Detailed Narrative
AI Custom Silicon Platform Expansion
Marvell is expanding its custom platform with innovative technologies such as custom high-bandwidth memory (HBM) compute architecture and co-packaged optics. The custom HBM compute architecture optimizes I/O interfaces for AI accelerators, increasing performance and reducing run times. Co-packaged optics enable a transition from copper to optical fiber for scale-up AI clusters, significantly expanding Marvell's interconnect revenue and market opportunities by enabling larger AI servers with higher memory capacity and processing capability.
Strategic Partnerships and Advanced Packaging
The company announced a partnership with NVIDIA to integrate NVLink Fusion technology into its custom platform, providing customers with an accelerated path to custom scale-up solutions. Marvell also introduced a new multi-die packaging platform, which is already qualified and in production for a customer-specific XPU program. This platform offers more efficient die-to-die interconnect, lower power consumption, increased yields, and lower product costs, presenting an alternative to traditional silicon interposers for custom cloud applications.
Custom AI XPU Program Momentum
Marvell's lead XPU program for a large U.S. hyperscale data center customer is a key revenue driver, having achieved volume production with A0 silicon. The company has secured 3-nanometer wafer and advanced packaging capacity for this program, with production expected to start in calendar 2026. Engagement on follow-on generations is progressing for this customer and another U.S. hyperscaler, reinforcing confidence in achieving long-term custom revenue goals.
Interconnect Portfolio Leadership
Marvell's PAM and DCI franchises continue to lead in enabling AI and cloud infrastructure build-out. Recent showcases at OFC 2025 included the industry's first 400-gig per lane PAM technology, a critical step towards 3.2T optical interconnects. Other innovations include 3-nanometer 1.6T PAM4 DSPs, next-generation 800-gig DCI modules, production-ready 1.6T AEC DSPs, Coherent light DSPs, and PCIe Gen 6 and Gen 7 SerDes, all designed to power next-generation AI deployments.
End-Market Recovery and AI Tailwinds
Marvell is observing an encouraging recovery in its carrier infrastructure and enterprise networking end markets, with the Q2 FY26 forecast marking the fifth consecutive quarter of sequential revenue growth for these combined segments. The company continues to benefit from strong AI tailwinds, including robust capital expenditure plans from hyperscalers, increasing sovereign data center announcements, and the emergence of new hyperscalers, all contributing to long-term confidence in the data center business.