Marvell: the 50x forward P/E is not this year's

The screen says Marvell trades at about 50 times forward earnings. That multiple uses roughly the next twelve months of earnings. On the fiscal year the company is in the middle of, my estimate of $4.15 a share puts it at 66 times. The gap between those two numbers is the whole argument of this analysis: the price is already living in next year.

At the October 2, 2026 close of $272.29 the shares trade 97.0% above my triangulated Base value of $138.20, a downside of 49.2%. The Bear value of $75 is 72.4% below the price, and the Bull value of $251 is 7.9% below it, so even my optimistic case sits under today's quote. The required margin of safety (MOS) is 50%, which puts the maximum buy price at $69. The shares would have to fall about 75% to get there. This is not a verdict on the business. Marvell is one of the strongest suppliers to AI data centers. It is a verdict on what the price already contains.

Route: SF-15 Semiconductors and ecosystem, subtype fabless. Economic classification: Quality Growth (8/12). Primary method: a multiple of fiscal 2027 free cash flow per share. Secondary method: ten-year free cash flow model. Marvell's fiscal year ends close to January 31, so fiscal 2027 runs from February 2026 to January 2027. Values are present fair value estimates, not 12 month price targets.

What Marvell is

Marvell designs the chips that move, switch, store and process data in data centers: custom AI accelerators and related chips for hyperscalers, electrical and optical interconnect (the DSPs and drivers inside high speed links), Ethernet switching, and storage controllers. It does not make chips itself; foundries do. The Data Center segment was $2.17 billion of the $2.74 billion second quarter revenue (79%, up from 74% a year earlier) and grew 46%. The rest, Communications and other, grew 10%.

Two acquisitions changed the shape of the company this year. Celestial AI, a photonic interconnect developer, closed on February 2, 2026 for about $1.3 billion in cash and about 24.5 million shares, with more cash and shares payable through fiscal 2029 if revenue milestones are met. XConn Technologies closed on February 10 for about $280 million in cash and 2.1 million shares. First-half acquisition outlays net of cash acquired were $1,270.9 million. Goodwill and acquired intangibles now total $16.2 billion, 59% of total assets and 88% of equity.

The other change is Google. In late July 2026 Marvell signed an expanded custom chip agreement with Google and issued it a warrant, disclosed on August 18, for 58,970,907 shares at $206.58, expiring August 18, 2033. A first slice of 1,360,867 shares vests over the first year. The other 57,610,040 shares vest one tranche per $500 million of custom product revenue Marvell recognizes from Google, from the third quarter of fiscal 2027 through fiscal 2033. Google pays $206.58 a share on exercise, so full exercise would bring in $12.2 billion and would require $120 billion of purchases. The purchases are at Google's discretion.

I could not verify customer concentration percentages for fiscal 2027, and the first draft gave none. The release itself warns that major customers are an increasing share of revenue.

The latest quarter

Management raised its revenue outlook for both fiscal 2027 and fiscal 2028 for the second quarter in a row. Coverage of the call puts the new figures at about $12 billion and $18 billion, against $8.2 billion in fiscal 2026. The Q3 gross margin guidance is lower than Q2 at both ends of the range, which fits custom silicon mix.

Why free cash flow needs the stock compensation line next to it

Free cash flow was the same in fiscal 2025 and fiscal 2026, to within $0.2 million, while revenue grew 42% and the share count rose. The first draft called the starting point a low base. It was a flat base, and free cash flow per share did not grow in fiscal 2026. In fiscal 2026 SBC was $590.8 million, so the hard figure after SBC is $805.6 million, 57.7% of the $1,396.4 million headline. In the first half of fiscal 2027 SBC was $533.8 million against free cash flow of $961.9 million, and in the second quarter it was 68% of free cash flow.

