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

Broadcom shares fell about 5% on Aug. 19 after Marvell disclosed an expanded custom chip agreement with Google tied to warrants for as much as $120B of future purchases, prompting investors to worry about Google order concentration. The article argues the risk is likely more about share of AI chip orders than a full customer loss, noting Broadcom’s April long-term TPU-related agreement and that AI semiconductor revenue reached $10.8B (+143% YoY) with management guiding $16.0B in the next quarter (+200%+). Overall, the headline creates near-term uncertainty for Broadcom’s AI growth trajectory rather than evidence of a break in its broader customer relationships.

Analysis

The market is treating this as an immediate loss of franchise value, but the more important mechanism is bargaining power. In custom silicon, a second source usually starts as leverage before it becomes displacement; that tends to pressure the incumbent’s multiple faster than its revenue, because investors discount future share before the tape-out cycle can prove anything. That said, if Google is now structurally dual-sourcing TPU-related workloads, the incremental dollar at risk is not the old wireless-style side business — it is the growth engine that has been underwriting AVGO’s premium valuation.

Second-order effects matter here: a successful Google dual-source strategy would embolden other hyperscalers to demand the same treatment, which is bearish for margin durability across AVGO, MRVL, and even NVDA’s custom/adjacent ecosystem over 6-18 months. The near-term winner is MRVL’s narrative multiple; the longer-term winner may be Alphabet, which gets supply optionality and negotiating leverage. The key question is not whether AVGO loses the account overnight, but whether future TPU ramps become shared economics instead of sole-source economics.

Contrarian view: the selloff may be overdone if investors are extrapolating a binary customer-loss story onto a multi-year design cycle. The real falsifier is AVGO’s next few earnings prints: if AI semiconductor growth stays above ~100% y/y and guidance remains intact, this is a headline-driven compression event, not a fundamental break. If, however, management stops quantifying acceleration or commentary implies Google is a smaller contributor to the next leg of AI growth, the stock’s 60x earnings multiple has room to de-rate further.

Key Decisions for Investors

• Relative value: long MRVL / short AVGO for 1-3 months, sized to isolate TPU-share-shift sentiment rather than sector beta; target is multiple dispersion if Google’s dual-source narrative keeps gaining credibility.
• Do not chase AVGO weakness intraday; wait for the next earnings cycle and use AI revenue guidance as the real catalyst. If AVGO reiterates rapid AI growth, cover shorts quickly — the thesis is timing-sensitive, not binary.
• If already long AVGO, hedge into the next print with a modest put spread or collar rather than an outright exit; the main risk is valuation compression, not an immediate earnings miss.
• Watch-list rather than trade: GOOG/GOOGL. Any evidence that dual sourcing lowers cost or improves supply resilience without slowing TPU rollout is structurally bullish for Alphabet and bearish for AVGO’s scarcity premium.