Marvell

Marvell designs custom AI silicon for hyperscalers and supplies data-center networking, storage, and interconnect chips. Custom silicon revenue growing triple-digits; partnerships with AWS and others. Strong position in optical interconnects for AI.

Marvell is a major supplier of custom AI ASICs and electro-optics for AI clusters, competing with Broadcom.

marvell.com Networking & Custom SiliconUSApublic · MRVL

At a glance

Total raised

$90M

last round ipo

Headcount

—

no public estimate on record

Headquarters

Santa Clara

USA

Founded

1995

31 years active

Status

Public

Ticker · MRVL

Subsector

Networking & Custom Silicon

For subscribers

Subscribe to see the full profile

Including Funding rounds, Bull / Bear thesis, Stock + earnings, Roster changes, Patents, News, and Open roles.

Founding

50% off

/month

91 of 100 spots left

Full

$10

/month

Available once all 100 Founding Member spots are claimed.

Get full access

Already subscribed? Sign in →

Funding history

Total raised

$90M

Rounds tracked

Latest stage

IPO

Date	Stage	Amount	Valuation	Lead investors
Jun 27, 2000	IPO	$90M	—	—

Stock performance

Latest earnings

Q1 2027Filed May 27, 2026

Revenue$2.4B

YoY+28%

GAAP EPS$0.04

Marvell reported Q1 FY2027 net revenue of $2.418 billion, a record and up 28% year-over-year from $1.895 billion. GAAP gross margin was 52.1% and GAAP diluted EPS was $0.04, compared to $0.20 in the prior-year quarter. Cash flow from operations reached a record $638.8 million. Data center revenue was $1.833 billion (76% of total), up 27% YoY, while communications and other revenue was $585 million, up 29% YoY. The company closed acquisitions of Celestial AI and XConn Technologies during the quarter. Management noted exceptional AI-related bookings and said it is significantly raising its revenue outlook for both fiscal 2027 and fiscal 2028 versus prior guidance. For Q2 FY2027, Marvell guided revenue of $2.7 billion ± 5%, representing 35% year-over-year growth at the midpoint, with GAAP diluted EPS expected at $0.37 ± $0.05.

Guidance

Q2 FY2027 revenue $2.7B ±5%; GAAP EPS $0.37 ±$0.05

Read the 8-K

How they make money

Hardware + recurring servicesPattern

Gross margin52.1%

Operating margin14%

Net margin1.4%

Q1 2027 · 8-K →

Marvell runs a fabless semiconductor model where it generates revenue from designing and selling data-center networking, storage, interconnect, and custom AI accelerator chips to hyperscalers and enterprise customers. The company collects revenue through two primary channels: one-time chip sales (standard products sold at a per-unit ASP) and custom ASIC engagements that combine upfront non-recurring engineering (NRE) fees with per-chip revenue streams tied to volume shipments. Customers like AWS pay for custom AI silicon as part of long-term design-win agreements, with revenue recognized upon delivery of fabricated chips or completion of design milestones. The economics scale with wafer volume and process-node migration — Marvell does not own fabs, so gross margin is a function of chip complexity, yield, and the spread between foundry cost and selling price. Design services and software toolkits attached to its silicon platforms add recurring revenue and deepen customer lock-in. Custom ASIC programs carry multi-year development cycles and high upfront engineering expense, subsidized by the NRE payments and back-end volume economics. The networking and electro-optics product lines benefit from the broader AI infrastructure buildout, with each new data-center generation driving both standard-part and custom-silicon refresh cycles.

Generated by deepseek-v4-flash · June 2, 2026Analytical summary · not investment advice

Founder backgrounds

Sehat Sutardja

Co-founder (former Chairman & CEO; deceased 2024)

Earned a B.S. in electrical engineering from Iowa State University and M.S. and Ph.D. degrees in electrical engineering and computer sciences from UC Berkeley in 1985 and 1988. Began his career as an analog circuit designer at Micro Linear and Integrated Information Technology, building chips for hard disk drives and digital video. Co-founded Marvell in 1995 with his wife Weili Dai and brother Pantas Sutardja. Served as President and CEO, growing Marvell into a multi-billion-dollar chipmaker covering storage, networking, and communications silicon. Holds more than 350 U.S. patents and was recognized as Silicon Valley Intellectual Property Law Association Inventor of the Year and IEEE Fellow. Passed away in 2024.

