MRVL

Marvell Technology

Q1 FY2027 earnings · 2026-05-27$0.82 consensus

Summary

What they do:

Design custom AI accelerator chips (Amazon Trainium, Microsoft Maia) and PAM4 optical DSPs that sit inside every high-speed data center transceiver — the only semiconductor company spanning both custom AI silicon (L06) and optical connectivity silicon (L11).

Why they matter:

Marvell is the #2 custom ASIC designer behind Broadcom, with a five-year multi-generational AWS supply agreement and Microsoft Maia, plus dominance in PAM4 optical DSPs powering 400G/800G/1.6T data center transceivers — creating unique dual-layer exposure to both AI compute and AI networking.

Recent performance:

FY2026 revenue $8.2B (+42% YoY), record Q4 revenue $2.22B. Custom AI ASIC revenue went from near-zero to $1.5B in one year. Stock at ~$140 (all-time high), market cap ~$120B. Q1 FY2027 guided $2.4B.

Our Verdict

Play TypeEmerging
Rel. ValueAttractive

The #2 custom ASIC designer with the best dual-layer positioning in AI infrastructure — custom chips for AWS/Microsoft plus optical DSPs for every data center transceiver — at ~55x forward earnings with 42% revenue growth.

Structural trends

Custom ASIC proliferation (AWS Trainium multi-generational commitment)optical bandwidth scaling (400G→800G→1.6T driving DSP content)data center network fabric expansioninference compute scaling

Structural

75

/ 100

Moat

7/10

#2 custom ASIC + #1 optical DSP — dual-layer AI moat with 5-year AWS deal

AI Exp.

High

~74% AI

Play Type

Emerging

AI Growth

~42%

Rel. Value

61

ATTRACTIVE

PriceLIVE

$133.83

+1.93%

Live via Yahoo Finance · refreshes every 5 min

Market Cap

$117.0B

P/E Ratio

43.7

P/S Ratio

14.3x

52W High

$136.90

52W Low

$48.09

52W Chg

178.3%

Beta

1.82

The Catch

Marvell's dual-layer positioning is a strength and a vulnerability. On custom ASICs, Marvell is unambiguously #2 behind Broadcom — fewer confirmed customers, smaller engineering force, and later to market in most hyperscaler programs. The five-year AWS deal is transformative but concentrates risk: if AWS Trainium underperforms vs. NVIDIA GPUs in inference benchmarks, Amazon could reduce deployment targets without breaking the contract. On optical DSPs, Broadcom's acquisition of the Salina DSP line creates a well-funded competitor with switch ASIC bundling capability that Marvell cannot match. The non-AI business segments (enterprise, carrier, consumer) are declining and dilute Marvell's AI growth narrative — at 74% data center mix, the remaining 26% is a drag. At all-time-high stock price, any deceleration in the ASIC ramp or optical DSP share loss would compress the multiple sharply.

If They Win

If Marvell executes on both pillars — if Trainium becomes the dominant inference chip at AWS, if Maia scales at Microsoft, if a third ASIC customer signs a multi-year agreement, and if 1.6T optical DSPs capture 50%+ of the next-gen transceiver market — Marvell becomes the only semiconductor company with dominant positions in both AI compute and AI connectivity. Revenue compounds to $15-18B by FY2028 with 60%+ gross margins. The custom ASIC business reaches $5-6B annually, rivaling Broadcom's scale. The optical DSP business generates $6-7B as every fiber in every AI data center carries Marvell silicon. Market cap reaches $200-250B as the market recognizes that Marvell's dual-layer moat is the most structurally advantaged position in AI semiconductors after NVIDIA and TSMC.

Not financial advice. All scores generated via AI algorithms using public data.