Short answer: NVIDIA operates the far larger AI data-center business, reporting $193.7 billion in Data Center revenue for fiscal 2026, while Marvell reported $6,100.3 million ($6.1003 billion). The companies also occupy different positions in the stack: NVIDIA sells an integrated accelerated-computing and networking platform, whereas Marvell emphasizes custom silicon, optical and electrical interconnects, Ethernet, storage and related infrastructure semiconductors.
They are competitors in selected components but not interchangeable vendors. Their March 31, 2026 NVLink Fusion agreement makes the relationship more complementary: Marvell plans to provide custom XPUs and compatible scale-up networking for NVIDIA’s ecosystem. The announcement describes planned cooperation, not verified customer deployments, shipment volumes or partnership revenue.
As an Amazon Associate I earn from qualifying purchases.
Marvell and NVIDIA at a glance
| Measure | Marvell | NVIDIA |
|---|---|---|
| Fiscal period | Fiscal year ended January 31, 2026 | Fiscal year ended January 25, 2026 |
| Reported Data Center revenue | $6,100.3 million | $193.7 billion |
| Year-over-year Data Center growth | 46%, according to Marvell’s fiscal 2026 annual report | 68%, according to NVIDIA’s fiscal 2026 results |
| Share of reported company revenue | 74% of Marvell’s $8,194.6 million total revenue | Approximately 89.7% when $193.7 billion is divided by NVIDIA’s $215.9 billion total revenue; the companies use different segment definitions and fiscal dates |
| Core description | Custom ASICs, interconnects, Ethernet, Fibre Channel, processors and storage controllers | Accelerated-computing platform for AI and data processing, including GPUs and networking |
| Role in the announced NVLink Fusion relationship | Custom XPUs, compatible scale-up networking and silicon-photonics collaboration | Vera CPU, ConnectX NICs, BlueField DPUs, NVLink, Spectrum-X switches and rack-scale AI-compute technologies |
These are company-reported segment results, not a market-share table. A simple division of the two Data Center figures produces a ratio of about 31.8 to 1 in NVIDIA’s favor, but that arithmetic does not measure shipments, installed systems, customers, profit or dominance in every AI infrastructure category.
What Marvell’s AI data-center business includes
Custom silicon is the central differentiator
Marvell’s fiscal 2026 filing describes a portfolio built around custom application-specific integrated circuits (ASICs) and connectivity. A custom ASIC is designed for a particular customer’s workload or system rather than sold as a broadly programmable processor. In an AI data center, that can mean an accelerator, networking device, storage controller or another chip optimized for a defined architecture, power envelope and software environment.
The practical implication is that Marvell generally participates as an enabling supplier inside a customer’s system. Its products can be essential to an AI server or network without Marvell selling a complete, branded rack-scale platform. This is a different commercial and technical role from a vendor whose reported Data Center segment includes a broad accelerator, interconnect and systems platform.
Connectivity and infrastructure semiconductors
Marvell lists interconnects, Ethernet solutions, Fibre Channel adapters, processors and storage controllers alongside custom ASICs. Those products address the movement and management of data around AI compute: links between chips, traffic across servers, connections to storage and interfaces used in data-center networks.
Marvell also said it was developing UALink and Ethernet for Scale-Up Networking switches for the emerging scale-out AI market. These statements describe development activity and product direction. They should not be read as evidence that every announced design was shipping at a particular volume or had generated a specified amount of revenue by the end of fiscal 2026.
Free tools Windows power users keep installed
One-click scans. No signup required.
How Marvell describes its revenue mix
In its 2026 proxy statement, Marvell characterized optical interconnect as roughly half of its data-center revenue and custom silicon as approximately 25%. Those are management descriptions of the segment, not an independently audited product-by-product bridge. The categories may reflect different ways of discussing the business and should not be added together as if they were mutually exclusive portions of a pie chart.
The mix explains why Marvell’s AI exposure can grow even when it is not selling the primary accelerator in a system. More AI compute increases the need for high-bandwidth links, optical modules and custom chips that connect or support those accelerators. It also leaves Marvell exposed to design-win timing, customer concentration and the success of the architectures into which its components are designed.
What NVIDIA’s Data Center segment includes
A broader accelerated-computing platform
NVIDIA’s annual review describes its Data Center platform as focused on accelerating compute-intensive workloads, including AI and data processing. Its fiscal 2026 reporting combines a much broader platform story than Marvell’s component-oriented disclosures: GPUs and networking are both part of the Data Center business, and NVIDIA presents the hardware, interconnect and software environment as a coordinated architecture.
That breadth matters when comparing revenue. A customer buying an NVIDIA-based AI system may obtain compute accelerators, networking, interconnect technology and rack-scale infrastructure from the same platform family. Marvell can supply important pieces of such a system, but its reported segment is not defined in the same way.
Recommended Free Tools
Rank #2
- Part number 900-53651-2500-000 and model: P3651
- This is the 2 slot version for when there is no empty slots between 2 slot cards. If you have one or more empty slots between the cards or the cards are 3 slot this NVLink will not work. See the attached images showing the card layout.
