, YouTube — is growing solidly, with operating income up 30% on 24% revenue growth. That's real. But the valuation multiple that looks attractive is partly a function of a one-time accounting event that could reverse in the next earnings report.

Portfolio role: core index holding, but treat the trailing PE as misleading. The forward multiple and the operating momentum are what you should be pricing on. If SpaceX stays depressed through September, the next GAAP report will look ugly even if the core business is fine. That's a timing risk, not a thesis risk.

Amazon (AMZN) — Cloud cash flow plus a big paper windfall

Amazon's trailing PE sits at 20.8x with a market cap of $2.8 trillion. Like Alphabet, Amazon's headline growth is inflated by its Anthropic stake — a $53.4 billion pre-tax non-operating gain that represented roughly 65% of net income for the quarter. Strip that out and Amazon's earnings growth was approximately 17%. Still solid, but not the 240%+ headline number.

The core story here is AWS and the e-commerce margin expansion, both real. Amazon is also a major backer of OpenAI, having committed $50 billion in February. These are long-lead investments with uncertain payoffs. The Anthropic valuation has surged because the company is valued at nearly $1 trillion in private markets and is expected to go public within the next year. But mark-to-market gains from a private startup that hasn't shown a path to profitability are not the same thing as AWS free cash flow.

Portfolio role: growth sleeve. Hold for the cloud revenue trajectory, not the investment gains. If the Anthropic mark reverses, it doesn't change AWS economics.

Microsoft (MSFT) — The cleanest story, at a premium

Microsoft traded at $482.04 today. Its trailing PE of 26.8x is the highest among the mega-caps outside Nvidia. But Microsoft's investment gains — $3.2 billion from Anthropic and $480 million from OpenAI — added roughly 10 percentage points to earnings growth, a fraction of Alphabet's or Amazon's distortion. The core business carries the weight here: Azure, Windows cloud, Copilot monetization, LinkedIn.

Microsoft also has the most diversified exposure. It's not one company propping up the valuation with a single private stake. The AI investments are material but they're not rewriting the income statement the way they are for Alphabet.

Portfolio role: quality anchor. You pay more for Microsoft because the earnings stream is more predictable. In a framework that weights durability alongside growth, Microsoft earns its premium.

Nvidia (NVDA) — The real AI cash flow engine

Nvidia is at $217.39, trading at a 33.0x trailing PE — the most expensive of this group on a trailing basis, but justified by actual revenue growth rather than paper marks. Nvidia is the company that's selling the shovels, not writing up a stake in the gold mine. Its earnings are cash, not accounting adjustments.

The concentration concern here is different. Nvidia's own earnings dominate the semiconductor supply chain. When the AI buildout slows, Nvidia feels it first. But right now, the demand pipeline from hyperscalers — including the ones marking up their private stakes — is feeding Nvidia's real operating results.

Portfolio role: momentum growth. This is the barbell's growth rail. High conviction in the AI infrastructure cycle, but watch for the inflection point where hyperscaler capex guidance starts to decelerate.

What the factor stack says about the broader index

The S&P 500 is not in trouble. Earnings growth of 32% without the two biggest marks is still exceptional by historical standards. Revenue growth is at its strongest since late 2021. The health care sector's contraction was driven by one-off charges at two firms; without them, the sector grew nearly 18%. Energy's 146% surge was oil-price-driven and unlikely to repeat, but it doesn't negate the underlying strength elsewhere.

What the factor stack does say is that concentration has become the dominant risk. The same companies that benefit most from the AI cycle are the ones whose reported earnings are most distorted by the accounting treatment of their own investments. Alphabet's SpaceX mark went from a non-cash gain in June to a 34% decline by late July. That swing alone could wipe out a full quarter of core earnings growth when it hits the September report.

BofA Securities' Savita Subramanian put it bluntly: dependence on uncontrollable factors reduces earnings visibility. The mark can turn into a loss just as fast.

How to position

The answer isn't to abandon the index. The answer is to understand what you're exposed to and build structure around it.

A barbell approach makes sense here. On one side, the AI infrastructure names — Nvidia, Microsoft — where real revenue growth is underwriting the multiples. On the other side, dividend yield and cash-flow durability in sectors less exposed to the mark-to-market volatility of private AI valuations. The energy sector's surge was temporary, but durable cash-flow businesses in industrials, select financials, and health care outside the two-company contraction offer ballast when the big marks swing.

Watch three things in the third quarter: SpaceX's valuation trajectory into September 30, Amazon's next Anthropic mark, and whether AI infrastructure capex guidance from hyperscalers holds or starts to crack. Those three signals will tell you whether the paper profits stay on the books or reverse — and whether the broad index still looks as strong as the headline numbers suggest.

The S&P 500's earnings boom is real. It's just not as big as the headline says, and it's not as stable as the narrative implies. The difference between 50% growth and 32% growth is the difference between comfortable and cautious. In a framework that compares before it convicts, that difference matters.