My valuation uses consensus free cash flow, which adds SBC back, and charges SBC through the multiple and a higher required margin of safety. The adversarial review shows what happens if the full cost is deducted instead. Fiscal 2027 free cash flow consensus is about $2.65 billion (third-party datasets, which I could not reproduce). The first half delivered $961.9 million, so the second half has to deliver about $1.69 billion, or $844 million a quarter, against $478.8 million in the second quarter. That is plausible with the Q3 revenue outlook 15% above Q2 and a custom ramp, but it is not yet in the bank.

Balance sheet and shares

Net debt is $1,030 million: $4,962.9 million of long-term debt less $3,932.8 million of cash. The first draft listed cash and borrowings but never used them in a bridge. I leave out operating leases, because rent is already inside operating cash flow. Two debt-like items are not quantified in the figures I could verify and are excluded from the bridge: the contingent consideration liability for Celestial and XConn (revalued upward by $433.7 million in the first half) and the forward stock purchase contract. Each $1 billion of such items would cost $1.05 a share.

The share count has three layers. Basic weighted shares of 897.4 million already include NVIDIA's $2.0 billion Series A convertible preferred stock as if converted, because Marvell reports under the two-class method: common shares outstanding are 876.9 million, which implies about 20 million preferred-equivalent shares. The 921 million diluted figure includes it too, so the first draft's list of additional dilution that named the preferred conversion counted it twice. What is genuinely extra is the Google warrant (14.2 million shares at today's price by the treasury method, up to 59.0 million shares if fully vested) and the Celestial and XConn earn-outs, which are payable partly in shares. I use 950 million economic shares for all per share values: 921 million, plus about 14 million for the warrant, plus about 15 million for earn-out shares and award growth. The last part is my estimate.

Buybacks do not offset it. Marvell spent $400 million on repurchases in the first half but also paid $365.2 million of tax on employee share settlements and $107.7 million of dividends. The dividend is $0.06 a quarter, a yield of 0.09%, with the next ex-dividend date on October 9, 2026.

Primary valuation: a multiple of free cash flow per share

Fiscal 2027 consensus free cash flow of $2.65 billion divided by 950 million shares is $2.79 a share. The first draft gave a Base of about $140 with no inputs. That value equals a multiple of 50x on this figure, so I use 50x as an explicit assumption. It is not the output of a build: Marvell is growing fast and has a strong design pipeline, which supports a multiple well above a mature chip company's 20x to 25x, but I cannot derive 50x from first principles. The result is $139.47. The grid tests the assumption.

Sensitivity: value per share

No cell reaches $272.29. The most generous cell, 60x on $3.00 billion, gives $189.5.

Secondary valuation: ten-year free cash flow model

The secondary method starts from fiscal 2027 free cash flow of $2.65 billion, grows it along an explicit path, discounts it at 9.5%, adds a terminal value at 2.5% growth, deducts net debt of $1.03 billion and divides by 950 million shares. Growth runs 45%, 35%, 25%, 20% and 15% in years 1 to 5, then 12%, 10%, 8%, 6% and 5%. Free cash flow reaches $3.84 billion in fiscal 2028, $5.19 billion in fiscal 2029 and $13.25 billion in year 10. Terminal value is 60% of the total. The result is $136.29.

That is 28% below the consensus fiscal 2029 figure of $7.17 billion, so the Base path is deliberately more cautious than the analysts. The first draft described a DCF range of $140 to $155 and then used about $165 in its triangulation table. Neither the range nor the $165 follows from its own inputs. With 921 million shares the same model gives $140.6, and with 950 million, $136.3.

Sensitivity of the Base model: value per share

The first draft's grid had $174 to $205 in the 8.5% row and $102 to $116 in the 10.5% row. Those cells do not reproduce.

Market multiples and growth diagnostics

The shares trade at about 90 times trailing GAAP earnings, 98 times fiscal 2027 consensus free cash flow on an enterprise value of $259.7 billion, and 21.6 times the roughly $12 billion of revenue management is now pointing to for fiscal 2027. A synchronized multiple history and a peer table are not available from free sources, so I do not use them.