- NVLink 3.0 for any brand of RTX Ampere model graphics cards: 3090, A30, A40, A100 / H100 (Requires three NVLinks), A800, A4500, A5000, A5500, A6000
- This is the same as PNY part number: NVLAMP-2SLOT-BSP and RTXA6000NVLINK-KIT
- This is the same as Dell part number: 0RWJ7Y
Fiscal 2026 scale
NVIDIA reported $193.7 billion in Data Center revenue for fiscal 2026, up 68% year over year. Its fourth-quarter Data Center revenue was $62.3 billion for the quarter ended January 25, 2026. NVIDIA’s total fiscal 2026 revenue was $215.9 billion.
The quarterly figure is useful for showing the pace of the business, but it should not be combined with Marvell’s full-year number. Marvell’s fiscal year ended January 31, 2026, six days after NVIDIA’s, and the companies report different segment scopes.
NVLink Fusion and customer-specific compute
NVIDIA’s fiscal 2026 report says it introduced NVLink Fusion so hyperscalers and custom-ASIC designers can integrate custom CPUs and XPUs with NVIDIA’s platform. This positions NVIDIA not only as a seller of its own accelerators, but also as a platform provider that can connect customer-specific compute designs to its interconnect and networking technologies.
The disclosure establishes the integration concept. It does not establish how many customers deployed it, the volume of integrated systems, or revenue attributable to those integrations.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Scale, growth and concentration: what the numbers really say
Revenue scale
Marvell’s $6,100.3 million Data Center revenue represented 74% of its $8,194.6 million company revenue in fiscal 2026. Data Center is therefore the majority of Marvell’s business and a central driver of its corporate results.
NVIDIA’s $193.7 billion Data Center revenue was approximately 31.8 times Marvell’s reported Data Center revenue when the two fiscal-year figures are divided. NVIDIA’s Data Center figure also represented approximately 89.7% of its $215.9 billion total revenue. These percentages are simple calculations from company totals; they are not equivalent measures of market share because the segment boundaries, product breadth and fiscal calendars differ.
Growth rates
Marvell reported 46% year-over-year growth in Data Center sales, attributing the increase to strong AI-related demand. NVIDIA reported 68% year-over-year Data Center growth. NVIDIA therefore grew faster on the reported metric and from a far larger base during these fiscal periods.
Rank #3
- Video/Sound Cards
- Passive Cooling
A growth percentage alone does not identify the source of future returns. Marvell’s result can be highly sensitive to a small number of custom programs moving into production, while NVIDIA’s result reflects a much larger platform business. Neither figure is a forecast for fiscal 2027, and neither should be treated as a guarantee of continued growth.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Why concentration has two meanings
For Marvell, a 74% Data Center share means AI infrastructure demand has become important to company-wide performance. For NVIDIA, the roughly 89.7% calculated share indicates an even stronger concentration in the reported Data Center category. Concentration can magnify upside when AI infrastructure spending rises, but it can also magnify exposure to customer budgets, supply constraints, product transitions and changes in system architecture.
Different positions in the AI infrastructure stack
| Layer or need | Marvell’s disclosed emphasis | NVIDIA’s disclosed emphasis |
|---|---|---|
| Compute | Custom ASICs and, through the announced partnership, custom XPUs | Data-center GPUs and an accelerated-computing platform; NVLink Fusion also supports integration of custom CPUs and XPUs |
| Scale-up links | Interconnects and development of Ethernet for Scale-Up Networking switches; NVLink Fusion-compatible networking is planned | NVLink interconnect and supporting platform technologies |
| Scale-out networking | Ethernet solutions and development work aimed at AI markets | ConnectX NICs, BlueField DPUs and Spectrum-X switches in the announced platform description |
| Optical connectivity | Optical interconnect described by Marvell as roughly half of Data Center revenue in its 2026 proxy | Silicon-photonics collaboration with Marvell announced as part of the partnership |
| Storage and adjacent infrastructure | Fibre Channel adapters, processors and storage controllers | Data Center reporting is centered on accelerated computing and networking rather than the storage-controller portfolio listed by Marvell |
The table shows why a direct “which chip is better?” comparison is misleading. Marvell’s disclosed strengths are often inside the system, connecting, moving or customizing compute. NVIDIA’s disclosed business is organized around a larger integrated platform whose center is accelerated computation.
The March 31, 2026 NVLink Fusion partnership
What Marvell is expected to contribute
The announcement says Marvell will provide custom XPUs and NVLink Fusion-compatible scale-up networking. It also describes collaboration on silicon photonics. In practical terms, Marvell is seeking to connect its custom-silicon and connectivity capabilities to NVIDIA’s AI factory and AI-RAN ecosystem.
What NVIDIA is expected to contribute
NVIDIA’s announced supporting technologies include the Vera CPU, ConnectX NICs, BlueField DPUs, NVLink interconnect, Spectrum-X switches and rack-scale AI-compute technology. These components provide the surrounding platform in which a custom XPU can operate and communicate.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Why the agreement is both competitive and complementary
Marvell and NVIDIA can compete for parts of the same infrastructure budgets: custom accelerators, networking silicon and high-bandwidth interconnects are all strategic choices for AI system designers. At the same time, Marvell’s custom devices can be integrated with NVIDIA technologies under NVLink Fusion. That makes the partnership an example of coopetition rather than proof that one company has replaced the other.
