Marvell Technology

Real‑time visibility into interconnect health is becoming essential as AI infrastructure scales. In a new blog, Todd Rope and Wesley Yeh introduce Marvell RELIANT, a software platform that delivers actionable, real‑time insight across scale‑up, scale‑out and scale‑across networks. By tapping live operational data at the silicon level, RELIANT helps data center operators visualize and monitor interconnects, conduct diagnostics and proactively prevent issues that can lead to downtime. Learn how RELIANT provides fleet‑wide visibility and predictive analytics for optimizing AI fabrics: mrvl.co/4sr2B7a

Matt Murphy, @MarvellTech Chairman and CEO, to Keynote at COMPUTEX 2026! ✨ Marvell is a global leader in data infrastructure semiconductor solutions, powering the next generation of AI infrastructure with a comprehensive portfolio of end-to-end connectivity technologies and custom silicon solutions for hyperscalers and cloud data center customers. 📅Date: June 2 at 10:30 a.m. (GMT+8) 📍Venue: Taipei Nangang Exhibition Center, Hall 2, 7F More Information: events.computextaipei.com.tw/en/ #computex2026 #taipei #techevent #global #marvell

Excited to announce Navin Bishnoi, AVP & India Country Manager, @MarvellTech, as a speaker at the upcoming Asia Dialogues Forum 2026 in Bengaluru. #ADFBangalore #AsiaDialogues2026 #AsiaDialoguesForum #CapabilityBuilding #DeepTech #SpeakIn

Marvell has been named one of the World’s Most Ethical Companies for 2026 by @ethisphere. This recognition highlights the Marvell commitment to business integrity through strong ethics, compliance, sustainability, and governance programs. Marvell is one of only five companies in the semiconductor industry to receive this designation, based on Ethisphere’s comprehensive assessment of corporate governance, ethical culture, and responsible business practices. Learn more about the recognition and what it reflects about Marvell values: mrvl.co/4dtwEq0

Marvell has been recognized for two key optical connectivity innovations in the 2026 @lightwaveonline Innovation Reviews. The 1.6T silicon photonics light engine and the active copper cable linear equalizers each received a 4.0 outstanding honoree rating for delivering clear technical advantages in performance, power efficiency, and system integration. The 1.6T light engine provides eight lanes of 200G PAM4 optical connectivity in a single package and is purpose‑built for AI scale‑up, supporting future CPO architectures. The ACC linear equalizers extend copper reach while maintaining ultra‑low latency, offering a practical on‑ramp to 1.6T deployment in existing data center designs. These recognitions demonstrate how Marvell continues to advance the optical technologies that enable efficient, scalable AI infrastructure. Read the full article for more details: mrvl.co/474d6Vp

The @ofcconference has officially kicked off and Marvell is excited to showcase its latest end-to-end connectivity innovations. On Monday, the @OpticaWorldwide Executive Forum held its annual event bringing together industry thought leaders to discuss the future of scale‑out and scale‑up data center networks and the technologies driving next‑generation AI infrastructure. Marvell was proud to be part of the discussion. Across the show floor, technology partners are showcasing more than 80 demos powered by Marvell devices and technologies, underscoring the pivotal role Marvell plays in driving the next evolution in AI infrastructure innovation. Stop by booth 1600 to experience more than 20 live demonstrations featuring the latest Marvell advances in its end-to-end connectivity portfolio for scale-up, scale-out and scale-across applications.