The growth valuation cross-check (SM-19) is a diagnostic only. Fiscal 2026 non-GAAP EPS was $2.84. The first half of fiscal 2027 delivered $1.75 and the third quarter is guided to $1.10. With about $1.30 for the fourth quarter, which is my estimate, the year comes to roughly $4.15, up 46%. At $272.29 that is 65.6x and a forward PEG of about 1.4. The first draft divided a 50x multiple, which is a next-twelve-months figure, by growth of 48% for the fiscal year, and got 1.0. The two numbers cover different periods. On next-twelve-months earnings of about $5.40 (implied by the quoted 50x, which I could not reproduce) the PEG is about 1.6. Growth durability: structural with execution risk. Cyclical distortion: medium, because demand rests on hyperscaler capital spending. Flags: accounting basis mismatch (non-GAAP EPS is $0.94 and GAAP is $0.33), peer period mismatch. SM-19 does not set fair value.

The free cash flow growth quality check (SM-20) starts at $1,396.4 million in fiscal 2026 and ends at the consensus $7,172 million in fiscal 2029 (third-party estimate): an increase of $5.78 billion, or 414%, and a compound growth rate of 72.5% a year. Per share, on 868 million fiscal 2026 diluted shares and 950 million for fiscal 2029, that is $1.61 rising to $7.55, or 68% a year. The low base flag is therefore medium rather than high, because fiscal 2026 is flat on fiscal 2025, but revenue itself grew 42% in that year, so the cash conversion of the growth is what has to improve. On the 950 million share value of $258.7 billion, free cash flow yields are 0.5% for fiscal 2026, 1.0% for fiscal 2027 and 2.8% for fiscal 2029. Even the consensus endpoint is not cheap. Distortion flags: low base medium, acquisition medium to high, dilution high, cycle medium. Estimate reliability: medium. Classification: structural with execution risk. I do not publish a numeric SM-20 score, since the first draft's 68/100 came with no components. SM-20 does not set fair value, MOS or the verdict.

Valuation triangulation

The two methods differ by 2.3%, well under the 20% warning threshold. That agreement is not independent evidence, since I picked the 50x multiple to match the draft's central value, and both methods lean on the same fiscal 2027 consensus. The first draft showed the secondary value as $165 to reach a Base of $150; on its own DCF inputs the figure is $136. The Base is $138.20.

What is the market already pricing in?

At $272.29 and 950 million economic shares, market value is about $258.7 billion and enterprise value about $259.7 billion. At my 50x multiple the price needs fiscal 2027 free cash flow of about $5.2 billion, 95% above consensus. In the model, the price equals the Base path at roughly twice the cash flow in every year (fiscal 2028 near $7.6 billion instead of $3.8 billion), or the Base path at a discount rate of about 6.3%. The full consensus path through fiscal 2029, with growth continuing after it, needs a discount rate of about 8.2%.

A buyer today needs consensus to be met and then beaten: custom silicon ramping to its full potential, connectivity staying in short supply, gross margin holding near 58%, the Celestial and XConn investments paying back, and the Google warrant being earned in full, which means very large purchases. The price is not asking for a good outcome. It is asking for a result above the analysts' plan.

Bear, base and bull scenarios

The Bear path grows 30%, 22%, 16%, 12%, 10%, 8%, 6%, 5%, 4% and 3% from $2.65 billion. The Bull path takes the fiscal 2029 consensus figure as given and interpolates fiscal 2028 at $4.36 billion (my estimate, the geometric midpoint), then grows 33%, 25%, 19%, 14%, 11%, 8%, 6% and 5%. Terminal value is 52%, 60% and 65% of the total. I assign no probabilities. The Bear is not a collapse: free cash flow still grows 2.9 times in ten years. The first draft gave $75 and $250 with no inputs; these paths reproduce both values.

