Marvell Technology, Inc. reported $8.195B in fiscal 2026 net revenue, employed 7,480 employees, and is led by Matthew J. Murphy. Marvell's latest fiscal year shows the AI data-center transition becoming the center of the company. Data center generated nearly three quarters of fiscal 2026 revenue, while a large gain from the sale of the automotive Ethernet business also shaped GAAP net income.
Marvell: Key Facts
- Revenue: $8.195B (fiscal 2026)
- CEO: Matthew J. Murphy
- Employees: 7,480
- Headquarters: Santa Clara, California
How Does Marvell Make Money?
Marvell makes money by designing and selling complex semiconductors for data centers, cloud infrastructure, communications networks, storage, and custom silicon programs. It is fabless, so manufacturing is outsourced to foundry and packaging partners, while Marvell focuses on architecture, IP, customer design wins, software, and long-cycle infrastructure platforms.
Marvell Financials
Marvell reported fiscal 2026 net revenue of $8.1946 billion, up 42% from $5.7673 billion in fiscal 2025. GAAP net income was $2.670 billion, helped by the sale of the automotive Ethernet business to Infineon for $2.5 billion and a related pre-tax gain of about $1.8 billion. Fiscal 2026 revenue was $6.1003 billion from Data Center, or 74% of total revenue, and $2.0943 billion from Communications and Other, or 26%. Direct customers accounted for $4.6304 billion and distributors for $3.5642 billion.
Marvell Competitive Advantage
Marvell competes with Broadcom, Nvidia, AMD, Intel, MediaTek, and specialized networking or optical semiconductor suppliers. Its strongest position is in data infrastructure, where custom silicon, electro-optics, SerDes, storage, and networking IP can be bundled into platforms for hyperscale and enterprise customers.
Marvell Outlook
Marvell's outlook depends on AI data-center capex, custom silicon ramps, optical connectivity demand, communications infrastructure spending, customer concentration, and foundry execution. The upside is continued data-center mix expansion; the risk is that hyperscale design wins and optical ramps are lumpy and technically demanding.
