On July 29, Marvell Technology signed a commercial agreement with Google. On August 18, it issued Google a warrant to buy 58,970,907 shares of Marvell common stock at $206.58 apiece — a stake worth roughly $12.2 billion if fully exercised, and about 6.7% of the company.
Marvell did not receive $12.2 billion for it. It received the right to sell Google chips.
The stock went up nearly 10% on the news. Marvell reports fiscal second-quarter results Thursday, August 27, after the close — and Thursday is the first time management has to explain, on the record and under questioning, what that trade actually buys.
Wednesday's tape gave it a quiet runway. The S&P 500 closed at 7,675.83, down 1.45 points, or 0.02%. The Dow fell 0.21% to 53,463.88, snapping a three-day win streak. The Nasdaq Composite slipped 0.08% to 26,130.20. The VIX eased to 15.21. The 10-year Treasury yield sat near 4.65%, and oil fell a third straight day — Brent settled at $87.84 and WTI at $82.23 — after Iran and Oman signaled a shared-control arrangement in the Strait of Hormuz.
Marvell closed at $245.11, up 1.97% on volume of 16.2 million shares, against an intraday range of $236.79 to $245.17.
What the Warrant Actually Says
The structure is in the 8-K, and it is unusually explicit.
1,360,867 shares vest on time alone, in equal quarterly installments over the first year.
The remaining 57.6 million shares vest only on purchases. They are divided into 240 equal tranches, and one tranche vests for every $500 million of custom-product revenue Google buys.
Do the multiplication: 240 × $500 million = $120 billion of cumulative revenue through fiscal 2033 if every tranche vests.
The exercise price, $206.58, was struck near where the stock traded when the deal was papered. It closed Wednesday well above that.
The warrant runs until August 18, 2033.
Read that vesting schedule again. Google's equity in Marvell grows every time Google buys more Marvell silicon. The customer has been converted into a shareholder whose stake is a direct function of its own order book.
That is either the most aligned commercial relationship in semiconductors or the most expensive customer acquisition in its history, and Thursday's call is where analysts start pressing on which.
The Business Underneath It
The dilution question only matters because the underlying numbers are real.
Q1 fiscal 2027 revenue: $2.418 billion, a record, up 28% year over year and $18.0 million above the midpoint of the company's own guidance.
Q1 non-GAAP gross margin: 58.9%; non-GAAP diluted EPS $0.80 (GAAP $0.04, the gap being stock compensation and intangible amortization).
Q2 guidance: $2.700 billion ±5%, which management framed as 35% year-over-year growth — an acceleration from 28%, not a deceleration.
Guided non-GAAP EPS: $0.93 ±$0.05. Guided non-GAAP gross margin: 58.25%–59.25%. Diluted share count guided to 915 million.
Wall Street's consensus lands on the guide: $2.72 billion in revenue (+35%) and $0.93 in EPS (+39%).
CEO Matt Murphy's language on the last call was not hedged: "We expect revenue growth to continue accelerating each quarter throughout fiscal 2027," and the company "significantly raised" its fiscal 2027 and 2028 outlook. The drivers he named: 800G and 1.6T scale-out optics, 51.2T Ethernet switches, co-packaged optics, datacenter interconnect, and custom XPU silicon.
Marvell now has a design relationship inside all three major U.S. hyperscalers: Amazon's Trainium, Microsoft's Maia, and now Google's TPU ecosystem. That third name is what the market has been paying for. The bottleneck in AI infrastructure has moved down the stack, away from the headline accelerators and into the components around them — which is precisely where Marvell sells.
Nvidia Already Answered Half the Question
Wednesday night, Nvidia reported the quarter the entire complex was waiting on, and it was not close. Revenue of $96.22 billion beat the $92.1 billion consensus by 4.5% and more than doubled year over year, up 105.8%. Adjusted EPS was $2.22 against $2.09 expected. Data center revenue hit $89.02 billion, up 117%, at a 75.0% non-GAAP gross margin. The guide is the number that matters here: $108 billion for the October quarter, ±2%, against a street looking for roughly $104 billion — with no China data-center compute revenue assumed in it.
Nvidia had closed the regular session down 1.59% at $209.66. Marvell traded up more than 3% after hours on the read-through, to roughly $253. That is the market saying the capital-expenditure budget Marvell's warrant is written against is real. It is not the market saying Marvell earns its share of it. Thursday night is where that gets tested.
The Price Is the Risk
Here is the part that cuts the other way.
Marvell trades at a market capitalization of roughly $213 billion on 877 million shares, against trailing twelve-month revenue of $8.72 billion. That is a trailing P/E near 84 and a forward P/E near 53. The stock is up more than 183% year to date and its 52-week range runs from $61.44 to $329.88 — a low-to-high span of better than five times inside one year. Beta: 2.25.
The analysts are almost uniformly positive and almost uniformly scrambling to keep up. In the three days before the print, Rosenblatt went from $240 to $300, Susquehanna's Christopher Rolland from $230 to $265 (expecting "better results/guidance, driven mainly by AI networking strength," and modeling ~$4 billion of custom AI revenue in fiscal 2028 against the company's reaffirmed ~$10 billion fiscal 2029 figure), and Citi's Atif Malik held Buy at $275. Morgan Stanley's Joseph Moore kept a Hold and raised his target from $195 to $224 — still below where the stock trades.
The consensus target across 44 analysts is $269.28. The high is $400. The low is $126, which is roughly half of Wednesday's price. That spread is the honest picture: nobody on the Street agrees on what this business is worth, and the money that moves before the crowd does has been repositioning around exactly this kind of dispersion.
The Macro Behind the Print
Wednesday morning's July PCE report did not cooperate. Headline ran +3.7% year over year against 3.6% expected; core ran +3.3% against 3.2% expected. Both were unchanged from June, and core rose 0.25% on the month versus a 0.18% forecast and June's 0.15%. The personal saving rate was 3.0%. The Fed's target is 2%, its preferred gauge has printed 3.3% core two months running, and this is the last PCE reading the September 16 FOMC will see — the next release is September 30, two weeks after the vote.
Jackson Hole opens Thursday. Chair Kevin Warsh delivers his first symposium keynote as chair Friday at 10:00 a.m. ET, at a program themed on financial innovation and payments — a structural agenda, not a cyclical one. Also reporting Thursday: Dollar General and Dollar Tree before the bell (consensus $2.01 and $1.15 in adjusted EPS, with DG comparable sales seen up 2.6%), and Workday after the close, where Silver Lake buyout reports from August 13 have added a takeout premium. Trade policy runs underneath all of it, and the specific line items in the newest tariff lists matter more than the headline numbers.
What Lands Next
Four things to watch Thursday evening:
Whether revenue beats the $2.700 billion midpoint by more than the $18 million Q1 beat. Murphy promised acceleration every quarter this year. The guide for Q3 is the test of that promise, not the Q2 result.
Any disclosure on Google custom-product revenue timing — the warrant vests in $500 million tranches starting in fiscal Q3, which means the first tranche is a dated, checkable event.
Non-GAAP gross margin against the 58.25%–59.25% guide. Custom silicon carries lower margins than optics. A rising custom mix is a revenue tailwind and a margin headwind at the same time.
The share count. Guided to 915 million diluted. The Google warrant sits on top of that, and every dollar of future revenue brings a piece of it into the count.
Marvell has spent the year being repriced from a networking company into an AI infrastructure company. Thursday night is the first quarter where it has to show that the contract it paid $12.2 billion of its own equity to win is actually turning into revenue.
Found this helpful? Share it with others.