Broadcom Has Seen Its Biggest Customer Drop a Key Chip Before. History Says What the Stock Did Next.

Broadcom shares fell ~5% on Aug. 19 after Marvell disclosed an expanded custom chip agreement with Google, including a warrant tied to up to $120B of future purchases. The article argues this may reflect Google adding a second silicon partner rather than fully moving away from Broadcom, citing Broadcom’s April long-term TPU/rack components deal through 2031 and its strong AI momentum (AI semiconductor revenue $10.8B in FY2Q, +143% YoY). It flags concentration risk—broadcasters note Google’s share isn’t disclosed—and suggests the true read-through will come on upcoming earnings rather than the one-day drop.

Analysis

The market is pricing a substitution story, but the more important mechanism is bargaining power: Google is signaling it wants optionality across TPU generations, which can pressure Broadcom’s pricing and cadence even if units don’t migrate meaningfully. That matters because AVGO is now valued on AI growth durability, not legacy chip cash flows; a small hit to the growth vector can compress the multiple faster than the revenue delta would imply.

Near term, Marvell is the clearest relative winner because it gets a stronger strategic foothold without needing to win the whole socket. The second-order effect is that other hyperscalers may push harder for dual-sourcing or in-house designs, which could dilute the economics of custom silicon broadly and shift bargaining leverage away from incumbents like AVGO over the next 1-3 quarters. The key question is not whether Google keeps Broadcom; it is whether Broadcom’s AI mix can keep compounding at a >100% rate if the customer base becomes less captive.

The contrarian view is that the selloff may be too binary: hyperscaler silicon programs usually expand in layers, and dual-source setups often increase total spend before they replace anything. Still, at ~60x earnings, AVGO is vulnerable if the next couple of earnings prints show even modest deceleration in AI revenue growth or weaker commentary on Google/T TPU ramps. Falsifiers: maintained or raised AI semis guide, evidence that Google purchases are additive rather than reallocative, or any indication MRVL’s role is narrow instead of program-wide.

Key Decisions for Investors

• Relative-value trade: long MRVL / short AVGO for 1-3 months, sized modestly. Thesis: market is underpricing Marvell’s strategic optionality and overpricing Broadcom’s customer concentration risk. Risk/reward improves if AVGO sells off on any soft AI commentary while MRVL keeps getting re-rated as a second-source beneficiary.
• Do not chase AVGO weakness into the first reaction window; wait for the next earnings print and AI revenue commentary. If Broadcom reiterates or raises AI semiconductor growth without visible Google share loss, the post-news drawdown likely becomes a buying opportunity rather than a structural short.
• Set a watch alert on AVGO’s next quarterly AI revenue guide and hyperscaler commentary. If AI growth guidance drops below the current hyper-growth trajectory or management hints at mix pressure from Google, that is the point to consider a tactical short or put spread.
• If you want convexity, prefer a small AVGO put spread into earnings rather than outright short stock. The thesis only pays if the market starts discounting 2027-2030 TPU economics, but downside is capped if this remains a slow partial transition.