MRVL · Forward model · Google warrant case
The Google warrant case, 20 quarters out
Each segment is projected from its own operating driver, rolled up into consolidated cash flow, and discounted back to a fair value per share. The assumptions below are editable — change them and every number on this page moves with them.
Quarters carry Marvell's own fiscal labels: 2027 Q1 is the fiscal first quarter ENDED 2 May 2026, and fiscal Q4 ends in late January. What Marvell discloses is exactly two end markets - data center and communications and other - and they sum to the dollar to reported net revenue in all nine quarters carried here ($1,832.7M + $585.1M = $2,417.8M in the basis quarter). Marvell runs ONE reportable segment and the CODM assesses performance on consolidated net revenue, so there is no disclosed margin, opex or capex by end market: the two margin lines and the two capex intensities on this page are the author's, calibrated so the consolidated result reproduces the reported quarter, and are not company data. The two earliest communications quarters (2025 Q1 and 2025 Q2) are marked estimated because they are our own sum of the four legacy end markets - enterprise networking, carrier infrastructure, consumer, automotive and industrial - that Marvell later combined into the single line; every later quarter is the reported figure. What is NOT split, and is not invented here: custom XPU silicon versus electro-optics versus switching versus storage inside data center. Marvell publishes no XPU units, ASPs, wafer allocation or backlog, so a unit or capacity driver on this line would be fabrication and both verticals run on plain growth rates. Margins are stated on Marvell's non-GAAP basis (35.0% consolidated non-GAAP operating margin in the basis quarter), and stock compensation of $207.6M, 8.6% of revenue, is charged back as corporate overhead because it is a real and dilutive cost that the non-GAAP margin excludes; the resulting free cash flow margin of about 18% in the first projected quarter sits just under the $483.1M, 20.0%, actually reported. Acquisition amortisation of $225.2M a quarter is NOT charged, which is why this model values a company that earned $34.5M of GAAP net income in the basis quarter. Share count is held flat at the 915 million diluted shares guided for fiscal 2027 Q2 and does not grow for the three live dilution overhangs - NVIDIA's $2.0B convertible preferred (about 21.8M shares at $91.84), the Google warrant (up to 58,970,907 shares at $206.58) and Celestial AI contingent consideration payable partly in stock through fiscal 2029. The reference price of $245.11 is the 26 August 2026 close, not the stale $195.22 in this site's stored profile. Seasonality was tested on both verticals and carried on neither; the method and the numbers are in each vertical's notes. Fiscal 2027 Q2 reports after the US close on 27 August 2026, so this model is built on the last reported quarter and will be one print behind within hours of publication.
The upper bound Marvell itself put on paper. The 18 August 2026 warrant vests 240 tranches of 240,042 shares, one for each $500 million of Custom Products revenue from fiscal 2027 Q3 through fiscal 2033: $120 billion of cumulative custom revenue for full vesting, against a whole-company fiscal 2027 outlook of nearly $11.5 billion. Spread evenly across the 25 quarters in that window it is $4.80 billion a quarter of custom revenue alone, roughly twice everything Marvell sold in the basis quarter. This case does NOT reach that. Even at a 2.25% quarterly tilt the model ends at $51.4 billion of annual revenue, so full vesting would require a step this page cannot express - and the warrant is an incentive structure and a disclosed ceiling, never guidance. It also ignores its own dilution: the 58,970,907 shares are not added to the 915 million share count, and at full vesting they would be about 6% of the equity.
Latest: $14.03B (2032Q1E)
| Period | Value |
|---|---|
| 2025Q1 | $1.16B |
| 2025Q2 | $1.27B |
| 2025Q3 | $1.52B |
| 2025Q4 | $1.82B |
| 2026Q1 | $1.90B |
| 2026Q2 | $2.01B |
| 2026Q3 | $2.07B |
| 2026Q4 | $2.22B |
| 2027Q1 | $2.42B |
| 2027Q2E | $2.78B |
| 2027Q3E | $3.16B |
| 2027Q4E | $3.58B |
| 2028Q1E | $4.02B |
| 2028Q2E | $4.48B |
| 2028Q3E | $4.97B |
| 2028Q4E | $5.48B |
| 2029Q1E | $6.01B |
| 2029Q2E | $6.56B |
| 2029Q3E | $7.14B |
| 2029Q4E | $7.73B |
| 2030Q1E | $8.34B |
| 2030Q2E | $8.97B |
| 2030Q3E | $9.63B |
| 2030Q4E | $10.30B |
| 2031Q1E | $11.00B |
| 2031Q2E | $11.72B |
| 2031Q3E | $12.46B |
| 2031Q4E | $13.23B |
| 2032Q1E | $14.03B |
What drives each segment
Data center
Growth pathCustom AI accelerators and XPU-attach silicon for hyperscalers, plus the electro-optics, switching and storage that connect them. Three quarters of revenue and effectively all of the growth. It is programme revenue: won at design-in, ramped on the customer's schedule, and disclosed as one dollar line with nothing underneath it.