How the buy price is set

A Confidence C score carries a base MOS of 40%. I add 5 points for customer concentration and 5 points for the contingent consideration, warrant and preferred stock structure, and nothing for the balance sheet, because net debt is only 0.4% of the company's value. The required MOS is 50%. The first draft added a single 5 points for risk; with a flat fiscal 2026 free cash flow, an earn-out that was revalued upward by $434 million in six months and SBC at 68% of free cash flow, one modifier is too few. Applied to the $138.20 Base, the maximum buy price is $69, and a stronger 60% margin marks a strong-buy reference near $55. The shares would need to fall about 75% from $272.29 to reach $69.

Dividend and capital return

The dividend is $0.24 a year, a yield of 0.09%, so it does not matter for the investment case. Capital return is also smaller than the stock compensation it offsets: first-half buybacks were $400 million, SBC was $533.8 million and share withholding for taxes cost $365.2 million.

Reasons to own MRVL

• Broad exposure to AI infrastructure. Custom silicon, interconnect, switching and storage are all sold to the same small group of buyers.
• Deep hyperscaler design relationships. Custom programs take years to design in and years to replace, and the Google agreement now ties a customer's purchases to equity.
• Accelerating revenue. Data Center revenue grew 46%, and the Q3 outlook is above 50% growth for the company.
• New connectivity capability. Celestial AI and XConn broaden the optical and scale-up portfolio.
• Very large absolute cash flow potential. Management's $18 billion fiscal 2028 revenue outlook would multiply free cash flow if margins hold.

What could go wrong?

• Thesis breaker: free cash flow per share fails to scale fast enough. Free cash flow was flat in fiscal 2026 on revenue up 42%, and a repeat is the central risk to a price that needs it to double.
• Valuation. At 98 times fiscal 2027 consensus free cash flow, good execution can still produce weak returns. The 52 week high is $329.88 and the shares are 17% below it.
• Customer concentration. The company warns that major customers are a growing share of revenue, and custom programs can be delayed or lost.
• SBC and dilution. SBC was 11.9% of revenue in the latest quarter, diluted shares are up 5.8% in a year, and the Google warrant covers up to 59.0 million more.
• Acquisition risk. Goodwill and intangibles are $16.2 billion, and the earn-outs for Celestial and XConn are payable in cash and shares through fiscal 2029.
• Supply and cycle. Foundry capacity, product ramps and hyperscaler capital spending can all move away from consensus.

Management execution

Management has beaten its revenue guidance by a small margin in each recent quarter, raised its outlook twice in a row and signed a long custom agreement with Google. The offsets are a record of paying for growth with shares, a first-half earn-out revaluation of $433.7 million, and per share cash generation that did not grow in fiscal 2026. Assessment: 16/20, good. The next test is the Investor Day on October 6, 2026, where management will present its long term targets.

Stock Analyza scorecard

The first draft scored 10/12, which mechanically means Compounder, and then overrode it in words. Return on capital is 0/2 because GAAP operating margin is 16.8% on $18.5 billion of equity and $5.0 billion of debt, mostly goodwill and acquired intangibles, and free cash flow per share was flat in fiscal 2026 (1/2). Business quality is 72/100: moat and pricing power 18/20, return versus cost of capital 10/20, balance sheet 11/15, earnings and cash quality 8/15, growth and runway 15/15, management and capital allocation 8/10, dilution and governance 2/5. The first draft gave dilution and governance 5/5 while flagging dilution red, and economic return 15/20 on a return profile near the cost of capital. The valuation score is 18/100: discount to Base 0/40, protection against the Bear case 0/20, agreement between methods 12/15, historical valuation 3/10, reverse valuation 0/10, data and model quality 3/5.

Confidence score

Reporting is timely and the balance sheet is clear. The grade is held down by customer concentration I could not quantify, a primary multiple that is an assumption, a $558 million gap between GAAP and non-GAAP quarterly profit, and a range of $75 to $251 between Bear and Bull.

