Marvell (Nasdaq: MRVL) crashes 10% as Google AI payoff shifts to FY29

Marvell Technology, Inc. (Nasdaq: MRVL) fell 10.3% on August 28 despite reporting record fiscal second-quarter revenue, raising its fiscal 2027 and fiscal 2028 sales targets and guiding to another double-digit sequential revenue increase in Q3. Revenue reached US$2.739 billion, up 37% year on year, while Data Center revenue increased 46% to US$2.172 billion and represented 79% of the company. Investors instead focused on the timing of Marvell’s expanded custom-silicon agreement with Google, under which warrant vesting could ultimately be associated with as much as US$120 billion of cumulative purchases through fiscal 2033 but materially larger revenue contributions are not expected until FY29. MRVL closed Friday at US$216.62, leaving the stock down about 8.6% over five sessions but still roughly 150% higher during 2026, making expectations rather than current growth the central investment problem.

Why did Marvell fall 10.3% after reporting record Q2 revenue?

The earnings themselves were solid. Revenue increased from US$2.006 billion to US$2.739 billion, while non-GAAP net income reached US$865.9 million and adjusted diluted EPS increased to US$0.94.

Wall Street had been looking for roughly US$2.72 billion of revenue and US$0.93 of adjusted EPS, so Marvell cleared both estimates. The company also generated US$605.5 million of operating cash flow during the quarter.

Data Center remained the dominant growth engine. Revenue from the segment reached US$2.1715 billion compared with US$1.4905 billion a year earlier, an increase of 46%. It also rose 18% sequentially and accounted for 79% of total company revenue compared with 74% a year earlier.

The problem was that investors had already priced in something stronger.

MRVL had risen from US$61.44 at its 52-week low to more than US$250 before the results and has been one of the largest semiconductor winners of 2026. In that context, simply beating quarterly estimates was not enough.

The shares closed at US$216.62 on August 28 compared with US$241.45 on August 27, while trading volume approached 49 million shares. MRVL is about 8.6% below its August 21 close of US$237.04 but still approximately 14.5% above its July 27 close of US$189.17.

The market has therefore reduced near-term expectations without reversing the enormous rerating driven by AI custom silicon.

What does Marvell’s US$3.15bn Q3 guidance imply for growth?

Marvell expects fiscal Q3 revenue of US$3.15 billion, plus or minus 5%.

At the midpoint, that represents approximately 15% sequential growth from Q2.

It would also be the first time quarterly revenue has exceeded US$3 billion and would extend the acceleration management says is developing in both connectivity and custom silicon during the second half.

Non-GAAP diluted EPS is expected at US$1.10, plus or minus US$0.05, while non-GAAP gross margin is guided to 57.5% to 58.5%.

The company has also increased its longer-range expectations. Fiscal 2027 revenue is now expected around US$12 billion, up from the previous US$11.5 billion outlook, while fiscal 2028 revenue is targeted around US$18 billion compared with US$16.5 billion previously.

Moving from US$12 billion to US$18 billion would imply approximately 50% revenue growth in a single year.

That kind of trajectory explains why investors have been willing to pay an exceptional valuation for Marvell. It also creates a very high bar because the stock increasingly depends on revenue expected several years from now rather than merely on the current quarter.

Management has indicated that custom-silicon revenue should accelerate materially in H2 FY27 and more than double during FY28. The October 6 Investor Day is therefore likely to matter more to valuation than an ordinary quarterly conference call because investors want more visibility into the composition and timing of those long-term targets.

Why did the Google agreement disappoint investors after initially exciting them?

The agreement itself is large enough to transform Marvell’s custom-silicon business.

Marvell and Google expanded their collaboration across custom products linked to Google’s TPU ecosystem, including AI inference accelerators, storage controllers, network-interface controllers, memory-interface controllers and near-memory compute.

Google received a warrant to purchase as many as 58.97 million Marvell shares at US$206.58 each.

Only about 1.36 million of those shares vest on a time basis. The remaining shares vest through 240 purchase-based tranches, with each tranche associated with US$500 million of qualifying custom-product revenue.

Multiplying 240 tranches by US$500 million produces a maximum associated revenue threshold of US$120 billion through the end of Marvell’s fiscal 2033.

That headline number helped send MRVL sharply higher when the agreement was announced.

The August earnings call reset the timing expectations. Chief executive Matt Murphy indicated that Marvell’s current FY27 and FY28 custom-revenue objectives already include some Google contribution, while the more significant acceleration from the agreement is expected during FY29.

That distinction is what disappointed investors.

The commercial opportunity remains enormous, but the share price had moved as though much more of that opportunity might enter the financial statements sooner. Investors are now discounting a longer wait before Google materially changes the earnings base.

How much dilution could Google’s Marvell warrant eventually create?

The warrant structure provides Google with unusually strong economic alignment if the commercial relationship reaches very large scale.

Marvell currently has roughly 899 million basic shares outstanding based on current market data and the company’s Q3 share-count guidance. The maximum 58.97 million Google warrant shares would therefore equal approximately 6.6% of the current basic share count.

The actual dilution could be lower because vesting depends heavily on qualifying purchase volumes and the exercise terms, but the theoretical maximum is large enough that investors should include it when assessing the economics of the agreement.

The exercise price is US$206.58.