Latest: $12.70B (2032Q1E)
| Period | Value |
|---|---|
| 2025Q1 | $816M |
| 2025Q2 | $881M |
| 2025Q3 | $1.10B |
| 2025Q4 | $1.37B |
| 2026Q1 | $1.44B |
| 2026Q2 | $1.49B |
| 2026Q3 | $1.52B |
| 2026Q4 | $1.65B |
| 2027Q1 | $1.83B |
| 2027Q2E | $2.16B |
| 2027Q3E | $2.52B |
| 2027Q4E | $2.90B |
| 2028Q1E | $3.31B |
| 2028Q2E | $3.75B |
| 2028Q3E | $4.20B |
| 2028Q4E | $4.68B |
| 2029Q1E | $5.18B |
| 2029Q2E | $5.70B |
| 2029Q3E | $6.24B |
| 2029Q4E | $6.79B |
| 2030Q1E | $7.37B |
| 2030Q2E | $7.96B |
| 2030Q3E | $8.57B |
| 2030Q4E | $9.21B |
| 2031Q1E | $9.86B |
| 2031Q2E | $10.54B |
| 2031Q3E | $11.23B |
| 2031Q4E | $11.96B |
| 2032Q1E | $12.70B |
Assumptions & reasoning
- The 15.4% first sequential step is calibrated to management's two dated numbers, not to consensus. It puts this line at $2,115M in fiscal 2027 Q2, which with communications gives $2,715M against a guided $2.700B plus or minus 5%, and it lands fiscal 2027 data center at $9,066M, up 48.6% on the $6,100.3M reported for fiscal 2026 against a guided 'approximately 50%'. Fiscal 2028 comes out at $13,909M, up 53.4% against 'approximately 55%'.
- Note what the driver structurally CANNOT do: management says growth accelerates each quarter, and a decaying sequential rate cannot accelerate. The fit is to the annual totals, and the shape inside fiscal 2027 is flatter than the shape management describes - front-loaded relative to guidance in the second quarter, back-loaded relative to it in the fourth.
- Marvell publishes no XPU units, no ASPs, no wafer allocation, no capacity and no backlog for this line, and does not say what share of it is custom silicon versus optics, switching and storage. Any unit- or capacity-driven model here would be inventing the denominator, so this vertical runs on a growth rate and says so.
- Seasonality tested and NOT carried. Ratio to a centred four-quarter moving average over the nine disclosed quarters gives 1.028 for fiscal Q1, 1.001 for Q2, 0.984 and 0.964 for Q3 and 1.073 for Q4. Only fiscal Q3 has two windows to compare and they disagree by 2.0 points; Q1, Q2 and Q4 rest on a single window each, inside a line that grew 2.2 times over the sample. The apparent 11-point Q4-to-Q3 shape is the custom ramp, not a season, and it is not separable on this much history. Derived, method as stated; no factors are applied.
- Terminal growth of 1.5% a quarter, about 6% a year, is the assumption doing the most work after the exit multiple. It says the custom programmes that carry this line eventually behave like semiconductors again. Nothing Marvell has disclosed reaches past fiscal 2029, so the last eleven quarters here are argument rather than evidence.
- Celestial AI (closed 2 February 2026, $3.5B) and XConn (closed 10 February 2026) sit inside this line from their acquisition dates. Neither revenue contribution is quantified in the 10-Q, so the organic and acquired parts of the 27.2% year-on-year growth in the basis quarter cannot be separated here.
- Concentration is the risk the sliders do not express: one direct customer was 16% of consolidated revenue and one distributor 45% in the basis quarter, and the 10-K names customers vertically integrating as a live risk. The same hyperscalers are both the customer and the potential competitor.
Communications and other
Growth pathEnterprise networking, carrier infrastructure, consumer and automotive/industrial, combined into one disclosed line in fiscal 2026. A cyclical recovery off the fiscal 2025 trough that is still a third below its fiscal 2024 peak, growing at roughly the data-center rate off a base one third the size.