Weili Dai

Co-founder (former President, departed)

Holds a B.A. in computer science from UC Berkeley. Born in Shanghai, she moved to the U.S. at age 17, ultimately earning her degree at Berkeley where she later became a commencement speaker for the College of Engineering in 2012. Co-founded Marvell in 1995 with her husband Sehat Sutardja and his brother Pantas Sutardja, serving as President, COO, and EVP and GM of the Communications Business Group. She is notable as the only female co-founder of a major semiconductor company. The Sutardja Dai Hall at UC Berkeley is named in honor of the three co-founders.

Pantas Sutardja

Co-founder (former CTO, departed)

An electrical engineering Ph.D. and brother of Sehat Sutardja. Co-founded Marvell Technology in 1995 alongside Sehat and Weili Dai, serving as Chief Technology Officer. His deep expertise in mixed-signal design, storage controller chips, and high-speed networking silicon was foundational to Marvell's early technical roadmap and differentiation in the hard disk drive controller and Ethernet switching markets. Departed Marvell alongside his brother and co-founder Sehat Sutardja in 2016 following a board-level governance dispute.

AI bull / bear

Bull case

Marvell has secured the most important partnership in AI infrastructure: a $2 billion deal with Nvidia to integrate NVLink technology into its interconnect portfolio. This cements Marvell as the second supplier of choice behind Broadcom for the highest-value AI networking workloads. Its custom compute engine business is already growing faster than internal projections anticipated, with triple-digit revenue growth driven by hyperscaler customers like AWS who are eager for alternatives to merchant silicon. The company's optical interconnect strength positions it to benefit from the shift to higher-bandwidth cluster architectures, a key theme at Computex 2026. With Nvidia's explicit stamp of approval and hyperscalers diversifying custom ASIC supply, Marvell is capturing both the networking and compute vectors of the AI boom simultaneously.

Bear case

Marvell's custom silicon business is fundamentally a second-source play in a market where Broadcom holds commanding design-win share and deeper engineering relationships with the largest hyperscalers. The $2 billion Nvidia deal sounds large but is spread across multiple years and likely carries tight margins given Nvidia's negotiating leverage. Marvell's custom ASIC growth, while exceeding internal expectations, starts from a much smaller base than Broadcom's, making percentage gains misleading. On the networking side, optical interconnect demand is real but commoditizing rapidly as volume scales, and Marvell faces intense competition from both established players and startups. Without the scale advantages of a diversified semiconductor giant, Marvell must out-execute larger rivals across two demanding fronts simultaneously — a strategy that historically leaves little room for error when the AI capex cycle inevitably tightens.

Generated by deepseek-v4-flash · June 2, 2026Refreshes quarterly · see disclosures

In the news

ServeTheHome6/2/2026
Marvell Computex 2026 Keynote Live Coverage
Marvell CEO Matt Murphy addresses Computex 2026 keynote on connectivity solutions for AI data centers.

Watch

Leadership

Sehat Sutardja

Co-founder (former Chairman & CEO; deceased 2024)

—

Weili Dai

Co-founder (former President, departed)

—

Pantas Sutardja

Co-founder (former CTO, departed)

—

Matt Murphy

Chairman, President & CEO

—

Willem Meintjes

Chief Financial Officer

—

Products

Custom AI Silicon (Custom Compute / XPUs)
Marvell co-designs custom AI accelerators (XPUs) and compute ASICs for hyperscalers, taking a customer's architecture to silicon on leading-edge TSMC nodes with Marvell's IP for SerDes, packaging, and chiplets. This custom-compute business, anchored by multi-year programs with Amazon, Microsoft, and Google, is Marvell's fastest-growing segment and positions it as the primary alternative to Broadcom for bespoke data-center AI chips.
Electro-Optics & PAM4 DSPs
Marvell's optical-interconnect portfolio, including industry-leading PAM4 DSPs and coherent DSPs that drive the optical modules linking GPUs and switches across AI clusters. As AI training scales out, bandwidth between accelerators becomes the bottleneck, and Marvell's electro-optics, spanning 800G and 1.6T, are central to the high-speed connectivity fabric of modern AI data centers.
Teralynx Ethernet Switches
Marvell's high-radix data-center Ethernet switch silicon, such as the Teralynx 10 (51.2 Tbps), built for low-latency AI and cloud networking. As hyperscalers push Ethernet as an open alternative to proprietary interconnects for AI fabrics, Teralynx competes with Broadcom Tomahawk and Nvidia Spectrum to move traffic between thousands of accelerators at the top of the rack and across the spine.