stockrowWhat it does
Marvell Technology is a fabless supplier of data-infrastructure semiconductors: it designs chips but uses outside foundries and subcontractors to make, package and test them. Its products move, process, store and secure data from the core of a data center to the network edge.
The company serves two end markets. Data center products go into cloud and on-premise AI systems, AI and general-purpose servers, Ethernet switching, network-attached storage, storage-area networks and links between data centers. Communications and other products serve enterprise networks, carrier networks, consumer devices, and certain industrial and government applications.
Its portfolio combines compute, networking, security, interconnect and storage technologies. This includes custom ASICs—chips designed to an individual customer’s specification—as well as optical and electrical connectivity chips, Ethernet switches and controllers, PCIe and CXL connectivity products, fibre-channel adapters, infrastructure processors, security processors, and hard-disk and solid-state storage controllers.
Many of these chips solve a similar practical problem: moving rising volumes of data reliably and with limited power use. Optical DSPs, drivers and TIAs help transmit data over optical links; switches direct traffic among servers and other equipment; and custom chips can combine customer-specific compute, memory, networking and security functions. Marvell also develops UALink and ESUN switches for scale-up AI networks.
How it makes money
Marvell earns revenue from designing and selling integrated circuits. It operates in one reportable segment: the design, development and sale of integrated circuits. Its sales mix is organized into data center and communications and other end markets rather than separate operating segments.
FY2026 revenue was $8.2 billion; Q2 FY2027 revenue was $2.739 billion.
Data center was the larger end market in FY2026, producing $6.1 billion, or 74.4% of revenue. Communications and other generated $2.1 billion, or 25.6%. The company sells both standard products, which can be sold to multiple customers, and optimized solutions customized to a particular customer’s requirements. Custom ASIC work can include customer-specific designs for AI, data center, compute, networking, carrier, storage, aerospace and defense applications.
The portfolio creates several routes to a system sale. A customer building AI infrastructure may use Marvell interconnect products for data transmission, switches to connect systems, custom chips for specialized functions, and CXL products for memory expansion. Its products also address storage and network-processing applications outside AI systems.
Revenue is recognized when control of promised products or services transfers to customers. Selling prices face continuing pressure, so the company seeks to introduce new products and lower the cost of design, manufacture, assembly and test. Its product-development cycle can be long: customers generally evaluate products before incorporating them into designs, and commercial shipments can begin six months to three years after selection.
| Product or service | Revenue, FY2026 | Share | Change on the year |
|---|---|---|---|
| Data center | $6.1 billion | 74.4% | 46.5% |
| Communications and other | $2.1 billion | 25.6% | 30.6% |
Customers and how it sells
Marvell’s target customers are original equipment manufacturers, original design manufacturers and distributors. OEMs and original design manufacturers design and make end-market devices, while distributors support sales and marketing. The company expects a significant percentage of sales to continue to come directly from key customers, while distributors operate in the United States, Europe and Asia. Field application engineers help customers design, test and qualify systems containing Marvell chips.
Customer concentration is material. Distributor A accounted for 37.0% of FY2026 revenue, and Customer A accounted for 14.0%. The company monitors the creditworthiness of both distributor and direct customers.
Its revenue is heavily linked to Asian manufacturing and customer operations. Sales shipped to customers with operations in Asia represented approximately 77% of FY2026 net revenue. By reported customer location, China generated $3.0 billion, or 36.2%, of FY2026 revenue; Taiwan generated $1.7 billion, or 20.2%; the United States generated $1.2 billion, or 14.3%; and other locations generated $2.4 billion, or 29.2%.
Marvell has operations in Argentina, China, India, Israel, Japan, Singapore, South Korea, Taiwan and Vietnam.
| Region | Revenue, FY2026 | Share | Change on the year |
|---|---|---|---|
| China | $3.0 billion | 36.2% | 18.4% |
| Other | $2.4 billion | 29.2% | 37.4% |
| Taiwan | $1.7 billion | 20.2% | 195.6% |
| United States | $1.2 billion | 14.3% | 22.7% |
Suppliers and costs
Marvell is fabless. The vast majority of its integrated circuits are made using available CMOS processes at independent foundries, rather than in company-owned fabrication plants. It typically outsources production packaging and testing to several subcontractors, primarily in Taiwan, Canada, Korea, Singapore and China.
Most products are manufactured by third-party foundries in Taiwan, with other sources in China, Germany, South Korea, Singapore and the United States. Manufacturing partners’ available capacity and existing obligations can limit their ability to supply materials and services. The company places firm supplier orders up to 26 weeks before expected customer delivery and may make supply commitments up to 52 weeks ahead. It also uses capacity-reservation arrangements with certain foundries and substrate partners.
The cost structure includes outsourced wafer fabrication, packaging, testing, inventory, research and development, and supplier-capacity prepayments. In Q2 FY2027, cash flow from operations was $606 million, while inventory was $1.36 billion. The company planned approximately $1 billion of capacity prepayments to suppliers in FY2027, to be applied against future material purchases.