That is only about 4.9% below Marvell’s August 28 closing price of US$216.62, meaning the warrant is already near the money based on the current stock price. Google nevertheless receives most of the potential shares only if substantial commercial purchases occur.

This creates a trade-off for existing shareholders.

If Google generates enough purchases for all US$120 billion of revenue-linked vesting thresholds to be reached, the dilution would occur alongside extraordinary cumulative revenue for Marvell. If purchase volumes remain much lower, fewer warrant shares should vest.

The relevant question is therefore not simply whether the warrant is dilutive. It is whether the operating profit generated from the associated business comfortably exceeds the value transferred through that equity participation.

Can Marvell protect margins while AI revenue accelerates?

Gross margin is one of the more important constraints on the valuation.

Marvell reported a GAAP gross margin of 53.1% and non-GAAP gross margin of 58.9% in Q2. For Q3, the company expects GAAP gross margin between 52.9% and 53.9% and non-GAAP margin between 57.5% and 58.5%.

At the midpoint, adjusted gross margin would decline roughly 90 basis points sequentially.

That pressure matters because custom silicon can have different economics from highly standardised merchant semiconductor products. Marvell can generate large revenue volumes and strong strategic positions while still producing lower gross margins than investors associate with some other AI-chip businesses.

The company also remains dependent on advanced manufacturing partners, packaging, substrates and other constrained parts of the semiconductor supply chain. Rapid growth can therefore increase revenue while simultaneously creating pressure on input costs and operating expenses.

Marvell expects approximately US$655 million of non-GAAP operating expenses in Q3, meaning operating leverage becomes another important variable as revenue moves above US$3 billion.

The most favourable outcome is accelerating custom and connectivity sales accompanied by reasonably stable high-50s adjusted gross margins. If revenue reaches the long-term targets while margins fall substantially, the earnings power implied by the US$12 billion and US$18 billion revenue figures would be lower than headline growth suggests.

Is Marvell expensive after the 10% sell-off?

At US$216.62 and roughly 899 million shares, Marvell’s market capitalisation is close to US$195 billion.

Reuters cited an LSEG 12-month forward P/E of approximately 58 times even after the post-results decline, compared with roughly 32 times for Broadcom.

That valuation makes MRVL particularly sensitive to timing.

A stock at almost 60 times expected earnings can still generate strong returns if revenue and earnings compound exceptionally quickly. The problem arises when large pieces of the expected growth move further into the future without an equivalent reduction in the discount rate investors apply to those earnings.

Marvell is still up dramatically during 2026, and Friday’s close remains more than three times the 52-week low. The stock is also approximately 34% below its 52-week high around US$330, showing how volatile the valuation has become as investors repeatedly revise their assumptions about AI custom silicon.

The October Investor Day therefore has an unusually important role. Investors need more visibility into the customers, product categories, revenue ramps and margin implications underlying the US$12 billion FY27 and US$18 billion FY28 targets.

If management can make those numbers more tangible, the current multiple becomes easier to defend. If another substantial portion of expected custom revenue is pushed into FY29 and beyond, the market may continue demanding a lower valuation despite strong near-term growth.

Marvell stock key takeaways after the 10.3% post-earnings fall

• Marvell closed 10.3% lower at US$216.62 on August 28 even though Q2 revenue reached a record US$2.739 billion and increased 37% year on year.
• Data Center revenue increased 46% to US$2.172 billion and now accounts for 79% of total company sales.
• Q3 revenue guidance of US$3.15 billion implies approximately 15% sequential growth from Q2, while non-GAAP EPS is expected around US$1.10.
• Marvell increased FY27 revenue expectations to approximately US$12 billion and FY28 expectations to approximately US$18 billion, implying another roughly 50% expansion between those years.
• Google’s warrant can cover up to 58.97 million Marvell shares, with most vesting tied to cumulative purchases across 240 US$500 million revenue tranches through FY33.
• Significant incremental Google revenue is expected to become more visible in FY29, later than some investors had inferred when the agreement was announced.
• At approximately US$195 billion of market value and around 58 times forward earnings, MRVL still requires exceptionally strong execution despite the 10% decline.

What would strengthen or weaken the Marvell investment case from here?

The investment case would strengthen if Q3 revenue reaches or exceeds the US$3.15 billion midpoint while adjusted gross margin remains near 58%. That combination would demonstrate that custom and connectivity acceleration can translate into meaningful operating leverage rather than revenue growth obtained at progressively lower profitability.

The October 6 Investor Day could become the next major catalyst if management provides clearer customer-level and product-level bridges between today’s US$2.7 billion quarterly revenue base and the US$18 billion FY28 annual target. Greater visibility into Google’s FY27 and FY28 contribution would reduce one of the largest uncertainties created by the earnings call.

The thesis would weaken if Google-linked growth continues shifting further into the future, other hyperscaler custom-chip programmes fail to accelerate or margins contract faster than expected. Customer concentration also becomes progressively more important as large AI buyers represent a growing share of Marvell’s revenue.

The 10.3% decline does not reflect weak current demand. Data Center revenue is growing 46%, Q3 revenue is expected to increase another 15% sequentially and management has raised multi-year forecasts.

What changed is the timeline investors are willing to pay for. At US$216.62, MRVL remains priced for enormous custom-silicon success. The next stage requires Marvell to convert a headline US$120 billion Google opportunity into visible annual revenue quickly enough that the earnings catch up with a valuation that has already travelled several years into the future.