Skip to content
MRVL
~5 min read · 1,158 words ·updated 2026-04-28 · confidence 88%

Total Addressable Market (TAM) & Serviceable Addressable Market (SAM)

Custom AI Silicon (XPU) TAM

Market Size & Growth Trajectory

TAM Overview: Marvell’s 2024 Investor Day (✓) raised the overall data center TAM to $94 billion by calendar 2028 (up from prior $21 billion estimate), driven primarily by custom accelerated compute reaching $55.4 billion by 2028. This splits into $40.8B for custom XPU and $14.6B for XPU attach revenue.

A separate analyst estimate quantifies the custom AI accelerator TAM at $40.8 billion by 2028 with a 47% CAGR through that year (◐—secondary source). Cumulative lifetime pipeline value of 18 design wins and 50+ sockets is pegged at $75 billion by Barclays analysts, though Barclays cautioned these deals will not contribute revenue until 2028 at the earliest.

Marvell’s SAM: Management targets expanding data center share from ~13% in FY 2024 to 20% of the $94B TAM by FY 2028 (✓). This implies a SAM of roughly $18.8B by 2028 if achieved.

Revenue Trajectory (Marvell Reported)

  • FY 2026: Custom XPU contributed $1.5 billion in revenue (✓).
  • FY 2027E: Management guides minimum 20% YoY growth, implying ~$1.8B+ (✓).
  • FY 2028E: Custom expected to double from FY 2027, reaching ~$3.6B+ (✓).

Confidence: ✓ (Marvell Investor Day guidance and Q3 FY2026 earnings calls).

Optical Interconnect (DSP & Transceiver) TAM

Market Size & Roadmap

TAM Overview: The optical interconnect TAM for data center Ethernet transceivers and Co-Packaged Optics is projected to grow from $11.5 billion by 2030 (LightCounting, ✓). More granular 2025–2026 growth: optical transceiver and DSP sales for AI-driven scale-out and scale-up networks reached $16.5 billion in 2025 and are forecast to reach $26 billion in 2026 (60% growth, LightCounting ✓).

Technology Transition: 800G → 1.6T → 3.2T

  • 800G PAM4 DSP: Hyperscalers’ 2025 investments caused 800G PAM4 chipset shipments to nearly triple in 2025; expected to more than double in 2026 (LightCounting ✓). 800G module form factors (both pluggable optics and CPO) remain dominant.

  • 1.6T DSP (emerging ramp): Marvell was first-to-market with 200G/lane 1.6T DSPs in 5nm (Nova, 2023), followed by 3nm platform (Ara, 2024). Ara X, Ara T, Petra, and Aquila M DSPs are sampling starting Q1 2026 (✓). Market transition to 1.6T is underway with tens of millions of ports projected for 2026 (LightCounting, ✓).

  • 3.2T & Beyond: Marvell’s April 2026 acquisition of Polariton Technologies (Swiss plasmonics designer) enables scaling to 3.2T and higher speeds via advanced modulation (polymer electro-optic materials) (✓). Industry analysts project 3.2T volumes ramping 2027–2028.

PAM4 vs. Coherent vs. Linear Optics (LPO/LRO)

  • PAM4 DSP-based modules: Dominant through 2025–2026 for 800G and 1.6T.
  • Linear Pluggable Optics (LPO): Marvell introduced 1.6T LPO chipset (TIA + laser driver) enabling sub-8W modules with 40% power reduction vs. DSP-based equivalents (OFC 2025, ✓). LRO (Linear Receive Optics) increasingly discussed for 200G/lane (1.6T+) designs.
  • Co-Packaged Optics (CPO): NVIDIA and Broadcom driving CPO adoption; Marvell entered CPO market with $3.25B acquisition of Celestial AI (December 2025) and $2B investment from NVIDIA (April 2026) for optical semiconductor alliance (✓).

Marvell’s SAM: Marvell’s DSP and transceiver product portfolio (Ara, Ara X, Petra, Aquila lines) targets dominant position in PAM4 DSP for 800G→1.6T (2025–2027) and nascent CPO participation beginning 2027–2028. Market share in DSP/transceiver revenue not disclosed; estimated mid-to-high teens % in plug-gable optical (secondary).

