Marvell Technology has pursued an aggressive semiconductor manufacturing strategy since 2018. The company opted to skip the 7-nanometer process node, moving directly to 5-nanometer technology to accelerate its development of cloud infrastructure components.
The company continues to advance its process nodes, currently working with 3-nanometer, 2-nanometer, and 14-angstrom technologies. This roadmap serves as the foundation for its current position in the cloud and AI infrastructure markets.
Marvell is focusing on SerDes technology to improve power efficiency in AI clusters and data centers. In September, the firm introduced 2-nanometer optical interconnects designed to support scale-up architectures where processor-to-processor data transfer accounts for 85% of traffic.
In August, Marvell announced a Photonic Fabric Memory Module that integrates DDR5 memory, high-bandwidth memory caching, and optical connectivity. This module allows for the decoupling of memory capacity from compute resources within AI systems.
The company has secured significant industry partnerships and investments. Nvidia Corp. invested $2 billion in Marvell earlier this year, and the company has reached an agreement to design custom AI chips for Google LLC.
