Marvell Technology, Inc. (MRVL) has introduced a new 1.6-terabit pluggable optical module and coherent digital signal processor (DSP) chips manufactured on a 2-nanometer process node, featuring built-in security capabilities. The products are specifically engineered to address the growing data transfer demands within and between artificial intelligence cloud data centers.
Targeting AI Infrastructure Bottlenecks
The announcement comes as cloud service providers rapidly scale their AI compute capacity, only to encounter significant challenges in moving vast datasets swiftly between server clusters, equipment racks, and geographically dispersed data centers. The distributed nature of advanced AI workloads is dramatically increasing data center interconnect (DCI) traffic, creating a critical need for higher-bandwidth, more power-efficient optical solutions.
Pluggable optical transceivers, which insert directly into network switches or routers, simplify infrastructure by reducing the need for separate transport equipment. The new Marvell modules adhere to the ZR and ZR+ industry specifications for coherent optics used in these interconnects. The coherent DSP performs the intensive signal processing required for reliable high-speed data transmission over fiber optic cables. The integrated MACsec (Media Access Control Security) provides link-layer encryption for securing Ethernet traffic.
Product Details and Timeline
Marvell's COLORZ 1600 pluggable module is powered by the company's new Electra coherent DSP, fabricated on an advanced 2nm semiconductor process. The company also announced an updated Libra DSP to support its existing COLORZ 800 pluggable module. Marvell plans to publicly debut these technologies at the OFC 2026 conference in Los Angeles, scheduled for March 15–19.
Sampling of the Electra and Libra DSPs, along with the corresponding COLORZ optical modules, is anticipated to commence in the second half of 2026. The company emphasized that serving the global AI data center market requires demonstrated strength in high-volume manufacturing.
"Meeting the global needs of AI-driven data centers requires proven, large-scale manufacturing strength," stated Russ Esmacher, Senior Vice President and General Manager for Data Center Interconnect at Marvell. Industry analyst Scott Wilkinson of Cignal AI noted that maintaining technology leadership in this segment hinges on power efficiency, critical feature integration, and manufacturability at scale.
Competitive and Market Context
The market for coherent pluggable optics is highly competitive. Rivals include Cisco Systems' Acacia unit, which offers 800G ZR/ZR+ optics designed to deliver 800 gigabits per second from a single switch port over data center links. This competition pressures all vendors to continuously increase transmission speeds while simultaneously reducing power consumption—a key metric for data center operators.
Marvell's core business is designing data infrastructure semiconductors for the data center, enterprise networking, carrier, and edge computing markets, spanning networking, storage, and security solutions. The company is led by CEO Matthew Murphy.
In Thursday morning trading, shares of Marvell Technology declined approximately 2.5% to $76.12, giving the company a market valuation of roughly $80.8 billion. Investor attention is now turning to the company's quarterly earnings conference call scheduled for 4:45 p.m. Eastern Time later today. Commentary on the demand outlook for AI-driven networking solutions is expected to be as significant to investors as the financial results themselves.
Future Considerations and Risks
While the new optics and chips represent a significant technological step, they are not expected to reach customers for sampling until the latter part of 2026. Production ramps on leading-edge semiconductor nodes like 2nm are complex and can experience delays. Furthermore, the commercial success of these products could be influenced by several factors, including the timing of cloud provider infrastructure upgrades, competitive pressure from rival module vendors, and the potential adoption of in-house designed solutions by large hyperscalers. The pricing and volume trajectory for Marvell's new products may look different by the time they are fully available on the market.