What the announcement does not prove
- It does not prove that a named customer has deployed a Marvell XPU through NVLink Fusion.
- It does not disclose shipment quantities, production timing or realized revenue.
- It does not show how much of either company’s future Data Center revenue will come from the relationship.
- It does not make Marvell’s and NVIDIA’s products interchangeable; their roles remain different.
How to interpret the comparison for different questions
If the question is “Who has the larger AI data-center business?”
NVIDIA, by a very wide margin on reported fiscal 2026 Data Center revenue. Its $193.7 billion segment is about 31.8 times Marvell’s $6,100.3 million figure, subject to the segment and fiscal-period caveats above.
Rank #4
- CUDA Cores: 4608 / NVIDIA Tensor Cores: 576 / NVIDIA RT Cores: 72
- GPU Memory: 24 GB GDDR6 with ECC / Bandwidth: 624 GB/Sec
- System Interface: PCI Express 3.0 x16
- Four DisplayPort 1.4 Connectors
- 3D Stereo Support with Stereo Connector
If the question is “Who is more dependent on data centers?”
Both companies are highly exposed, but the reported shares are different. Data Center was 74% of Marvell’s fiscal 2026 revenue. Using NVIDIA’s reported totals, Data Center was approximately 89.7% of fiscal 2026 revenue. The comparison is informative about concentration, not about identical business composition.
If the question is “Who supplies the complete AI platform?”
NVIDIA’s disclosures support that description more strongly. Its Data Center platform combines accelerated computing and networking, and its announced stack includes GPUs, CPUs, NICs, DPUs, interconnects, switches and rack-scale systems. Marvell’s disclosures support a description as a supplier of custom silicon and connectivity that can be designed into broader systems.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11If the question is “Can a custom accelerator work with NVIDIA infrastructure?”
NVLink Fusion is NVIDIA’s announced mechanism for enabling that type of integration. Marvell is one announced participant, contributing custom XPUs and compatible scale-up networking. The commercial availability, deployment scale and financial contribution of those integrations remain unspecified in the announcement.
Important limits on a financial comparison
Segment definitions are not standardized
“Data Center” is a company-defined reporting category. NVIDIA includes a broad accelerated-computing and networking platform. Marvell’s category includes custom ASICs, interconnects, Ethernet, storage and other semiconductor solutions. A dollar in one segment is not necessarily comparable to a dollar in the other in terms of product mix, software content, system value or gross margin.
Fiscal dates are close but not identical
Marvell’s fiscal year ended January 31, 2026; NVIDIA’s ended January 25, 2026. That six-day difference is small, but the more important issue is that each company closes its books under its own calendar and reporting conventions. Quarterly comparisons require matching periods rather than mixing a full year with a quarter.
Revenue is not market share
Reported revenue can be affected by pricing, product mix, accounting treatment, acquisitions, customer concentration and whether a company sells a component or a complete platform. The ratio between Marvell’s and NVIDIA’s revenue therefore cannot answer how many AI servers use each company’s technology or what percentage of the addressable market each controls.
Plans are not completed results
UALink development, Ethernet for Scale-Up Networking, NVLink Fusion integrations and silicon-photonics collaboration are strategically important disclosures. Unless a company reports production, customer adoption or revenue, they should be treated as plans or announced capabilities rather than achieved operating results.
What to watch after fiscal 2026
- Marvell’s custom-program conversion: whether announced or developing ASIC and XPU programs move into production and contribute recurring Data Center revenue.
- Optical-interconnect demand: whether the roughly half-of-Data-Center characterization in Marvell’s proxy remains representative as product mix changes.
- Scale-up networking adoption: customer use of UALink and Ethernet approaches in AI systems, and whether those designs produce measurable shipments.
- NVLink Fusion milestones: public evidence of supported products, customer deployments, production schedules or revenue rather than announcement language alone.
- NVIDIA platform breadth: the balance among GPUs, networking, rack-scale systems and integrations with customer-specific CPUs or XPUs.
- Revenue concentration and cyclicality: whether the exceptionally high Data Center shares of both companies’ revenue persist through product transitions and changes in AI infrastructure spending.
Bottom line
NVIDIA is the clear leader in reported AI Data Center business scale: $193.7 billion of fiscal 2026 revenue versus Marvell’s $6,100.3 million. But the comparison is not a contest between identical products. NVIDIA reports a broad accelerated-computing and networking platform, while Marvell supplies custom silicon and the connectivity that can make large AI systems work.
The most accurate view is therefore two-part. NVIDIA has the larger platform, revenue base and reported growth rate. Marvell is a smaller but strategically important infrastructure supplier whose custom ASIC, optical-interconnect and networking capabilities can compete with parts of NVIDIA’s stack while also integrating with it through the announced NVLink Fusion partnership. Future customer deployments and reported revenue—not the announcement alone—will determine how significant that cooperation becomes.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problems
