Marvell Technology closed its fiscal second quarter with $2.739 billion in revenue — a company record, up 37% year over year and $39 million above the midpoint of its own guidance. Data centre revenue growth accelerated to 46%. Non-GAAP net income came to $865.9 million, or $0.94 a share; on a GAAP basis, $308 million and $0.33. The company then guided the current quarter to $3.15 billion, plus or minus 5%, with adjusted earnings of $1.10 a share.
Marvell also raised its full-year fiscal 2027 revenue outlook to roughly $12 billion and set a fiscal 2028 target near $18 billion. Every line cleared.
The stock fell about 6% Friday.
The reason is a date. Investors wanted a bigger lift to the longer-term numbers on the back of the Google custom-silicon agreement, and got limited detail on fiscal 2028 instead — with the meaningful financial impact landing in fiscal 2029. That is later than a stock up 187% year to date was positioned for.
The warrant everyone quoted and nobody read
On July 29 Marvell signed an agreement with Google covering AI inference accelerators, storage controllers and other silicon built around Google’s in-house TPUs. Alongside it, Marvell issued Google a warrant for 58,970,907 shares — about 7% of the company — at an exercise price of $206.58.
The headline wrote itself: Alphabet could own 7% of Marvell. The fine print is the actual deal. The warrant vests in $500 million revenue increments through fiscal 2033. Full dilution only happens if Google first hands Marvell roughly $120 billion in revenue. Google is not being given a stake; it is being offered one, priced in chips it has to buy.
That structure has quietly become standard in AI silicon. The customer takes equity in the supplier, the supplier books a demand commitment, and both sides get a headline number that is technically true and practically conditional. It aligns incentives genuinely well. It also lets a $120 billion figure into a press release without a $120 billion order behind it.
Our take: This is the first clean read on how the market is pricing AI timing rather than AI magnitude. Marvell did not miss. It did not guide down. It raised the full year, and lost 6% of its value in a morning because the big money lands in fiscal 2029. Compare Nvidia, which beat on every line last week and still slipped, and Salesforce, which grew 11% and jumped 14%. The market has stopped paying for eventual AI revenue and started discounting it — and a three-year gap between the announcement and the cash is now worth real multiple compression. Deal structures that push the payoff out are about to get harder to sell.
What is actually working now
Strip out the Google timing and the quarter is strong on its own terms. Data centre growth accelerated rather than decelerated, and management pointed to broad-based strength across the portfolio — connectivity in particular, which is the less glamorous half of the business and the half with fewer competitors. The custom business is set to accelerate in the second half of the current fiscal year, before any Google contribution matters.
The Q3 guide of $3.15 billion implies about 15% sequential growth. That is a company executing, not a company waiting.
What to watch
- The first $500 million vesting tranche. It is the only public evidence of whether the Google relationship is running on schedule.
- Connectivity versus custom mix. Custom silicon carries lower margins; a faster custom ramp is a gross-margin story, not just a revenue one.
- Whether other suppliers copy the warrant. If two more chipmakers pay customers in equity this quarter, it is a market structure, not a one-off.
- Fiscal 2029 language on the next call. Any softening of that date is the whole thesis.
Record quarter, better outlook, worse stock. In this cycle those three go together more often than they should.