Phase Space AI

Valuation

Marvell Technology [MRVL]

Marvell Technology, Inc. [MRVL] — Valuation

Spot $174.36 (2026-07-28) · two outputs, both required · references/valuation.md

---

Verified valuation inputs

On the share count. The scan's 874.8m is contradicted by Marvell's own guidance in the same earnings release the revenue figure came from. Common outstanding was 847.3m at 2026-01-31; the Q1 FY27 diluted weighted average was 893.3m; Q2 FY27 is guided to 915m diluted because "The Company calculated EPS under the two-class method as a result of the issuance of the Series A Convertible Preferred Stock on March 31, 2026." The NVIDIA preferred "represents a second class of common stock for purposes of computing net income per share" with rights "identical to common shares, except for certain voting rights" — it is common equity for valuation purposes.

Internal-consistency note. The $3,843.6m cash balance includes the $2.0bn NVIDIA proceeds. Using that cash while using a pre-conversion share count would credit the proceeds twice. This memo pairs 915m diluted shares with −$1,117.7m of net debt, which is the consistent combination. The alternative consistent pair — 847.3m shares and −$3,117.7m of net debt — gives an EV of $150,846m; the 915m/−$1,117.7m pair is used because it matches how the company will report EPS.

---

Output 1 — the 12-month target

Own EV/Sales history, 2022-03-10 → 2026-07-28 (1,099 trading days)

Today's multiple sits at the 90.5th percentile of Marvell's own four-and-a-half-year history, and 1.76x the median.

Construction as for the cluster: Alpaca split-adjusted daily bars against point-in-time TTM revenue lagged to filing dates; current shares and net debt held constant.

EV/EBIT is declared UNIDENTIFIED as an anchor. Marvell's continuous four-quarter EBIT series spans 291 trading days, and — more damagingly — it straddles two years of operating losses, so the distribution contains negative multiples (the 25th percentile is −86.5x). A percentile drawn from a distribution that crosses zero is meaningless. Per valuation.md the correct response is to declare it unidentified rather than substitute a peer median. EV/Sales is the anchor.

NTM revenue

Uniform cluster method: next quarter at company guidance, then +5% sequential for three quarters.

Management states: "We expect revenue growth to continue accelerating each quarter throughout fiscal 2027." A flat +5% sequential build is therefore neither aggressive nor conservative relative to guidance; it is the same rule applied to all three cluster names, which is what makes the targets comparable.

Target

12-month target: $156, −10.5% to spot. Base case is the own-history 75th percentile — the same rule applied to AVGO (+58.5%) and AMD (−28.5%).

Bull: $233 (+33.7%) — requires the multiple to hold at the 90.5th percentile for another twelve months. Bear: $132 (−24.3%) at the historical median.

The arithmetic to hold in mind: revenue grows 33.5% and the multiple compresses 33% (18.43x → 12.36x), netting −10.5%. Marvell's target is closer to spot than AMD's because its multiple is less stretched (90.5th versus 95.9th percentile), not because its business is better positioned.

External sanity band. No professional target for MRVL is on file in this project, so no gap is reported.

---

Output 2 — the implied-path test (the Valuation Criteria)

Parameters

On the terminal margin. The screen capped this at 15.6% from Marvell's own 16.1% — a trivial haircut here compared with what the same cap did to AVGO, but wrong in the same way. 17.0% is used because the company's own Q2 FY27 guidance implies it: 2,700 × 52.6% gross margin − $960m GAAP opex = $460m, or 17.0%. Using a figure above the demonstrated level means the FAIL below is not an artefact of a harsh margin.

Result

Required revenue CAGR: 54.2% for five years.

Implied FY2031 revenue: $76.1bn (from $8.7bn — an 8.7x increase).

Implied multiple compression: today's 115.4x EV/EBIT to a 20.0x exit = −82.7%.

Sensitivity — over the exit multiple

Sensitivity over the terminal margin, at the 20.0x anchor: 14% → 60.3% required; 17% → 54.2%; 20% → 49.3%.

No cell produces a PASS. At twice the anchor — a 40x EBIT exit, or 6.8x EV/Sales, comfortably inside Marvell's historical 5.62x–33.35x range — the required 34.3% still exceeds the latest quarter by 6.7pp and the trailing record by 21.7pp. Even against Marvell's single best quarter in three years (+63%, Q1 FY2026) the anchor case is only 8.8pp of headroom on a rate the company has since fallen 42pp below.

Terminal value is 100% of modelled EV, above the 60% threshold, so the reverse DCF is the primary long-horizon output and this table is the analysis.

