Key Highlights

  • Broadcom's Q2 results disappointed elevated AI Revenue growth expectations, triggering a broad semiconductor sector selloff that erased approximately $1.3 trillion in chip stock Market Capitalisation.
  • The selloff extended to Marvell Technology (Nasdaq: MRVL) and the XLK technology ETF, reflecting systemic repricing of AI infrastructure growth assumptions.
  • Stronger-than-expected jobs data concurrently raised rate hike fears, compressing the discount rate applied to long-duration Growth Stocks.
  • Broadcom's custom ASIC ramp, which is the closest commercial analogue to Marvell's custom silicon programme, appears to be progressing more slowly than analysts had assumed.
  • The key distinction between AVGO's disappointment and MRVL's thesis is whether hyperscaler AI capex is decelerating or merely smoothing — a question the next two quarters will clarify.

 

Analysis

When Broadcom (NASDAQ: AVGO) reported its Q2 results and the AI revenue growth figures disappointed against expectations that had been inflated by months of bullish analyst commentary and management guidance, the semiconductor sector discovered once again the danger of pricing in optimistic scenarios as base cases. The resulting selloff — approximately $1.3 trillion in chip stock market capitalisation erased in a matter of sessions — was not merely a repricing of Broadcom. It was a repricing of the assumptions that underpin the entire AI infrastructure Investment thesis, including those applied to companies like Marvell Technology.

 

The Mechanism of Guidance Inflation

Semiconductor companies serving AI infrastructure customers operate in an environment where the Demand signal from hyperscalers is genuinely strong, but where the timing of revenue recognition — from design win to tape-out to qualification to Volume production to invoice — creates a long and uncertain path between the commercial opportunity and the financial statement. The temptation for management teams, Investor relations departments, and Sell-Side analysts is to extrapolate the demand signal forward at the steepest plausible angle.

 

The result is that expectations for AI semiconductor revenue can become embedded in stock prices at levels that require not just strong growth, but the most optimistic timing assumptions for every ramp, every qualification, and every customer deployment cycle. When any one of those timing assumptions slips — when a custom ASIC qualification runs three months behind schedule, or when a hyperscaler spreads its procurement across more suppliers than expected, or when a programme is resized at the last minute — the gap between expectation and delivery creates a violent repricing.

 

The Custom ASIC Parallel

Broadcom's AI revenue is heavily weighted toward custom application-specific integrated circuits — chips designed in collaboration with specific hyperscalers for specific AI workloads. This is precisely the Business model that Marvell has been executing and growing through its own custom silicon programme, which doubled in FY26 and is guided to more than double again in FY28. The fact that Broadcom's custom ASIC ramp appears to have progressed more slowly than the market assumed is a direct read-across risk for Marvell's own custom silicon ambitions.

 

The read-across is not deterministic. Broadcom and Marvell serve different customers within the hyperscaler universe and their custom programmes are at different stages of the ramp cycle. Marvell's Q1 FY27 results — $2,418M in revenue, above the guidance midpoint, with management confirming several new design wins — do not show the same pattern of disappointment. But the risk that should be named is this: if hyperscaler AI capex is genuinely decelerating, or if the transition from infrastructure build-out to inference workloads reduces the pace of new silicon procurement, then both Broadcom and Marvell face a common headwind that is independent of their individual execution.

 

The Rate Environment Compound

The simultaneous release of stronger-than-expected jobs data on the day of Broadcom's results added a second compression Factor. Semiconductor companies with high growth expectations and correspondingly high valuation multiples are long-duration Assets — their value rests disproportionately on Earnings and cash flows years or decades in the future. When the Risk-Free Rate rises, those future earnings are discounted more heavily, reducing the present value of the growth premium. The combination of an earnings disappointment from the sector's highest-profile AI revenue reporter and an upward shift in rate expectations is about as unfavourable a dual shock as these stocks can absorb.

 

The disciplined response is not to extrapolate the bad news any more than one should extrapolate the good news. The underlying AI infrastructure build-out — the investment in compute, memory, and connectivity by the five major hyperscalers — continues. The question Broadcom's results raise is about the timing and distribution of that investment, not its direction. For investors in Marvell specifically, the next two quarters will clarify whether the company's own revenue trajectory — guided at $2.7 billion in Q2 FY27 — is tracking to plan or beginning to show the same timing delays that appear to have affected Broadcom.

Disclaimer

This article is for informational purposes only and does not constitute financial advice or a recommendation to buy or sell any security. All data is sourced from Marvell Technology's official earnings presentations (FY26 Q4 and Q1 FY27). Investors should conduct their own Due Diligence before making investment decisions.