Marvell Reports Thursday. Here's Where the Google Contract Would Actually Show Up in the Numbers.

Marvell is set to report fiscal Q2 2027 results on Aug. 27, after guiding revenue to about $2.7B (+35% YoY at the midpoint). Management also expects EPS of $0.37 (GAAP) and $0.93 (non-GAAP) for the quarter, each ±$0.05, with data center end-market revenue of $1.83B (+27% YoY) as the key line item. The prior expanded Google custom-chip deal includes a warrant with revenue-measuring starting Aug. 1, 2026, meaning Thursday’s numbers may reflect limited Google revenue and rely more on forward outlook for the ramp. The stock has rallied on the agreement despite uncertainty around pace, and investors will watch guidance and commentary for indications of accelerating growth into fiscal Q3.

Analysis

The market is likely pricing the wrong quarter. The Google expansion is more of a long-duration design-win signal than a near-term revenue contributor, so the next two prints are still mostly a test of whether the underlying data-center franchise can keep comping above consensus without help. That matters because at this valuation, MRVL does not need good news — it needs a clean sequence of upside revisions, and any hint of a flat guide would invite multiple compression even if the headline quarter looks fine.

The bigger second-order issue is bargaining power. A hyperscaler willing to pay partly with warrant economics is effectively saying it can shift some of the AI value transfer away from the supplier; that is positive for Google’s cost curve but a warning that future custom-silicon deals may come with lower margins, heavier customer concentration, and more revenue quality risk. If this pattern spreads, the broader custom-ASIC group — especially names with similar hyperscaler exposure like AVGO — could see the market demand better disclosure around mix, take-rates, and true recurring content.

Contrarian view: consensus may be underestimating the strategic significance of the Google win while overestimating the amount of P&L it contributes in the next 1-2 quarters. The real falsifier for a bearish read is not the reported quarter itself, but a materially stronger fiscal Q3 guide and commentary that quantifies a faster-than-expected ramp in customer-owned AI infrastructure. Absent that, the path of least resistance is a volatility event rather than an immediate re-rating.

For GOOG/GOOGL, the economics are modestly positive over 6-18 months because lower inference cost and better supply diversification protect AI margins; the effect is more defensible operating leverage than a near-term earnings pop. For MRVL, the near-term risk is that the stock has already capitalized too much of the story before the revenue actually shows up.

Key Decisions for Investors

• Do not add to MRVL ahead of earnings at current levels; if the stock rallies into the call and management only reiterates guidance, fade strength via a short-dated MRVL put spread or a short vs. SOXX overlay. Falsifier: fiscal Q3 guide meaningfully above consensus and explicit acceleration in data-center growth.
• Longer horizon: accumulate GOOG/GOOGL on AI hardware pullbacks. The thesis is not headline AI spend, but better unit economics from internal silicon that can preserve search/cloud margin structure over 6-18 months.
• Relative value into the Oct. 6 investor day: long GOOG, short MRVL. This isolates the beneficiary of lower AI compute cost from the supplier most exposed to valuation and dilution risk if the market gets ahead of the ramp.
• If already long MRVL, hedge into the print with a collar rather than adding exposure. The event is more likely to de-risk a future story than to re-rate the stock immediately.