Verdict: FAIL

criteria.md permits PASS WITH ARGUMENT where the excess over the demonstrated record is bridged by "a specific, evidenced reason (a named product cycle, mix shift, pricing action). Narrative does not qualify."

Marvell has a genuinely specific, genuinely new mechanism: the NVIDIA strategic partnership, funded with a $2.0bn preferred investment on 2026-03-31, "to connect our custom XPUs and compatible scale-up networking with NVIDIA's AI infrastructure ecosystem." That is not narrative — it is a signed, capitalised relationship with the dominant player in the end market. It is the single best piece of evidence in Marvell's filings.

It does not close the gap, for four reasons.

1. Magnitude. The price requires 54.2% for five years against a 12.6% trailing record and a 27.6% latest quarter. $76.1bn of FY2031 revenue is 8.7x the current business. That is not a "more than demonstrated" gap; it is a different business.
2. The trailing record moves the wrong way. Marvell's data-centre YoY growth went +98% → +78% → +76% → +69% → +38% → +21% across five quarters — a 77pp deceleration in the exact end market that has to compound at 54% for five years. Over the same period Broadcom's AI line went from +46% to +143%. Two companies, the same end market, overlapping customers, opposite directions. A required path far above the demonstrated one is hardest to argue when the demonstrated one is falling.
3. Nothing is contracted. No committed RPO is disclosed. DeferredRevenueCurrent is $63.7m on $8.7bn of TTM revenue. Meanwhile Marvell pre-commits foundry capacity "up to 52 weeks to secure capacity" — obligations on the supply side against cancellable orders on the demand side. Broadcom's $164.6bn of firmly committed, non-cancellable RPO is what a bridgeable gap looks like.
4. The concentration cuts the other way. Distributor A is 45% of revenue (from 24% two years ago) and three customers hold 75% of gross receivables. A required 54% CAGR delivered through two counterparties is not a lower-variance path than the same CAGR delivered through fifty.

Ranking input for the strategy: margin = −41.6pp against the corrected trailing CAGR, −26.6pp against the latest quarter. Third of three in the cluster on the trailing measure and third on the latest-quarter measure.

---

Cluster note — was one shared exit anchor justified?

Yes on growth level, and this is the homogeneous cluster where such an anchor is legitimate. But MRVL is the member that required separate treatment, and it got it.

Growth level. TTM year-over-year revenue growth: AVGO +32.3%, AMD +35.0%, MRVL +34.1% — a 2.7pp band. All three are fabless (SIC 3674), asset-light, high gross margin, design-win-driven, customer-concentrated, selling into the same hyperscaler capex budgets. Marvell's own 10-K names AMD and Broadcom as direct competitors. On level, the three genuinely bracket each other and a shared 20–24x EBIT growth-matched anchor is defensible.

Where MRVL does not bracket the other two: trajectory.

AVGO and AMD accelerated monotonically. MRVL decelerated by 35pp and is only now recovering. Its data-centre engine decelerated by 77pp. That is a different regime, not a different point on the same curve.

MRVL is therefore anchored separately at 20.0x EBIT — the low end of the cluster's growth-matched range — rather than at the cluster midpoint. Under the framework's own rule ("an exit multiple may only be drawn from a comparator set whose growth brackets the subject's growth at the exit year"), a name whose growth is falling toward the comparator set from above is not symmetrically placed with one rising toward it from below.

Where the shared anchor was genuinely illegitimate, and the screen did it anyway: the terminal margin. scan_v3 capped AVGO at 15.6% (from a demonstrated 39.9%) and MRVL at 15.6% (from 16.1%) on an "industry 75th percentile" basis, while leaving AMD at its own 10.7%. Actual TTM operating margins are AVGO 43.4%, MRVL 16.0%, AMD 11.7% — a spread of nearly 4x. Margin is the one axis on which this cluster is not homogeneous, and it is the axis the screen flattened. That single cap is what produced "reverse DCF unsolvable in range" and INDETERMINATE on all three names. For MRVL the cap was nearly harmless (15.6% vs 16.0%); for AVGO it was decisive.

---

Sources

Marvell 10-Q (2026-05-02, filed 2026-05-28); 10-K (2026-01-31, filed 2026-03-11); earnings releases 8-K EX-99.1 2026-05-27, 2026-03-05, 2025-12-02, 2025-08-28, 2025-05-29, 2025-03-05, 2024-12-03, 2024-08-29, 2024-05-30; SEC XBRL companyfacts CIK 0001835632; Alpaca daily bars and option snapshots 2026-07-28/29; reports/scan_v3/_scan.json.