Confidence: ✓ for 2025–2026 demand; ◐ for 3.2T TAM/SAM (ramp timing uncertain).

Switching Silicon (Ethernet Switch ASIC) TAM

Market Size & Competitive Landscape

TAM: Marvell’s management targets expansion of overall data center share from 13% (FY 2024) to 20% by FY 2028 against a $94B TAM. Within this, switching (Teralynx family) competes in the Ethernet switching ASIC market, historically dominated by Broadcom.

  • Broadcom Tomahawk dominance: Holds approximately 80% market share in Ethernet switch ASICs (as of 2025). Tomahawk 6, launched June 2025, is the world’s first 102.4 Tbps switch (64×1.6TbE ports).
  • Marvell Teralynx positioning: Marvell holds roughly 10% of Ethernet switching ASIC market (2025), but has gained traction by optimizing for hyperscale AI fabric requirements (high radix, predictable latency). Teralynx 9 (51.2 Tbps) competes with Tomahawk 5; Teralynx 10 (102.4 Tbps roadmap) targets Tomahawk 6 parity by 2027–2028.

SAM & Market Share

Marvell’s SAM: Teralynx is proven and deployed at scale in the largest cloud data centers (millions of ports shipped). Management does not separately disclose Teralynx revenue; estimated at 5–10% of total data center segment revenue.

2-Year / 5-Year Outlook: Switching TAM is growing at ~10–15% CAGR (lower than compute/memory/optics), driven by increasing east-west fabric bandwidth in AI clusters. Marvell’s ability to gain share hinges on hyperscaler adoption of custom designs (e.g., custom-silicon strategies at Meta, Google).

Confidence: ✓ for Broadcom/Marvell market share; ◐ for total switching TAM (not separately broken out by analysts).

Storage TAM

Market Overview

Storage segment revenue is declining as a % of Marvell’s data center portfolio, displaced by custom AI silicon and optical interconnect growth.

  • Status: Legacy HDD/SSD controller business (Marvell Cavium legacy). Not a growth driver for Marvell in 2025–2028.

Confidence: ✓.

Carrier Infrastructure TAM

5G RAN & Automotive Ethernet

TAM Sizing: Carrier segment (5G RAN processors, automotive Ethernet switches) contributes ~5–10% of Marvell’s total revenue; not core to AI infrastructure play.

  • 5G RAN: Marvell competes with Broadcom in base station processors; limited growth (mature market).
  • Automotive Ethernet: Emerging but nascent; not a major TAM driver through 2027–2028.

Confidence: ◐ (limited public disclosure).

TAM/SAM Summary Table

Segment2028 TAMMarvell SAMMarket Share2025–2028 CAGRConfidence
Custom AI XPU$40.8B~$8–10B20–25% (target)47%✓
Optical Interconnect (DSP/Transceiver)$11.5B + CPO (TBD)~$2–3BMid-to-high teens35–40%✓
Switching ASIC~$8–12B~$1–1.5B10% (stable)8–12%◐
Storage~$2–3BDeclining<2%-10% to 0%✓
Carrier Infrastructure~$3–5B<5%Single digit3–5%◐

Key Takeaways

  1. Custom AI silicon is the dominant growth vector for Marvell through 2028, with TAM expanding 47% CAGR and Marvell positioned to capture 20%+ share via design wins.
  2. Optical interconnect is the second-largest near-term opportunity, driven by AI capex; Marvell is well-positioned in PAM4 DSPs and gaining CPO relevance post-Celestial AI and NVIDIA partnership.
  3. Switching remains competitive but stable, with Marvell maintaining niche share in hyperscale-optimized designs.
  4. Overall data center TAM of $94B by 2028 is the primary driver of Marvell’s growth strategy, vs. legacy storage and carrier segments (declining).

Sources

  • Marvell Investor Day 2024 guidance
  • Marvell FY 2026 earnings call transcripts (Q3, Q4)
  • LightCounting optical interconnect forecasts (2025–2026)
  • 650 Group Ethernet ASIC market analysis
  • Polariton and Celestial AI acquisition announcements (April, December 2025)

Cross-references