Marvell Beats and Raises, but Shares Slide 8% as Investors Demand a Timeline for Google's $120B Chip Deal

Marvell Technology just delivered one of the strongest quarters in its history — and got punished for it. On Friday, August 28, the chip designer’s shares fell more than 8% to around $221 in early trading, deepening a selloff that began after hours on Thursday. The proximate cause wasn’t the results, which beat on every headline metric, nor the guidance, which was raised. It was a question investors couldn’t get answered to their satisfaction: when does the Google deal actually start paying out?

A record quarter on paper

Marvell’s fiscal second quarter of 2027 (reported Thursday, August 27) was unambiguously good. Revenue hit a record $2.739 billion, up 37% year over year, ahead of both guidance and analyst estimates. GAAP net income came in at $308.0 million, or $0.33 per diluted share; on a non-GAAP basis, profit reached roughly $866 million — a 36.6% operating margin — or about $0.94 per diluted share. Cash flow from operations was $605.5 million.

The data center segment did the heavy lifting, as it has for several quarters: revenue there climbed 46% year over year to approximately $2.17 billion, representing 79% of total revenue. AI infrastructure — custom AI silicon and the high-bandwidth interconnects that stitch accelerators together — remains the growth engine.

Guidance was equally assertive. For the current quarter, Marvell guided revenue to $3.15 billion, plus or minus 5%, with non-GAAP earnings of $1.10 per share, plus or minus $0.05 — both above Street consensus at the time. More significantly, CEO Matt Murphy raised the full-year outlook: Marvell now expects fiscal 2027 revenue of roughly $12 billion (about 45% growth) and fiscal 2028 revenue of approximately $18 billion, with data center revenue growing more than 60% into fiscal 2028. It was the second consecutive quarter the company lifted its annual forecast.

The problem: the Google ramp has no dates attached

None of that was enough, because the market’s attention is fixed on August 19 — the day Marvell announced a sweeping custom-silicon agreement with Google. Under that deal, Marvell attaches to Google’s TPU program across a broad stack of custom chips, with cumulative revenue potential of up to $120 billion tied to purchasing targets running through fiscal 2033. Google also received a warrant to purchase up to 58.97 million Marvell shares at $206.58 apiece — roughly $12.2 billion, or about 6.7% of shares outstanding — vesting against those same purchase milestones.

Investors who bid the stock up on the announcement now want the fine print translated into quarterly numbers. On the earnings call, Murphy said custom-silicon revenue would ramp significantly in the second half of fiscal 2027 and more than double in fiscal 2028, and noted there was “upside bias” to the company’s prior target of $10 billion-plus in custom revenue by fiscal 2029. He also said existing fiscal 2027–2028 targets already reflect some contribution from the Google programs.

That wasn’t granular enough. As Reuters put it in Friday’s follow-up, the selloff deepened “as investors seek clarity on Google AI deal payoff” — how much revenue lands in which quarters, at what margin, and what happens if Google’s purchasing targets slip. Seeking Alpha’s take was blunter: guidance disappointed relative to the deal-fueled run-up, and Wall Street “wants more details on Google revenue.” Shares slid over 6% in after-hours trading Thursday and the decline accelerated to more than 8% Friday morning.

Why the market is suddenly demanding receipts

The Marvell selloff is the latest signal that the AI trade is rotating from announcements to execution. For two years, headline partnerships and mega-valuation financing rounds were rewarded on sight. In 2026, the mood has shifted: Nvidia paused its AI-cloud revenue-sharing deals after internal antitrust warnings; hyperscaler capex is being stress-tested against actual returns; and investors increasingly treat conditional, milestone-based revenue as worth less than its face value until it shows up in a 10-Q.

There’s a structural reason for the skepticism, too. The Google agreement’s economics are back-loaded by design — the $120 billion is a ceiling tied to purchasing targets through fiscal 2033, not a committed order book, and the warrant’s vesting schedule mirrors that conditionality. If Google’s TPU volumes come in below plan, or if it rebalances workloads between Marvell-designed silicon and its long-standing Broadcom partnership, Marvell’s revenue curve flattens with little warning. Analysts including JPMorgan’s Harlan Sur pressed on exactly this during the call.

The company’s answer, in effect, is patience plus a date: Marvell says the full custom-silicon roadmap will be laid out at its Investor Day on October 6, 2026, where it intends to bridge the gap between the $120 billion headline and its audited forecast. Murphy’s fiscal 2028 figure of ~$18 billion already embeds a substantial custom ramp; if October’s disclosure quantifies the Google contribution quarter by quarter, the stock has a concrete catalyst to stabilize.

What to watch

Three things will determine whether Friday’s drawdown looks like a buying dip or an early warning:

1. October 6 Investor Day. Expect management to break custom-silicon revenue out by customer cohort and fiscal year. The more granularity on Google specifically, the better the re-rating odds.
2. Data center mix in the October quarter. Guidance implies total revenue of $3.15 billion; anything confirming the custom ramp started in 2H FY27 takes timing risk off the table.
3. Broadcom’s parallel disclosure. Google now has two major external silicon partners. Watching how Broadcom characterizes its own Google-related pipeline will reveal whether Marvell’s win came out of Broadcom’s share or out of incremental TPU volume.

The irony of Friday’s trading is stark: a company growing revenue 37% with record data center demand, raised guidance, and a signed agreement worth up to $120 billion in conditional revenue lost roughly a tenth of its market value in a day — not because the fundamentals deteriorated, but because the market has learned that in this cycle, deals don’t count until the revenue prints.