Goliath Investment Group
Goliath Investment Group

© 2026 Goliath Investment Group

Investment Research

Equity Research
All Research

Marvell Technology: Riding the Custom Silicon Wave

Jonathan Cao··6 min read

Executive Summary#

Recommendation: BUY | Price Target: $130 | 47.96% Upside

Marvell Technology has rapidly repositioned itself at the center of the AI infrastructure build-out. With custom silicon partnerships - including Microsoft's Maia program - a growing optical DSP business now tracking 50% YoY growth, and the Celestial AI acquisition unlocking a new photonic interconnect TAM, Marvell has the building blocks to double its data-center market share from ~10% in 2023 to ~20% by 2028. The AI infrastructure horizon is enduring, and Marvell is one of the best-positioned mid-cap plays to capture it.


Company Overview#

Marvell is a leading supplier of data infrastructure semiconductor solutions, developing high-performance System-on-a-Chip (SoC) architectures that integrate analog, mixed-signal, and digital signal processing functionality. Its product portfolio includes:

  • Ethernet solutions
  • Custom application-specific integrated circuits (ASICs)
  • Interconnect products
  • Storage controllers

These products are essential for data centers, enterprise networking, and communications infrastructure, serving a diverse range of end markets.

Marvell employs a fabless business model, outsourcing manufacturing to third-party foundries while focusing on chip design and go-to-market. Operations span the United States, China, India, Israel, and other countries.


Market Outlook#

  • The custom AI chips market is expected to surpass $94B by 2028, up from just $6.6B in 2023 - a roughly 15x expansion driven by hyperscaler demand for proprietary silicon.
    • Across the broader data-center semiconductor market, Marvell currently holds ~10% share and is targeting ~20% by 2028 as AI infrastructure spending accelerates.
  • Celestial AI's photonic interconnect technology targets extreme-scale AI compute clusters, unlocking a new TAM for Marvell.
    • AI scale-up interconnect TAM is expected to reach $30–50B by 2030.

Key Catalysts#

  • Microsoft XPU / Maia Program: Microsoft is building its own in-house AI accelerator chips - called Maia - for Azure data centers, reducing reliance on NVIDIA GPUs. Marvell has been one of the key silicon design partners on this program, representing a significant and recurring revenue opportunity.
  • Celestial AI Revenue Ramp: Marvell expects initial revenue contributions from Celestial AI's photonic interconnect technology to begin in the second half of fiscal 2028, opening an entirely new high-growth revenue stream tied to AI scale-up infrastructure.

Investment Thesis#

Strategic acquisitions expanding AI capabilities

  • Marvell recently acquired Celestial AI and XConn, adding photonic interconnect and PCIe switch capabilities that meaningfully strengthen its AI and cloud data center portfolio.

Growth and momentum far outpacing the sector

  • Forward revenue growth of 37% vs. the semiconductor sector's 10%
  • One-year price performance of 42% vs. the sector's 10%, reflecting sustained investor confidence in the AI infrastructure thesis

Structural competitive advantages with high switching costs

  • Marvell holds deep intangible assets in networking chip design that enable it to compete at the cutting edge
  • Hyperscalers co-designing custom silicon with Marvell are embedded in multi-year development programs - switching costs are high
  • Optical DSP demand is accelerating sharply: now expected to grow 50% YoY vs. prior expectations of 30%

Financial Analysis#

Figures in USD millions.

JAN '26JAN '25JAN '24JAN '23JAN '22
Income Statement
Sales8,1955,7675,5085,9204,462
Gross Income4,1812,3822,2942,9882,063
EBIT1,338(366)(437)360(253)
EBITDA2,6299919611,752992
Net Income2,670(885)(933)(164)(421)
Balance Sheet
Cash & Short-Term Investments2,639948951911614
Total Assets22,28520,20521,22922,52222,109
Total Debt5,0354,6794,7045,1245,115
Net Debt2,3963,7303,7534,2134,501
Total Liabilities7,9776,7786,3976,8856,407
Total Shareholders' Equity14,30813,42714,83115,63715,702
Cash Flow
Capital Expenditures(358.6)(291.6)(350.2)(217.3)(187.1)
Free Cash Flow1,396.41,396.61,034.21,082.6650.0
Metrics
Gross Margin51.02%41.31%41.64%50.47%46.24%
Operating Margin16.33%(6.35%)(7.93%)6.07%(5.68%)
Total Debt / EBITDA1.79x4.44x4.65x2.78x4.98x
EBITDA / Interest Expense12.98x5.23x4.54x10.27x7.12x
Price / Earnings25.71xN/AN/AN/AN/A
EV / EBITDA26.35x102.44x64.87x23.49x65.48x

Competitors#

Marvell's biggest competitors are NVIDIA, Intel, Broadcom, and AMD across the data-center semiconductor landscape. Additionally, emerging AI chip companies - Cerebras, Groq, and SambaNova - pose growing competition for Marvell's targeted share of AI inference and training silicon.


Key Risks#

  • AI Capex Cyclicality: If hyperscaler AI capital expenditure flatlines or declines - due to demand disappointment, macro pressure, or ROI skepticism - Marvell's growth story could decelerate sharply given its exposure to a single secular theme.
  • Customer Concentration: Revenue is heavily concentrated in a small number of customers. In Q3, one distributor accounted for 38% of revenue and one direct customer for 11%. In FY25, the top 10 customers represented 81% of total revenue - leaving Marvell meaningfully exposed to the pullback of any single major relationship.
  • ASIC Market Competition: Broadcom is an established, well-capitalized competitor in the custom ASIC space and poses a meaningful threat to Marvell's market share gains.
  • Stretched Valuation: At an EV/EBIT of 60x - significantly above the sector median of 26x - any miss on growth expectations could result in sharp multiple compression.

Conclusion#

There is a high likelihood that the AI infrastructure investment horizon is enduring and profitable for investors who stay the course. The goal is to remain vigilant amid cyclical concerns and capex spend dynamics. Based on the above research, the recommendation is BUY with a price target of $130 (47.96% upside).

Marvell is among the best-positioned mid-cap plays on the AI infrastructure build-out - with custom silicon partnerships at the hyperscaler level, optical DSP leadership, and a fabless model that scales efficiently with AI capex. It is a high-conviction addition to an AI-focused portfolio.