Menu

OFC 2022: Marvell Debuts CPO Platform to Spur Open Switch and Optics Integration

The News: Marvell announced its first-generation cloud-optimized co-packaged optics (CPO) technology platform, designed to enable faster connectivity while reducing power consumption. The new platform includes 2.5D/3D highly integrated silicon photonics using lasers, TIAs, drivers and its PAM4 DSP. Read the Marvell Press Release here.

OFC 2022: Marvell Debuts CPO Platform to Spur Open Switch and Optics Integration

Analyst Take: Marvell showcased its CPO technology platform in a demo at the 2022 Optical Networking and Communication (OFC) Conference that featured the Marvell Teralynx switch platform along with Marvell CPO electro-optics integrated into a 1 rack-unit (RU) 32 port optical switch. The demo touted the portfolio foundation for Marvell’s planned 3.2T CPO platform for the 51.2T switch generation and spotlights laser integration into the CPO platform as well as ODM integration aimed at broadening ecosystem support.

Marvell’s entry into the CPO sweepstakes is well-timed as I see CPO technology is now poised to support commercially meaningful shipment volumes and drive tangible innovation across cloud data center environments after lengthy speculation about its commercial viability. Through CPO, the laser source can be positioned inside the same package as crucial computing electronics. Input/Output (I/O) chiplets can serve to convert received laser light into electrical signals, or to power optical signals out based on electrical inputs that need to move an especially short distance from the neighboring ASIC.

Marvell’s Teralynx offering aligns with the industry-wide momentum away from initial proprietary offerings and toward more reliable, standards-based (e.g., Coalition for On-Board Optics, OIF) CPO solutions that encourage an open ecosystem approach for switch and optics integration to assure multi-vendor interworking.

I see the new offering also advancing Marvell’s overall 2.5D/3D portfolio investments which is crucial to meeting the rapidly expanding bandwidth demands throughout multi-cloud environments. Three CPO bandwidth-intensive drivers, for instance, include HPC clusters, AI/ML engines, and military/aerospace applications. Marvell now counters the various CPO moves of AMD, Broadcom, Cisco, and Intel, providing the cloud data center ecosystem a broader range of much-needed standards-based choices.

In 2.5D packaging methodology, there is no stacking of dies on dies, as communication between chips is accomplished using either a silicon or organic interposer, usually a chip or layer using through-silicon vias (TSVs) for communication. In other words, the dies are packed into a single package, and both are flip chipped on a silicon interposer. As a result, distances are shorter and increases conduits for signals, enabling faster overall communications with less energy required to drive signals. In addition, the 2.5D architectures are coupled with stacked memory modules, especially high-performance memory, to further boost performance.

3D packaging furthers integration density providing the ability to dramatically lessen interconnect delays and improve performance while reducing circuit power consumption through shortened wire length. 3D integration also offers a flexible approach to implement the heterogeneous system-on-a-chip (SoC) design by combining distinct technologies, such as opto-electric devices, memory & logic circuits, and radio frequency (RF) components onto different dies of a 3D integrated circuit.

Key Takeaways on Marvell’s CPO Platform Debut

I believe Marvell is now delivering the 2.5D/3D portfolio innovation needed to help kickstart adoption of standards-aligned CPO technology, providing cloud data center decision makers more flexibility and confidence to adopt CPO solutions. As new switch products using increasingly viable CPO capabilities become commercially available, CPO technology can play its long-awaited yet pivotal role in meeting skyrocketing bandwidth demands, improve energy efficiency, and boost performance throughout demanding cloud data center environments.

Disclosure: Futurum Research is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.

Other insights from Futurum Research:

OFC 2022: Marvell Adds Unique Touch to New Cloud-Optimized 400G/800G PAM4 DSPs

MWC 2022: Dell Intros New Open Telco Solutions and Works with Marvell to Spur Open RAN

Marvell Technology Q4 Revenue Soars 68% YoY to $1.34B as the Semiconductor Maker Continues Its Torrid Pace Following Strategy Shift to Enterprise

Image Credit: AiThority.com

Author Information

Ron is an experienced, customer-focused research expert and analyst, with over 20 years of experience in the digital and IT transformation markets, working with businesses to drive consistent revenue and sales growth.

Ron holds a Master of Arts in Public Policy from University of Nevada — Las Vegas and a Bachelor of Arts in political science/government from William and Mary.

Related Insights
Collapsing the Stack VAST Data’s Bid to Own the AI Data Loop
February 27, 2026

Collapsing the Stack: VAST Data’s Bid to Own the AI Data Loop

Brad Shimmin, Vice President at Futurum, analyzes the VAST Data platform updates from VAST Forward, detailing how the new Policy Engine, Tuning Engine, and Polaris architectures are simplifying the AI...
Are Enterprises Ready for the Virtualization Reset, or Just Swapping Out One Complexity for Another
February 27, 2026

Are Enterprises Ready for the Virtualization Reset, or Just Swapping Out One Complexity for Another?

Futurum’s Alastair Cooke shares his insights on new HPE research that finds that only 5% of enterprises are fully prepared for the so-called Great Virtualization Reset, even as two-thirds plan...
February 18, 2026

Hybrid and Multi-Cloud Object Storage for AI – Futurum Signal

AI workloads are reshaping enterprise infrastructure strategy. As organizations scale model training, fine-tuning, and inference across environments, traditional storage...
Arista Networks Q4 FY 2025 Revenue Beat on AI Ethernet Momentum
February 16, 2026

Arista Networks Q4 FY 2025: Revenue Beat on AI Ethernet Momentum

Futurum Research analyzes Arista’s Q4 FY 2025 results, highlighting AI Ethernet adoption across model builders and cloud titans, growing DCI/7800 spine roles, AMD-driven open networking wins, and a Q1 guide...
Enterprises Should Finally Stop Paying SSD Prices for Cold Data
February 16, 2026

Enterprises Should Finally Stop Paying SSD Prices for Cold Data

Alastair Cooke, Research Director, Cloud and Data Center at Futurum, shares his insights on how CTERA enables customers to use cloud object storage combined with small amounts of SSD to...
IBM’s New FlashSystem Might Be the Blueprint for AI-Driven Storage Resilience
February 16, 2026

IBM’s New FlashSystem Might Be the Blueprint for AI-Driven Storage Resilience

Alastair Cooke, Research Director at Futurum, shares his insights on IBM’s latest FlashSystem release with Agentic AI features to minimize manual operations and simplify compliance....

Book a Demo

Newsletter Sign-up Form

Get important insights straight to your inbox, receive first looks at eBooks, exclusive event invitations, custom content, and more. We promise not to spam you or sell your name to anyone. You can always unsubscribe at any time.

All fields are required






Thank you, we received your request, a member of our team will be in contact with you.