- Revenue$8.2B
- Cost of revenue−$4.0B
- Research and development−$2.1B
- Selling, general and administrative−$767.1M
- Other operating costs−$15.5M
- Operating income$1.3B
Competitive position
Marvell competes in markets characterized by rapid technology change, evolving standards, frequent product launches and pricing pressure. Its directly named competitors include Advanced Micro Devices, Alchip, Astera Labs, Ayar Labs, Broadcom, Cisco Systems, Credo Technology Group Holding, Intel, Global Unichip, Lightmatter, MACOM Technology Solutions Holdings, MediaTek, Microchip Technology, Montage Technology, Nvidia, NXP Semiconductors, Phison Electronics, Qualcomm, Rambus, Ranovus, Realtek Semiconductor, Semtech, Silicon Motion Technology and Socionext.
Competition spans custom and merchant semiconductor products for AI and cloud infrastructure, interconnects, switching, networking, storage, processors and communications equipment. Customers can also develop certain semiconductor products internally. The company expects average unit selling prices to remain under significant pressure.
Marvell’s stated competitive factors include product performance, features, quality and price; execution and timing of new product introductions; standards adoption; market demand; foundry capacity; and the number and nature of competitors in each market.
Our reading: Marvell’s breadth across custom ASICs, optical and electrical interconnects, switching, processing and storage is a medium-strength moat because it can supply several functions in one data-infrastructure design, but it competes against established specialists and broad semiconductor companies. (item1; call_prepared)
Our reading: Its system-level expertise, large intellectual-property portfolio and ability to customize chips are a medium-strength moat, because customers undertake comprehensive technical evaluations and product selection can precede shipments by six months to three years. (item1; item1a)
Our reading: Fabless manufacturing is not a moat on its own because production capacity remains dependent on third parties, especially foundries in Taiwan. (item1; item1a)
| Competitor | Where they compete | Threat (our reading) |
|---|---|---|
| Broadcom | AI and cloud data-center semiconductors | High |
| Advanced Micro Devices | data-center semiconductor products | High |
| Nvidia | AI infrastructure and scale-up networking | High |
| Intel | data-center and infrastructure semiconductors | Medium |
| Qualcomm | communications and infrastructure semiconductors | Medium |
| Cisco Systems | networking and switching | Medium |
| Astera Labs | data-center connectivity | Medium |
| Credo Technology Group Holding | high-speed interconnects | Medium |
Growth opportunities
Marvell’s principal growth vector is data-center infrastructure for AI, including custom chips, optical and electrical connectivity, Ethernet switching, CXL memory expansion and scale-up networking. FY2026 data-center revenue was $6.1 billion, up 46.5% from FY2025, while communications and other revenue was $2.1 billion, up 30.6%.
The company expects FY2027 revenue of roughly $12 billion, approximately 45% year-over-year growth, driven by data-center revenue growth of approximately 60%. It expects FY2028 revenue to grow approximately 50% year over year and data-center revenue to grow more than 60% year over year. It also expects its custom business to more than double in FY2028.
Within connectivity, the company reports strong demand for 800-gig optical DSPs and a rapid ramp of 1.6T products. It is developing for scale-out, scale-across and scale-up AI networks. The company expects copper and optical technologies to coexist for several years in scale-up networks, while customers plan optical deployments starting as early as the following year. It reports adoption of NPO solutions at multiple customers in addition to CPO activity.
The company sizes its total addressable market at $400.0 billion by 2030. It completed the Celestial AI acquisition to add Photonic Fabric technology for scale-up interconnect and the XConn acquisition to add PCIe and CXL switching silicon. An expanded commercial agreement with a key hyperscaler covers existing custom programs, new design wins and potential future programs; revenue through FY2028 from covered programs is already included in the company’s custom-revenue target.
- $4.5BFY2022
- $5.9BFY2023
- $5.5BFY2024
- $5.8BFY2025
- $8.2BFY2026
- $12.0BFY2027 guidance
Structural risks
-
Customer and end-market concentration. A few customers account for a significant portion of revenue, while data center is an increasing share of sales. The company also faces the risk of losing design wins with key customers.
-
Demand and order volatility. A significant portion of sales is made on a purchase-order basis, allowing customers to cancel, change or delay commitments on relatively short notice. Customers may also manage inventories differently, buy competing products or lose market share.
-
China and trade restrictions. Export restrictions, tariffs and changes in political, regulatory and economic policies related to China can reduce demand. Customers in China may build indigenous solutions, vertically integrate or replace products with other suppliers.
-
Third-party manufacturing dependence. The company relies on manufacturing partners for fabrication, assembly, testing and packaging. Capacity limits, supply-chain disruptions, shortages, yield or quality problems and regional concentration can constrain supply or raise costs.
-
Technology and competition. The company must develop, qualify and introduce new products in time for changing data-center and communications requirements while facing rapid technological change, pricing pressure and competitors’ product development.
-
Acquisition and execution risk. The company faces risks in integrating acquisitions and realizing anticipated benefits or synergies, as well as in scaling the business and securing capacity for growth.
Bottom line
Our reading
Our reading: Marvell is a fabless data-infrastructure chip designer whose business is increasingly centered on AI data centers, with custom silicon and connectivity spanning optical links, switching, memory expansion and storage. (item1; call_prepared)
Our reading: Its opportunity depends on customers scaling AI infrastructure and on Marvell executing complex designs while securing outside manufacturing capacity; revenue is also concentrated by end market, geography and major customers. (item1; item1a; call_prepared)
Our reading: The company’s broad technology portfolio can support multiple parts of a customer’s system, but competition, customer in-sourcing, price pressure and export restrictions remain material constraints. (item1; item1a)