Marvell Targets AI Data Center Bottlenecks with End-to-End Connectivity Push at OFC 2026
Event summary
- Marvell will showcase over 20 connectivity solutions for AI data centers at OFC 2026 (March 15–19).
- Key demos include 40G die-to-die IP, PCIe 8.0 SerDes, and 1.6T PAM4 optical interconnects.
- Marvell claims its portfolio addresses power, latency, and bandwidth bottlenecks in hyperscale AI infrastructure.
- The company will participate in executive forums and panels during the event.
The big picture
Marvell’s OFC 2026 showcase underscores the growing importance of connectivity in AI data centers, where bandwidth and latency demands are outpacing traditional semiconductor solutions. The company’s focus on die-to-die and optical interconnects aligns with the industry’s shift toward multi-rack AI clusters, where efficiency and scalability are critical. If successful, Marvell’s portfolio could become a foundational layer for next-generation AI infrastructure.
What we're watching
- Execution Risk
- Whether Marvell can deliver on its claims of power and latency efficiency in real-world AI deployments.
- Market Adoption
- The pace at which hyperscalers and cloud providers integrate Marvell’s connectivity solutions into next-gen AI clusters.
- Competitive Positioning
- How Marvell’s end-to-end portfolio will differentiate it from rivals like Broadcom and Intel in the AI data center space.
Related topics