Marvell Q2 FY2027: Record Revenue, Guidance Raised Twice, Stock Down

© 2026 Street Alpha. All rights reserved.

Disclaimer: This article is for informational and educational purposes only. It does not constitute financial, legal, or geopolitical investment advice. The views expressed are based on public intelligence and analytical frameworks as of the publication date and are subject to change without notice. Readers should conduct their own research or consult professional advisors before making any investment decisions.

Five Numbers From the Quarter

• Revenue $2.739B, up 37% year over year, an all-time record, 1.1% above consensus
• EPS $0.94, up 40% year over year, a penny above consensus
• Data center revenue $2.17B, up 46% year over year, now 79% of the company
• Gross margin 58.9%, in line with guidance
• Operating margin 36.6%, up 180 basis points year over year

Data center growth accelerated from 27% last quarter to 46% this quarter. Operating expenses grew slower than revenue, which is why the operating margin expanded.

Everything else came in roughly as expected: communications revenue of $568M (up 10%), operating expenses of $611M (slightly above guidance), and operating cash flow of $606M.

Guidance: Both Fiscal Years Moved Up

• Q3 revenue $3.15B, 3.8% above consensus, growth above 50% year over year
• This year (FY2027) roughly $12B, up 45%, versus the $11.5B target given three months ago
• Next year (FY2028) roughly $18B, up 50%, versus the $16.5B target given three months ago
• Q3 gross margin 57.5% to 58.5%, down about 90 basis points sequentially
• Custom silicon revenue more than doubles next year

The FY2028 target moved up $1.5B in a single quarter. It was $15B in May and $18B in August — a 20% raise across two quarters.

The tax rate goes from 11% to about 13% next year, and this year’s operating expense target moved from $2.45B to $2.55B.

Why the Stock Fell

Gross Margin Doesn’t Improve Next Year

CFO Dan Durn was asked why Q3 gross margin steps down sequentially. He said the driver is product mix, with custom silicon ramping hard. Then he gave his read on next year: FY2028 gross margin lands in the same range as the second half of this year, 57.5% to 58.5%.

Revenue grows 50% next year. Gross margin stays flat.

The cost of that sentence is easy to quantify. Moving revenue from $16.5B to $18B adds $0.84 to EPS. Gross margin, operating expenses, and the higher tax rate together take back $0.26. FY2028 EPS goes from $6.10 to $6.68 — a 9.5% raise.

Revenue up 9%, earnings up 9.5%. One for one.

A stock trading at 53 times this year’s guided earnings needs to see earnings growing considerably faster than revenue.

The custom silicon business model prevents that. When you design a chip for a hyperscaler, the customer captures most of the design value and your pricing power is compressed. Standard products — optical DSPs, switch silicon, broadband analog — carry much higher margins. As custom moves up from roughly 25% of revenue this year, the blended margin moves down.

Management’s long-term operating margin target is 38% to 40%. Q4 of this year is when the company first enters that range, and mid-FY2028 is when it reaches the top end. That number is the ceiling.

The Google Deal Adds Nothing Next Year

On August 19, Marvell disclosed an agreement with Google that includes a warrant letting Google buy 58.97 million shares at $206.58, worth about $12.2B. Full vesting requires $120B in cumulative purchases, with the agreement running through 2033. The stock rose 9.85% that day.

On the call, a JPMorgan analyst did the arithmetic ($120B over six years is about $18.5B a year) and asked when that revenue shows up. Matt Murphy’s answer: revenue from programs covered by the agreement through FY2028 is already inside the custom revenue target the company had previously given. The real impact starts in FY2029.

That 9.85% on August 19 was the market pricing in 2029 through 2033. The August 27 report confirmed nothing changes next year.

The Most Important Number Got Pushed Out Five Weeks

Murphy was asked repeatedly to size the FY2029 custom business. He declined, saying he would save it for the Investor Day on October 6. Durn added that the company’s long-term financial model will also be reset that day.

A company that just signed a $12.2B warrant refused to tell the market what its revenue looks like three years out. Holders were left with two options: wait five weeks in an information vacuum, or sell first.

Communications Drops Sharply Next Quarter

Q3 guidance has the communications segment down 13% to 15% both sequentially and year over year. That means the entire $3.15B quarter depends on data center growing more than 20% sequentially. Revenue concentration is increasing.

