$AVGO delivered a strong, but not flawless, Q3. The important result was AI execution, not the modest headline revenue beat.
Going into the print, my test was whether Broadcom could convert AI backlog into revenue, not whether it could manufacture a token headline beat. The
AWS and NVIDIA just announced 2M *more* GPUs for 2027 and 2028. Blackwell Ultra, Rubin, Rubin Ultra. Plus Vera CPUs on AWS, 100K GPUs for federal AI factories, and NVHBM going into Trainium racks.
Is 2M is a big number? It is. But here's what matters more. AWS committed to 1M+
Deal terms for $MRVL$GOOG *were* disclosed:
-$120B over 6 years (annual forecast is $11.5B) You do the math. Nearly doubled the size of the company?
-custom, NOT off the shelf networking or optics
-not TPU, but what $MRVL calls XPU attach
The “AI inference accelerator”
$CSCO delivered a clean Q4 FY26 beat, but the more important signal is broader networking demand beyond hyperscalers.
Revenue: $17.25B vs. $16.82B, BEAT by $433M / 2.6%.
Non-GAAP EPS: $1.22 vs. $1.17, BEAT by $0.05 / 4.3%.
Q1 FY27 non-GAAP EPS guide: $1.32-$1.34, with the
$COHR delivered and AI optics beat and a strong Q1 guide. The stock’s negative first read says the bar was obv a lot higher. Maybe it’s gross margins.
Revenue: $2.046B vs. $1.98B, BEAT by about $66M / 3.3%.
Non-GAAP EPS: $1.74 vs. $1.62, BEAT by $0.12 / 7.4%.
Non-GAAP gross
It looks like demand is off the charts for this $INTC offering. They’re upping the base offering AND the greenshoe.
Yet another vote of confidence in $INTC Foundry.
This is the largest primary, all-common-stock U.S. follow-on offering on record where the money goes to the
