In Defense of AVGO’s ASIC Share

Research evidence excerpt

In Defense of AVGO’s ASIC Share

IdeaMThis debate is somewhat reminiscent of last year’s MRVL / Alchip debate, not

because the share outcomes are identical, but because the market again appears to

be extrapolating new supplier participation into a much more extreme displacement

thesis. Marvell did lose meaningful Trainium share, but the view that it would be

fully replaced proved too draconian. Similarly, while MediaTek clearly has a role in

the TPU ecosystem, we think calls for Broadcom’s share to fall to 50% or below, or

eventually be fully displaced, are premature.

We agree MediaTek has a role, but disagree with converting that into a

disruptive Broadcom share-loss thesis. To be clear, we do not dismiss MediaTek’s

opportunity. Google is cost conscious, wants supplier optionality, and is developing

multiple TPU iterations. The Taiwan supply chain view also raises real points:

MediaTek appears to have a credible 3nm TPU opportunity, Google has an incentive

to reduce reliance on Broadcom, and advanced packaging constraints create a

reason to explore alternatives. Where we disagree is the conversion of those facts

into a disruptive share-loss thesis for Broadcom.

Our base case is that Broadcom maintains roughly 80% TPU share over time. We

do not agree with the supply-chain framework that implies a rapid move toward

Broadcom share loss. This view is consistent with MediaTek’s publicly stated long-

term strategy of reaching 15-20% share. Said differently, we think MediaTek

participation is real, but we believe the market is overstating the amount and pace

of Broadcom share loss within TPU. The key debate, in our view, is not whether

MediaTek participates, but whether that participation disrupts Broadcom’s broader

AI ASIC growth trajectory.