Archive for the 'AI' Category
-
May 05, 2026
MACsec: A Shift in Security for Scale-across Networking
By Joseph Chon, Senior Director, Product Marketing, Data Center Interconnect, Marvell
MACsec is moving to the module in scale-across networks.
Media Access Control security (MACsec) is a foundational technology for protecting data in motion. It encrypts and authenticates Ethernet traffic to guard against eavesdropping, denial-of-service attacks, intrusion and other security threats while also strengthening overall data integrity. Embodied in silicon, MACsec further establishes a robust root of trust for managing encryption keys and securing the boot process.
What’s changing is where the silicon for delivering MACsec gets located.
To date, the MACsec circuitry for long-distance scale-across networks has typically been embedded in the switch ASIC, where space and silicon real estate are at an absolute premium. Embedding MACsec into the tight confines of the ASIC raises the cost of integrating the technology. It also makes infrastructure less flexible: some upgrades require taking the system offline, reducing overall capacity.
-
April 20, 2026
OFC: Lumentum Optical Circuit Switching and Marvell Optics for Scaling AI
By Todd Rope, Vice President of Software Engineering at Marvell
Optical circuit switching (OCS) has become one of the fastest growing segments in networking with revenue expected to exceed $3.5 billion by 2029, more than 2x over 2025.1 The unique architecture of OCS systems, however, also mean that developers and data center operators need to ensure that these systems can seamlessly integrate into data infrastructure and interoperate with existing product lines.
Lumentum and Marvell took a significant step toward that goal with a live demonstration at OFC 2026 that combined the R300 OCS system from Lumentum with different classes of modules powered by Marvell optical DSPs. The modules included inside-the-data center modules powered by the Marvell® Ara 1.6T (5m-2km interconnects), coherent lite modules with 1.6T Marvell Aquila for campus-size connections (2 to 20km) and long-range COLORZ® 800T ZR/ZR+ modules for 10-1000km data center interconnects.
Marvell RELIANT™, a new software platform for analyzing equipment performance and optimizing networks in real-time, was also used to monitor data transmission, power consumption, bit error rate and other metrics in the demo. Michael DeMerchant, senior director of product line management at Lumentum and I walk you through more of what RELIANT can accomplish with OCS in the video.
-
April 15, 2026
OFC 2026: Next Generation Die-to-Die Interconnects for XPUs
By John Ma, Senior Director of Design Engineering, AMS COMPHY, Marvell
Although they are the shortest class of interconnect technologies in terms of length, die-to-die (D2D) interconnects and other so-called scale-inside technologies for moving data between compute and memory die inside XPUs and chip packaging have an outsized impact on performance, efficiency and total cost of ownership.
The growing scale and complexity of AI workloads, meanwhile, and the expected development of 3D and other multilayer chip designs mean that more data will be moving across more and longer scale-inside links and at faster speeds than ever before. Advances in packaging, high-speed I/O interfaces, and other technologies will be needed to meet the demands of tomorrow’s algorithms.
At OFC, Marvell showcased a recent advancement in scale-inside: a 3nm 40 Gbps D2D interface for linking HBM and compute die within the same chip package. The increase in speed enables designers to dramatically improve performance and latency while diving down power and silicon area needed for interfaces. The wide-open bathtub curve shown on the monitors in the video demonstrates exceptional signal integrity and reliability.
-
April 09, 2026
Live from OFC: Pioneering a New Class of Interconnects—Hybrid AEC/ACC Cables—for Optimizing Infrastructure
By Nicola Bramante, Senior Principal Engineer, Connectivity Marketing, Marvell
Why develop a hybrid cable? Because the quest for greater optimization in AI data centers never ends.
High speed cable developer and manufacturer Luxshare-Tech and Marvell showed off the industry’s first hybrid AEC/ACC solution at OFC 2026, the latest step in enhancing copper interconnects to meet the stringent power, performance and reach standards of AI infrastructure.
Active electrical cables (AECs) are designed for comparatively long (~4-9 meter) high-bandwidth connections within or between racks. The boost in reach over passive copper cables is accomplished by integrating optimized AEC DSPs into the terminal ends of a cable. Active Copper Cables (ACCs), by contrast, rely on equalizers and redrivers for extending reach. ACCs consume far less power than AECs but generally are deployed for in-rack connections running 2 meters or less.
Hybrid AEC/ACC cables combine technologies from both for a solution that delivers a longer, AEC-like reach and the low latency, low power, low cost and low complexity benefits of ACC designs.
-
March 19, 2026
Marvell RELIANT: Real-time Insight for Optimizing AI Fabrics
By Todd Rope, Vice President of Software Engineering, Marvell, and Wesley Yeh, UI Architect, Marvell
Interconnects have become one of the fastest growing markets in AI as hyperscalers race to expand their infrastructure. Annual port shipments and revenue for high-speed cables (100G and above) are expected to grow by 42% and 39% through 2030, respectively, reaching a cumulative 1.4 billion units, according to research firm LightCounting.1 That surpasses the unit and revenue growth rates for GPUs, CPUs, memory devices and other major components.2
As data center interconnects extend in reach and increase in speed, their impact on total cost of ownership and return on investment will only intensify. Achieving optimal performance at scale requires live visibility into interconnect fabrics.
Marvell RELIANT (Reliable Link Analytics and Intelligence) is a new software platform that improves network health through real-time intelligence across scale-up, scale-out, and scale-across networks. Tapping into live operational data at the silicon level, RELIANT enables data center operators to visualize and monitor interconnects, conduct diagnostics and proactively prevent link flaps and other problems that can lead to downtime.
Recent Posts
- MACsec: A Shift in Security for Scale-across Networking
- Celebrating Earth Day: Technology and Sustainability in Action
- OFC: Lumentum Optical Circuit Switching and Marvell Optics for Scaling AI
- HCDS Best Paper Award Winner: Bridging the Memory Hierarchy Gap with Photonic Fabric™ Technology
- OFC 2026: Next Generation Die-to-Die Interconnects for XPUs