PCIe is evolving to meet the demands of large‑scale AI systems. In a new Marvell blog, Krishna Mallampati examines how PCIe scale‑up fabrics are emerging as a critical foundation for AI infrastructure that spans hundreds of processors across multiple racks. The article outlines the architectural requirements for scale‑up environments and explains how monolithic PCIe switch designs can deliver the low latency, high lane density, and energy efficiency needed to support next‑generation AI workloads. Learn how PCIe scale‑up fabrics are shaping the future of AI data center networking: mrvl.co/3PGov7M

The AI memory wall is becoming one of the most pressing constraints in modern data centers. In a new Marvell blog, Jianping Jiang explains how Structera S CXL switches enable shared, scalable memory architectures that improve utilization, lower latency, and boost application performance across AI workloads. The article outlines how CXL memory pooling, acceleration, and expansion work together to help data center operators overcome capacity and bandwidth limitations while reducing cost and power. Read about how Structera S is helping scale AI infrastructure beyond the memory wall: mrvl.co/478vP21

@Lumentum is excited to partner with @MarvellTech for a joint demo on optical circuit switching for next-generation #AI scale-up infrastructure at this year’s @OFCConference. 🔗 Marvell press release: bit.ly/4s8QpYB 🔗 Lumentum #OFC26 updates: bit.ly/40xOd0N

Marvell today announced Structera™ S 60260, the industry’s first 260-lane PCIe 6.0 switch. Leveraging industry-leading interconnect solutions through the recent acquisition of XConn Technologies, the new industry’s highest-radix PCIe switch extends the Marvell end-to-end PCIe portfolio, offering unique levels of scale-up AI data center performance and design flexibility. As AI servers integrate increasing numbers of GPUs and accelerators, PCIe scale-up fabrics have become critical to maximizing compute density and resource utilization. Traditional switches require multiple devices to achieve scale, resulting in increased power, latency, board complexity and total cost of ownership. The Marvell Structera S 260-lane PCIe 6.0 switch eliminates the need for multiple smaller switches, enabling higher density, lower latency and improved system efficiency. Read the full press release: mrvl.co/4lzpKSl

Marvell today announced Structera™ S 30260―a new 260-lane CXL switch device that enables rack-level memory pooling. Working in concert with Structera A near-memory accelerators, Structera X memory-expansion controllers and Alaska P PCIe/CXL retimers, the new switch allows data center operators to further increase memory bandwidth and capacity by providing access to disaggregated memory resources outside the server. Leveraging industry-leading switch solutions through the recent acquisition of XConn Technologies, the Structera S CXL switch device enables true memory pooling across the rack, allowing hyperscalers and data center operators to dynamically expand and allocate memory resources across CPUs, GPUs, XPUs and other accelerators without replacing existing platforms or relying solely on HBM stacking. It enables a large scalable, composable resource that can deliver higher memory utilization, improved data flow efficiency and AI application performance, greater infrastructure design flexibility and scalability, and reduced overall total cost of ownership. Read the full press release: mrvl.co/3PHmhVJ

Marvell has joined the eXtra dense Pluggable Optics (XPO) Multi‑Source Agreement (MSA) as a founding member. In this article, Xi Wang discusses how the XPO MSA is defining a new optical transceiver form factor designed to increase bandwidth density and support AI‑scale infrastructure through innovations such as liquid‑cooled, high‑lane‑count modules. It examines how XPO can enable denser switches, more efficient data center designs, and greater flexibility across scale‑up, scale‑out, and scale‑across architectures. Learn how XPO is accelerating innovation in optical modules for AI data centers: mrvl.co/3NGU3d9

Power efficiency is becoming a defining factor in AI infrastructure ROI. In a new Marvell blog, Henry Chen explains how Ara T, the industry’s first 1.6T transmit‑only optical DSP, delivers meaningful power savings for mid‑reach AI and cloud interconnects. Designed to reduce optical module power consumption by more than 35%, Ara T extends Marvell 1.6T leadership while helping operators manage power, cost, and scale as AI systems grow. Learn how optimized optical DSPs can improve AI infrastructure efficiency: mrvl.co/4bu5RaL

