There Are 16 Billion Reasons to Pay Attention to Broadcom Earnings on Sept. 2

The market is likely over-interpreting this as a zero-sum share shift when the more important mechanism is capacity expansion. For hyperscalers, the first constraint is not chip ideology but time-to-deploy and packaging/test capacity; adding Marvell reduces single-vendor bottlenecks and can be additive to Broadcom rather than displacing it. That said, the incremental dollars are still a competitive signal: if Alphabet is broadening its custom silicon stack, other cloud buyers may push harder on supplier leverage, which can cap long-run pricing power across ASIC vendors.

Broadcom’s real setup is not headline share loss but whether AI revenue can keep compounding fast enough to justify the current multiple after a year of underperformance versus the index. If AVGO clears the next print with a credible Q4 guide, the stock can rerate quickly because the bear case has been about durability, not absolute demand. The falsifier is not one new customer announcement; it is any sign that AI growth is decelerating while non-AI semis fail to offset, which would force estimate cuts and multiple compression over the next 1-3 months.

Marvell is the cleaner event-driven beneficiary, but the move may be getting ahead of fundamentals. A warrant-linked commercial structure can inflate sentiment faster than revenue, so the key watch item is whether it translates into repeatable design wins and margin-accretive volume over 2-4 quarters. Contrarian view: the consensus is underestimating how much custom silicon spending can support both names simultaneously; the bigger risk is that the market starts to treat the entire custom-ASIC complex like a winner-takes-all trade, which is not how hyperscaler sourcing usually works.