Latest: $1.33B (2032Q1E)
| Period | Value |
|---|---|
| 2025Q1 | $344M |
| 2025Q2 | $392M |
| 2025Q3 | $415M |
| 2025Q4 | $452M |
| 2026Q1 | $455M |
| 2026Q2 | $516M |
| 2026Q3 | $557M |
| 2026Q4 | $567M |
| 2027Q1 | $585M |
| 2027Q2E | $613M |
| 2027Q3E | $642M |
| 2027Q4E | $671M |
| 2028Q1E | $701M |
| 2028Q2E | $732M |
| 2028Q3E | $764M |
| 2028Q4E | $796M |
| 2029Q1E | $830M |
| 2029Q2E | $864M |
| 2029Q3E | $900M |
| 2029Q4E | $936M |
| 2030Q1E | $974M |
| 2030Q2E | $1.01B |
| 2030Q3E | $1.05B |
| 2030Q4E | $1.10B |
| 2031Q1E | $1.14B |
| 2031Q2E | $1.18B |
| 2031Q3E | $1.23B |
| 2031Q4E | $1.28B |
| 2032Q1E | $1.33B |
Assumptions & reasoning
- Modelled from its own sequential history, deliberately NOT as the residual of the consolidated and data-center guidance. Backing it out of two figures management rounded to the nearest five percent would show this line flat to slightly down through fiscal 2028, which is a rounding artefact and not a forecast anyone made. The last four sequentials are +13.4%, +8.0%, +1.9% and +3.1%; 2.5% fading to 0.5% is a maturing recovery, and no management growth rate exists for this line at all.
- The two earliest quarters are our arithmetic, not Marvell's disclosure. Communications and other did not exist as a reported line before fiscal 2026; the 10-K says the four legacy end markets were combined into it with the composition of data center unchanged, so 2025 Q1 and 2025 Q2 are the sum of the four legacy figures and are flagged estimated.
- The series has a structural break inside it. The 14 August 2025 sale of the automotive ethernet business took automotive and industrial from $76.0M in 2026 Q2 to $35.0M in 2026 Q3, roughly $40M a quarter out of this line, so year-on-year comparisons are flattered on the way out and depressed on the way in through fiscal 2027 Q2.
- Seasonality tested and NOT carried. Ratio to a centred four-quarter moving average gives 0.953 for fiscal Q1, 1.013 for Q2, 1.001 and 1.031 for Q3 and 1.018 for Q4. The only quarter with two windows, fiscal Q3, spreads 3.0 points - as wide as the whole apparent shape - and one of those two windows is the quarter the automotive divestiture landed in. On a line that is part derived, part restated and carries a disposal, none of this is separable from trend. Derived, method as stated; no factors are applied.
- Terminal growth of 0.5% a quarter, about 2% a year, treats this as mature networking and consumer silicon growing roughly with GDP. It never regains the $3,291.0M of fiscal 2024 inside this horizon, which is the honest read of a portfolio that has been divested from and repositioned rather than re-grown.
- Margin glides DOWN from the consolidated 35.0% to 33.0%, and capex intensity down to 4.0%. Both are assumptions with no disclosure behind them: Marvell reports one segment, so the only defensible starting point is the consolidated rate applied to both lines, with this one set slightly below the company average because it is the cyclical, lower-content half.
Where each case comes from
Google warrant case — primary sources
The primary sources this case is built from — filings, calls and posts. Where they are claims by an interested party rather than disclosures, every number in the Google warrant column is what happens if they are taken at face value.
The warrant itself
- Aug 19, 2026 the Company issued to Google a warrant to purchase up to an aggregate of 58,970,907 shares of common stock of the Company at an exercise price of $206.58 per share
- Aug 19, 2026 The remaining Warrant Shares vest based on discretionary purchases from the Company's third quarter of fiscal 2027 through the end of the Company's fiscal year 2033 by or on behalf of Google and its affiliates in 240 equal tranches, with one tranche vesting for each $500 million in Custom Products revenue.
What the company has actually guided, for scale
- May 27, 2026 We remain confident in achieving our target model for our custom business to deliver on over $10 billion in revenue in fiscal 2029.
- May 27, 2026 Net revenue for the first quarter of fiscal 2027 was $2.418 billion, $18.0 million above the mid-point of the Company's guidance provided on March 5, 2026.
From cash flow to fair value
| Present value of free cash flow, 20 quarters | $24.94B |
| Terminal-year revenue | $51.45B |
| Terminal-year EBITDA | $15.26B |
| Exit multiple, on revenue | 9.5x |
| Terminal value | $488.75B |
| Discounted at 9.5% a year, terminal value becomes | $310.47B |
| Enterprise value | $335.40B |
| Net cash | -$1.12B |
| Equity value | $334.29B |
| Shares | 0.92B |
| Fair value per share | $365.34 |
| Against the current price of $241.45 | +51% |
The exit multiple is the whole argument, and it is an assumption with no verified comparable set behind it. 6.2 times terminal revenue is 22.4 times this model's own terminal EBITDA, because the last four projected quarters carry $30.8B of revenue at a 27.7% EBITDA margin after the stock-compensation charge. That is deliberately generous for a business the model has slowed to 12.8% year-on-year growth by the final quarter. At a 10% discount rate the base case is $146.22 a share, measured against the $245.11 close of 26 August 2026 this model carries rather than against the site's live capture, and the sensitivity is one-sided: 4x terminal revenue is $100, 6.2x is $146, 8x is $184, 10x is $226 and it takes 10.9x - about 39 times terminal EBITDA, for a company by then growing at 13% - to reach today's price. Moving the discount rate does almost nothing by comparison: 8% gives $160 and 13% gives $129. So the honest statement of this page is that the operating path is close to management's own and the gap to the price lives entirely in what a $30B-a-year Marvell should be worth. What this valuation does NOT charge: acquisition amortisation of about $225M a quarter, cash interest of about $53M a quarter on the $5.0B of senior notes (free cash flow here is pre-interest, with the $1.12B of net debt netted off the equity instead), the Celestial AI contingent consideration payable in cash and stock through fiscal 2029, and any growth in the share count from the NVIDIA preferred, the Google warrant or ordinary equity compensation.