Marvell and @Lumentum today announced a live demonstration of Marvell optical connectivity solutions interoperating with Lumentum’s optical circuit switching (OCS) platform. The integrated, rack‑level system will be showcased in the Marvell booth (#1600) at OFC 2026. As AI workloads continue to drive unprecedented growth in data center traffic, network architectures are rapidly evolving to deliver higher bandwidth, lower latency and greater power efficiency. OCS enables direct optical paths between endpoints to support scalable, high‑signal‑integrity, real‑time data movement and ultra‑low‑latency connectivity across large AI fabrics. The demo will feature Aquila 1.6T coherent‑lite DSPs, Ara 1.6T PAM4 optical DSPs, and COLORZ 800 ZR/ZR+ DCI modules interoperating with the Lumentum R300 OCS, along with Marvell RELIANT, an end‑to‑end advanced telemetry, analytics and intelligence platform. Read the full press release for details: mrvl.co/4lwRwP9

At DesignCon 2026, Sonam Sadhukhan was honored in the 40 Under 40 program and selected as one of the top six speakers. The recognition highlights her contributions as a senior staff engineer on Marvell’s Optical PHY team, where she is helping build high‑speed infrastructure for the AI data center era. The article also reflects on her prior work in automotive SerDes and clocking architectures, as well as her commitment to mentoring and supporting underrepresented engineering talent. Read to learn more about Sonam’s work and leadership: mrvl.co/4bGhmgk

As AI workloads outpace the limits of standard memory technologies, the industry is accelerating development of custom and semi‑custom solutions. In a new article from @eetimes, Khurram Malik explains why hyperscalers can no longer wait for multi‑year JEDEC cycles and are turning to custom memory to gain more bandwidth, more capacity, lower power and the ability to move at the pace AI demands. He outlines how custom HBM approaches, including C‑HBM4E, can place specialized logic closer to DRAM, reduce data‑movement overhead and align memory architectures with rapidly evolving training and inference needs. Read the full piece for a deeper look at how custom memory is emerging as a critical element of next‑generation AI platforms: mrvl.co/3PfWDY1

Marvell demonstrated new advances in active electrical cable technology at DesignCon 2026.   In this video, Michael Arsenault highlights a live demo of the Golden Cable running on the first‑to‑market 224G Alaska A 1.6T DSP. The setup showcases a clean 3‑meter, 30‑gauge channel with a 10⁻¹⁰ pre‑FEC BER, along with validation that the same DSP can extend performance out to 6.5 meters.

We’re excited to share that Matt Murphy, Marvell Chairman and CEO, will join the CEO Panel at the Optica Executive Forum, one of the industry’s premier gatherings for business and technology thought leaders in optical networking. Marvell is proud to contribute to this influential dialogue with global leaders who are driving the next wave of advancements in high‑performance networking. Learn more about Marvell speakers at OFC: mrvl.co/4b8lnK9

Marvell will participate in an @ofcconference panel on recent advances in AI cluster interconnects. Preet Virk will join industry leaders to discuss the technologies shaping next‑generation AI infrastructure, including advanced switching, co‑packaged optics, optical circuit switches and high‑performance transceivers. The session will explore the market drivers and design considerations that will influence how AI clusters scale in the years ahead. Learn more about Marvell speakers and sessions on the OFC event page: mrvl.co/4b8lnK9

Marvell demonstrated its 200G/lane active copper cable technology for AI scale‑up and in‑rack connectivity at DesignCon 2026. In this video, Nicola Bramante showcases 200G/lane active copper cable technology in AI scale up and in rack connectivity. ACC is a technology that is an emerging alternative to pure copper cables; it allows for lower power, lower latency and lost cost. explains how Marvell ACC solutions deliver longer reach, thinner cables and lower power compared to traditional copper, enabled by advanced analog equalization that exceeds competitive approaches. The demo features a 1.6T, two‑meter cable connected to a SerDes system emulating in‑rack traffic, with measured bit error rates between e‑8 and e‑9 across all lanes. Marvell is excited to have its ACC as an addition to the company's AI scale up portfolio, which already includes PCIe, AEC, and AOC.