FactSet published its Q2 scorecard on Friday and the top line is the kind of number that ends arguments. The blended earnings growth rate for the S&P 500 — actual results for the 88% of companies that have reported, estimates for the rest — is 50.4%. A week ago it was 47.4%. On June 30 it was 23.1%. If it holds, it is the index’s best quarter since Q2 2021, the second straight above 25%, and the seventh straight of double-digit growth. The S&P 500 closed Friday at a record 7,757.64, up 3.6% on the week.
Now the asterisk, which FactSet puts in its own second paragraph rather than burying: strip out Alphabet and Amazon and the blended growth rate falls to 32.0%. Two companies out of 500 account for more than 18 points of the index’s earnings growth.
They got there without selling anything extra. Alphabet reported GAAP EPS of $9.11 against a $2.88 estimate; that actual included roughly $98 billion of other income, primarily net unrealized gains on equity securities — stakes in private companies including Anthropic and SpaceX. Alphabet’s own release put realized and unrealized equity gains at $99.0 billion, worth $6.26 of diluted EPS by itself. Amazon reported $5.75 against $1.82, including $53.4 billion of non-operating other income, primarily from its Anthropic investment. Together: about $152 billion of portfolio marks, flowing through GAAP earnings and up into the index’s headline growth rate.
Our take: The mechanism matters more than the number. These are mark-to-market gains on private AI companies, and the marks come from funding rounds priced by roughly the same cohort of investors. When AI valuations go up, the reported earnings of AI’s largest backers go up with them — and get rolled into “S&P 500 earnings growth,” a phrase most people hear as businesses made more money. It isn’t fraud, it’s GAAP. But it means a widely quoted measure of American corporate profitability now partly tracks the valuation of a handful of private startups. That line moves in both directions, and almost nobody is modelling the down leg.
The surprise number is the real tell
Of the companies that have reported, 86% beat EPS estimates — against a five-year average of 78% and a ten-year average of 76%. That would be the highest share since Q2 2021. The magnitude is the strange part: in aggregate, companies are reporting earnings 29.2% above estimates, versus a five-year average of 7.0%. If that holds, it is the largest earnings surprise FactSet has recorded since it began tracking the metric in 2008, breaking the Q2 2020 record of 23.2%.
Excluding Alphabet and Amazon, the surprise falls to 10.9% — still above both long-run averages, and that is the honest good-news portion. The rest is an artefact of scoring: most analysts submitting estimates used the GAAP actual, other income included, as the comparable figure. The “beat” is measured against forecasts that were never trying to predict a $98 billion revaluation.
What’s clean
Revenue. You cannot mark a sale to market. Blended revenue growth is 15.0%, up from 14.2% a week ago and 12.2% at quarter-end, and the best since Q4 2021. Three-quarters of companies beat on revenue, by an aggregate 3.2% — the widest since Q2 2022. Ten of eleven sectors are growing earnings and eight of those in double digits, led by Energy and Communication Services; all eleven are growing revenue. Health Care is the only sector reporting an earnings decline.
That is a genuinely strong quarter. It is just a 32% quarter with a 50% headline stapled to it.
What to watch
- August 28. FactSet is skipping its August 14 and 21 editions, so Friday’s reading is the last full picture for three weeks. Expect 50.4% to be quoted unqualified for most of that gap.
- The forward estimates. Q3 at 27.4%, Q4 at 25.2%, calendar 2026 at 30.0%. Worth asking how much of that assumes portfolio marks keep going the same way.
- The multiple. Forward 12-month P/E is 20.0 against a five-year average of 19.9, down from 20.4 on June 30. A P/E looks cheaper when the E is doing this.
- The reversal. Unrealized gains are unrealized. A flat quarter for private AI valuations turns this tailwind into a year-over-year hole in the Q2 2027 comparison.
The habit worth keeping is small: when a growth rate jumps 27 points in five weeks without the economy doing anything to deserve it, the move is usually in the accounting, not the operations. Friday’s numbers were good. They were not 50.4% good.