Read the other way round: at $241.45 the market is paying 6.0x terminal-year revenue, holding every other assumption on this page fixed. That is the number to argue about.
The projected path
| Quarter | Data center | Communications and other | Revenue | YoY | EBITDA | Capex | FCF | R40 | PV of FCF |
|---|---|---|---|---|---|---|---|---|---|
| 2027 Q2E | $2.16B | $613M | $2.78B | +38% | $791M | $173M | $550M | +58 | $538M |
| 2027 Q3E | $2.52B | $642M | $3.16B | +52% | $904M | $191M | $634M | +72 | $606M |
| 2027 Q4E | $2.90B | $671M | $3.58B | +61% | $1.03B | $212M | $724M | +81 | $677M |
| 2028 Q1E | $3.31B | $701M | $4.02B | +66% | $1.16B | $233M | $821M | +87 | $750M |
| 2028 Q2E | $3.75B | $732M | $4.48B | +61% | $1.29B | $255M | $924M | +82 | $825M |
| 2028 Q3E | $4.20B | $764M | $4.97B | +57% | $1.44B | $278M | $1.03B | +78 | $901M |
| 2028 Q4E | $4.68B | $796M | $5.48B | +53% | $1.59B | $302M | $1.15B | +74 | $978M |
| 2029 Q1E | $5.18B | $830M | $6.01B | +50% | $1.75B | $327M | $1.27B | +71 | $1.06B |
| 2029 Q2E | $5.70B | $864M | $6.56B | +46% | $1.91B | $353M | $1.39B | +68 | $1.13B |
| 2029 Q3E | $6.24B | $900M | $7.14B | +44% | $2.09B | $380M | $1.52B | +65 | $1.21B |
| 2029 Q4E | $6.79B | $936M | $7.73B | +41% | $2.27B | $407M | $1.65B | +62 | $1.29B |
| 2030 Q1E | $7.37B | $974M | $8.34B | +39% | $2.45B | $436M | $1.79B | +60 | $1.36B |
| 2030 Q2E | $7.96B | $1.01B | $8.97B | +37% | $2.64B | $466M | $1.94B | +58 | $1.44B |
| 2030 Q3E | $8.57B | $1.05B | $9.63B | +35% | $2.84B | $497M | $2.08B | +57 | $1.52B |
| 2030 Q4E | $9.21B | $1.10B | $10.30B | +33% | $3.04B | $529M | $2.24B | +55 | $1.59B |
| 2031 Q1E | $9.86B | $1.14B | $11.00B | +32% | $3.25B | $562M | $2.39B | +54 | $1.67B |
| 2031 Q2E | $10.54B | $1.18B | $11.72B | +31% | $3.47B | $596M | $2.56B | +52 | $1.74B |
| 2031 Q3E | $11.23B | $1.23B | $12.46B | +29% | $3.70B | $631M | $2.73B | +51 | $1.81B |
| 2031 Q4E | $11.96B | $1.28B | $13.23B | +28% | $3.93B | $668M | $2.90B | +50 | $1.89B |
| 2032 Q1E | $12.70B | $1.33B | $14.03B | +28% | $4.17B | $706M | $3.08B | +50 | $1.96B |
Every row is projected. A year-over-year change is shown only where the quarter it compares against exists — an em dash means there is no comparable quarter, not a flat year.
Model revisions
Assumptions are marked to reality as each quarter prints. Every change is appended here, with the fair value the model produced at the time, so the model's own history stays visible.
| Date | Changed | Fair value then | Note |
|---|---|---|---|
| 2026-08-27 | verticals, corporate, valuation, scenarios | $146.22 | First cut, built on the fiscal 2027 Q1 release of 27 May 2026 and the research brief of 26 August 2026. Two verticals because Marvell discloses exactly two end markets and one reportable segment: data center on a decaying growth rate calibrated to the guided 50% and 55% fiscal 2027 and 2028 data-center growth, and communications and other grown from its own sequentials rather than backed out as a residual. Seasonality was tested on both lines by ratio to a centred four-quarter moving average and carried on neither. |