What I would watch from here

Today three readings are red: SBC (42% of free cash flow in fiscal 2026 and 68% in the second quarter), diluted share growth (+5.8% in a year) and free cash flow per share (flat in fiscal 2026). Gross margin is yellow, because the Q3 outlook is 57.5% to 58.5% against 58.9% in Q2, and integration is yellow because of the earn-out revaluation. Revenue growth (36.5%), Data Center growth (45.7%) and custom silicon ramps (management expects acceleration in the second half) are green. First half free cash flow of $961.9 million is on track for the $2.65 billion Base only if the second half averages $844 million a quarter. The next review points are the Investor Day on October 6, 2026 and the third quarter report, expected around December 1, 2026 (an estimate).

The three most important thesis breakers

1. The loss or major delay of a custom silicon program.
2. Free cash flow per share misses materially despite revenue growth.
3. Dilution, including the warrant and earn-outs, outruns the value it creates.

Adversarial review

I recomputed the DCF, the primary value, both grids, the triangulation, the MOS, the free cash flow growth figures and the share bridge, and checked the reported inputs against the second quarter release, the fiscal 2026 results and the closing price. The reported figures reconcile. The corrections came in the valuation model, the scores and several statements.

None of these corrections changes the direction of the verdict. They lower the Base by $11.80 and the maximum buy price by $13.40. The one correction that moves in the company's favor is the Bull, which I rebuilt on the consensus path and which reproduces the draft's $250.

The strongest counter-thesis, for the bulls: Marvell's custom AI silicon and optical interconnect opportunity could be much larger and longer lived than consensus, as the Google agreement and the $120 billion vesting ceiling suggest. If free cash flow keeps compounding at exceptional rates well beyond fiscal 2029, the price is reasonable. The response is that the Bull case is exactly this, it gives $251, and today's price is above it.

Audit flags: consensus free cash flow for fiscal 2027 and 2029, the revenue outlook figures for fiscal 2027 and 2028, and the 50x forward P/E come from third-party coverage that I could not reproduce from free sources. The fiscal 2028 consensus free cash flow was not available, so I interpolated it. The fourth quarter EPS of $1.30 is my estimate. The SBC path is my estimate: the Q3 outlook implies about $310 million for the quarter, and I assume about $320 million in Q4. The stress test partly overlaps with dilution, which the share count already carries. Customer concentration percentages for fiscal 2027, the fair value of the contingent consideration and the forward stock purchase contract, and a peer table are unavailable. The publication gate status is PASS WITH WARNING because long-duration growth, dilution and consensus free cash flow remain uncertain.

Final verdict: reduce at $272.29

Marvell is a high-quality AI semiconductor business with accelerating revenue, a new anchor agreement with Google and a record of raising its own outlook. The shares, though, trade 97% above my Base and above my Bull, and they need free cash flow per share to roughly double against a base that did not grow last year. Google can buy the stock at $206.58 if it keeps buying the chips. At $272.29 the rest of us would be paying 32% more for the same shares.

Verdict: REDUCE at $272.29. Triangulated Base value $138.20, downside 49.2%, Confidence C (13/25). Required margin of safety 50%, maximum buy price $69.

The case changes if free cash flow per share starts compounding faster than SBC and the share count, if the October 6 Investor Day raises the long-term targets above the consensus path, or if the price falls toward the buy zone with the business still executing.

Sources

Primary sources include Marvell's second quarter fiscal 2027 release of August 27, 2026, the Form 10-Q for the quarter ended August 1, 2026, the Form 10-K for fiscal 2026, the fourth quarter fiscal 2026 release, the earnings call coverage and the Google warrant disclosure of August 2026. The October 2, 2026 close comes from Yahoo Finance. Consensus estimates and trailing multiples come from public aggregators cited in the framework analysis and are labeled as estimates.

The reference market price is the October 2, 2026 regular session close of $272.29. A broker app showed $273.04, which is the after-hours quote. Balance sheet and cash flow figures are as of August 1, 2026, and the common share count of 876.9 million is from StockAnalysis. The 950 million economic shares, the ten-year paths, the 50x multiple and the SBC stress test are analytical estimates, not company guidance.