Google Is Getting Paid in Marvell Stock Warrants for Buying Marvell's Chips

Google/Alphabet received a Marvell warrant for 58,970,907 shares (~7% of Marvell) at a $206.58 exercise price as part of an AI chip partnership, implying Marvell shares jumped ~10% to close near $237. The full warrant would cost Google about $12.2B to exercise and is already worth roughly $2.1B more than exercise at current prices, potentially lifting Alphabet earnings via mark-to-market/“other income” effects. Alphabet also reported Q2 revenue up 24% YoY to $119.8B and a $98B net gain (primarily unrealized gains on equity securities), contributing to EPS nearly quadrupling to $9.11.

Analysis

This reads less like a one-off customer perk and more like a pricing signal in hyperscaler procurement: the buyer now has enough volume leverage to force suppliers to share equity upside. That is structurally negative for the chip vendor class because it implies future design wins may come with hidden financing costs, margin givebacks, or dilution, especially where a single customer dominates backlog. The second-order winner is the large buyer ecosystem: the more concentrated the AI capex stack becomes, the more optionality and bargaining power accrues to the cloud platforms rather than the silicon names.

For MRVL, the near-term stock pop can persist if investors focus on headline “AI exposure,” but the 1-3 month risk is that the market realizes the economics are customer-optioned, not contracted. Vesting tied to discretionary spend means the warrant is effectively a contingent rebate on future gross margin, so the true value to equity holders depends on whether incremental revenue arrives with acceptable ROIC. Watch for any sign Google shifts spend to other custom silicon partners; that would turn this from a growth signal into a competitive wedge against smaller ASIC vendors.

Contrarian take: the market may be underpricing the signal for GOOG/GOOGL. This is a small, asymmetric way for Alphabet to monetize capex and strengthen TPU supply optionality without committing balance sheet cash. The bigger issue is not the warrant itself but the precedent: if other hyperscalers negotiate similar structures, AI infrastructure returns can migrate from suppliers to customers over 6-18 months. That would support the large-platform multiples while compressing valuation support for narrow-moat chip suppliers when the capex cycle cools.

Key Decisions for Investors

• Tactically short MRVL into post-disclosure strength; use a 1-3 month horizon and cover on any follow-through in Google capex or evidence the warrant is accretive to booked demand. Falsifier: upward revision to MRVL AI revenue guidance or evidence of materially better gross margin on Google-related programs.
• Prefer long GOOG/GOOGL over MRVL as a pair trade for the next 3-6 months: Alphabet gains supplier flexibility and potential equity upside with limited balance-sheet risk, while MRVL absorbs dilution and customer concentration risk.
• For semiconductor exposure, underweight custom-silicon enablers versus diversified AI compute leaders; pair long NVDA against short MRVL only if market starts rewarding "AI adjacency" indiscriminately. This expresses the view that monopoly-like GPU pricing power is more durable than ASIC design-win economics.
• Set an alert on any follow-on disclosures from AVGO or other hyperscaler suppliers that include equity-linked concessions; if this becomes a pattern, it is a medium-term negative for semicap equipment and design-services margins.