Three Things the Market Overlooked

The $1.5B raise for next year is driven mainly by scale-up optics and switching. Murphy called this the most meaningful piece of the raise, and elsewhere said connectivity was the largest net driver. Scale-up optics was guided to roughly $300M for FY2028 last quarter, and that number is now accelerating. His words: the year after next is going to be “way larger than I sort of could have ever comprehended” when he first looked at Celestial AI.

Marvell is building both NPO and CPO packaging, supports three different modulator technologies, and supports all three scale-up protocols — UALink, ESUN, and NVLink Fusion. Customers haven’t decided which path to take, and Marvell is on every path.

Three connectivity product lines are each approaching $1B in annualized revenue: broadband analog, data center interconnect modules, and scale-out switching. All three are at or ahead of that trajectory, with scale-out switching more than doubling this year.

CXL is benefiting from expensive memory. Murphy called CXL a “home run” and said it’s deployed at multiple hyperscalers in extremely high volumes. Customers are shifting toward memory expansion because memory is scarce. DRAM shortages are a cost problem for most chip companies; for Marvell’s CXL business, they are a source of demand.

Operating cash flow of $606M came in below expectations. Durn explained it as capacity prepayments to suppliers, roughly $1B for the full year — guidance he had already given last quarter. Inventory declined sequentially, the company bought back $200M of stock in the same period, and net debt to EBITDA sits at 0.27 times. Revenue up 37%, inventory flat to down, and capacity locked in ahead of time. When supply is tight, that’s competitive behavior.

Valuation

Every input below is anchored to guidance given on this call.

Probability-weighted fair value: $210.

The base case operating margin of 40.3% matches Durn’s comment about reaching the top of the 38% to 40% range during FY2028. The bear case sits at the bottom of that range. The bull case at 42.3% already exceeds the company’s own target and requires management to raise the long-term model at Investor Day.

At $222.58, the stock trades at 53 times this year’s guided earnings and 33 times next year’s. Put 25 times on what the company has actually earned over the last twelve months and you get roughly $95 — meaning about 57% of the current price rests on earnings that don’t exist yet.

On odds: the bull case is 23% upside, the bear case is 38% downside. The probability-weighted expected return is negative.

A Word About the Multiples

The three multiples above (24x, 30x, 36x) reflect what I observe the market currently willing to pay.

In my June 3 piece I published a two-year target of $503 using 40x multiple. The stock touched $329.88 on June 18 and fell to $162.90 by July 29 — down 51% in six weeks. Nothing changed in the fundamentals during that stretch. What changed was how much the market would pay for earnings three years out.

Valuations for companies like Marvell depend heavily on profits that haven’t been earned yet. When sentiment expands, that structure works as leverage. When sentiment contracts, it works the same way in reverse. July’s semiconductor drawdown demonstrated the second half of that.

If semis run hot again next year and 36x becomes 45x, the bull case is $343. The reverse applies equally.

What I can assess is earnings. This report was clear on that point: revenue guidance went up, gross margin guidance did not, and earnings rose by the same percentage as revenue. That holds true in any multiple environment.

The levels below are built on earnings.

Reference entry/add points

• Above $222 — no adds. The price sits above fair value of $210 and the weighted expected return is negative.
• $200 to $215 — small adds, no more than a third of the intended position. This range crosses fair value, and $206.58 is Google’s strike price. Below that line Google has no reason to exercise, which creates a psychological floor.
• $180 to $200 — main entry. There’s margin of safety against the $201 base case.
• $160 to $178 — size up. The July 29 low of $162.90 has already been tested once.

The reason for holding back two-thirds is October 6. Murphy said FY2029 custom revenue “definitely goes higher” and is “a lot larger than anybody’s been modeling,” then declined to give a number. If that day brings $14B or more, the bull case probability rises and fair value goes above $230. If it’s $11B to $12B, that’s in line rather than a surprise.

Conclusion

Marvell keeps beating on revenue. All three major hyperscalers are covered, the position in optical interconnect and scale-up switching is stronger than the market credits, CXL has an unexpected tailwind from memory scarcity, the balance sheet is clean, and inventory is falling while the company locks in future capacity.

All of that is true.

Also true: this company’s ability to convert technical advantage into margin advantage is weaker than I had assumed. The low-margin structure of custom silicon means that as revenue grows, each dollar of revenue carries less profit. The 38% to 40% range is the ceiling management drew for itself.

A company that grows revenue 50% and earnings 50% is worth a very different amount than one that grows revenue 50% and earnings 80%. Before August 27 the market thought it owned the second one. Now it knows it owns the first.

Fair value $210, stock at $222. No adds here. Wait for sub-$215, and keep the ammunition for October 6.

Refer a friend