Blayne’s Bytes: Revisiting AVGO, Korean Memory Buildout, META’s Cloud Plans

Research evidence excerpt

Blayne’s Bytes: Revisiting AVGO, Korean Memory Buildout, META’s Cloud Plans

USA | Semiconductors EquityJulyResearch2, 2026

Blayne’s Bytes: Revisiting AVGO, Korean Memory

Buildout, META’s Cloud Plans

We kicked off the week with a review of the AVGO bull case, and our analysis

suggests C28 EPS of $30-40. Elsewhere in compute, META is reportedly

building a cloud business to sell excess compute, which would support

.

continued Capex in the near term, we believe, but Meta will need to show model

progress for this to be sustainable. Including additional news on Memory and

more below. As always, reach out to TeamBlayne@jefferies.com with anything

you need.

For our take on D-Wave's National Science Foundation grant, see KG's Kernels section below.

Revisiting AVGO Bull Case – Scenario Analysis Points to $30-40 in C28: The post-earnings

pullback has us revisiting the bull case for AVGO. We would note that there’s work to be done on

XPV funding and supply to make the C28 scenario analysis possible, but it does point to $30-40 of

EPS at $200-250B of ASIC revenue. A midpoint of $225B at 20GWs and $11B/GW. We see the TPU

v9 (Pumafish) on schedule for a late C26 launch featuring N2/3 3D stacked logic and 400G SerDes

(Co-packaged Copper), while MTK's Zebrafish remains in testing/development, with our checks

pointing to only 60k units in C26 vs MTK analyst forecasts of 400-450k. The recently announced

20GW XPV (Apollo/Blackstone) reflects Anthropic and OpenAI demand exclusively and excludes

Google internal and Meta volumes. Customer breadth continues to expand with the unveiling of

the OAI/AVGO Jalapeno XPU, the first in a multi-gen compute platform. Initial testing suggests

significantly improved performance per watt vs industry-leading competitors (we estimate ~50%