Marvell to present on networking innovations at Open Compute Summit

As AI workloads grow in scale and complexity, data centers face new challenges in managing bandwidth, thermal performance and power efficiency. Marvell is expanding its connectivity portfolio with the addition of active copper cable (ACC) linear equalizers, extending the reach of copper interconnects while maintaining minimal latency and power efficiency. Leveraging Marvell industry-leading PAM4 technology and expertise in 100G/lane and 200G/lane analog devices, the new Marvell® ACC linear equalizers support 800G and 1.6T copper interconnects and expand the company’s scale-up connectivity solutions, which include AEC and AOC chipsets. Marvell is showcasing its latest advancements in accelerated infrastructure at the Open Compute Project Foundation Global Summit this week. Read the full announcement: https://mrvl.co/42GdmHY

At the Open Compute Project (OCP) 2025 Global Summit, Broadcom introduced Ethernet Scale Up Networking (ESUN), a groundbreaking approach designed to advance Ethernet-based AI infrastructure. As AI workloads continue to expand, modern data centers require networks that can handle massive bandwidth, ultra-low latency, and efficient data movement between thousands of GPUs. ESUN addresses this challenge by redefining how Ethernet can scale vertically within AI clusters. Unlike traditional Ethernet designs optimized for general-purpose connectivity, ESUN is built specifically for scale-up AI systems. It enables direct, high-bandwidth communication between accelerators while maintaining the open and interoperable nature of Ethernet. This innovation helps bridge the gap between proprietary interconnects and open networking standards, making AI clusters more flexible, efficient, and cost-effective to deploy. Broadcom’s ESUN integrates advanced features such as adaptive routing, congestion control, and precise flow management, ensuring predictable performance even under intense AI workloads. By leveraging Broadcom’s extensive silicon and networking expertise, ESUN provides the scalability needed for next-generation training and inference environments. As AI continues to reshape industries, from cloud computing to semiconductor design, open innovation through OCP and technologies like ESUN will be crucial in accelerating adoption while reducing infrastructure complexity. Ethernet, now more than ever, stands at the center of scalable AI networking, enabling organizations to build faster, smarter, and more sustainable data centers. Related News: https://lnkd.in/g53re7vG #Broadcom #2025OCP #Ethernet #AIInfrastructure #DataCenter #Networking #AI #CloudComputing #Semiconductors #Innovation #ESUN #EthernetScaleUpNetworking

Cadence tapes out its next-generation 32G UCIe advanced package IP subsystem on TSMC's 3nm process, offering: • Enhanced high-speed die-to-die connectivity with flexible data rates up to 32Gbps • Universal interoperability • Self-calibration, hardware-based bring-up • Robust performance across industrial condition • Protocol support to meet the demands of AI/ML, HPC, and datacenter applications. Learn more here: https://ow.ly/svQf30sPSiA

Cadence tapes out its next-generation 32G UCIe advanced package IP subsystem on TSMC's 3nm process, offering: • Enhanced high-speed die-to-die connectivity with flexible data rates up to 32Gbps • Universal interoperability • Self-calibration, hardware-based bring-up • Robust performance across industrial condition • Protocol support to meet the demands of AI/ML, HPC, and datacenter applications. Learn more here: https://ow.ly/WIbb30sPRso

Cadence tapes out its next-generation 32G UCIe advanced package IP subsystem on TSMC's 3nm process, offering: • Enhanced high-speed die-to-die connectivity with flexible data rates up to 32Gbps • Universal interoperability • Self-calibration, hardware-based bring-up • Robust performance across industrial condition • Protocol support to meet the demands of AI/ML, HPC, and datacenter applications. Learn more here: https://ow.ly/8omF30sPVKN

Cadence tapes out its next-generation 32G UCIe advanced package IP subsystem on TSMC's 3nm process, offering: • Enhanced high-speed die-to-die connectivity with flexible data rates up to 32Gbps • Universal interoperability • Self-calibration, hardware-based bring-up • Robust performance across industrial condition • Protocol support to meet the demands of AI/ML, HPC, and datacenter applications. Learn more here: https://ow.ly/2EWV30sPYVU

Cadence tapes out its next-generation 32G UCIe advanced package IP subsystem on TSMC's 3nm process, offering: • Enhanced high-speed die-to-die connectivity with flexible data rates up to 32Gbps • Universal interoperability • Self-calibration, hardware-based bring-up • Robust performance across industrial condition • Protocol support to meet the demands of AI/ML, HPC, and datacenter applications. Learn more here: https://ow.ly/mKm330sQ3wj

Quantum breakthroughs, AI-native processors, and photonic interconnects are reshaping how data is computed, transferred, and secured. From intelligent edge systems to hyperscale data centers, microelectronics innovation is driving the next leap in performance, connectivity, and energy efficiency. Our latest analysis identifies 8 key growth areas and 20+ leading companies building adaptive, scalable microelectronics ecosystems — from quantum & neuromorphic processing to advanced memory, networking, and next-gen packaging. https://hubs.la/Q03LhGWd0 Engage with us to understand how microelectronics advancements are fueling APAC’s leadership in next-gen computing and connectivity. https://hubs.la/Q03LhHbZ0

Leveraging our cutting-edge silicon photonics technology, the 800G OSFP DR8 delivers low-latency, highly-reliable connectivity for AI networks and data centers. Compatible with InfiniBand and Ethernet, it supports up to 500m reach for high-speed connections. #800G #DataCenter #OpticalTransceivers #Networking 

Broadcom’s new Tomahawk Ultra processor marks a major step forward in AI-era networking. Unveiled in July 2025, the Tomahawk Ultra is designed to accelerate large-scale chip-to-chip communication — a critical bottleneck in modern AI infrastructure. The processor enables higher bandwidth efficiency inside data centers, connecting more chips simultaneously while reducing latency across high-performance computing systems. In the AI acceleration race, compute power alone is not enough — network efficiency now defines real performance. Broadcom’s latest innovation reinforces the importance of interconnect technologies that bind the next generation of GPUs and accelerators together. Connectivity is becoming the new frontier of computation. #AI #Semiconductors #Broadcom #Networking #DataCenters #HPC #Innovation