Marvell Technology has demonstrated its first 2nm silicon IP, built on TSMC’s 2nm process, for next-generation AI and cloud infrastructure. The new platform is designed to enhance hyperscalers’ performance and efficiency, supporting the development of custom AI accelerators, CPUs, and networking solutions. The technology includes high-speed 3D I/O for vertically stacking die inside chiplets, a key innovation to increase bandwidth and reduce connections.

This 2nm platform continues Marvell’s leadership in advanced semiconductor manufacturing, following its 5nm platform in 2020 and 3nm platform in 2022. The company’s approach integrates high-speed SerDes, die-to-die interconnects, custom high-bandwidth memory, and compute fabrics like PCIe Gen 7 to build custom AI and cloud solutions.

Marvell's collaboration with TSMC is critical to developing high-performance, power-efficient silicon solutions for AI infrastructure. TSMC's senior vice president Kevin Zhang emphasized their joint effort in utilizing advanced silicon and packaging technologies to push the boundaries of AI processing capabilities.

(Source: Marvell Technology)