CEO's Column
Search
More
AI Infrastructure

Marvell Debuts World's First 2nm Custom SRAM for AI

ByNeelima N M
2025-06-18.2 months ago
Marvell Debuts World's First 2nm Custom SRAM for AI
Marvell unveils the first 2nm custom SRAM, delivering faster speeds and reduced power for advanced AI and cloud infrastructure.

Marvell Technology, a leader in semiconductor solutions for data infrastructure, has launched the world’s first 2nm custom Static Random Access Memory (SRAM), marking a significant advancement in memory technology.

This new custom SRAM is designed to enhance the performance of custom XPUs (accelerated processing units) and devices that power cloud data centers and AI clusters. By combining Marvell’s custom circuitry and software with cutting-edge 2nm process technology, the custom SRAM delivers up to 6 gigabits of high-speed memory while significantly reducing power consumption and die area at comparable densities.

High-Speed Memory with Energy Efficiency

Marvell’s custom SRAM promises to deliver the highest bandwidth per square millimeter in the industry, making it an essential component for accelerating memory hierarchy performance in cloud infrastructure. It enables chip designers to recover up to 15% of the total area in a 2nm design.

This recovered silicon real estate can be used for additional compute cores, expanded memory, or to optimize device size and cost. Additionally, Marvell custom SRAM consumes up to 66% less power than traditional on-chip SRAM at equivalent densities, while operating at speeds up to 3.75 GHz.

Pioneering the Future of AI Infrastructure

Will Chu, Senior Vice President of Custom Cloud Solutions at Marvell, “Custom is the future of AI infrastructure. The methodologies and technologies used by hyperscalers today to develop cutting-edge custom XPUs will percolate to more customers, more classes of devices, and more applications.”

He added, “We look forward to collaborating with our partners and customers creating the leading technology portfolio for the custom era.”

Memory remains one of the key challenges in AI clusters and cloud systems. These systems require as much memory as possible, at the fastest speeds, to support the growing demand for processing power

Also read: Marvell Launches Custom UALink Scale-Up Solution for AI Infrastructure

Extending the Marvell Custom Technology Platform

The introduction of custom SRAM is a part of Marvell’s broader custom technology platform designed for the post-Moore’s Law era. Over the past 50 years, semiconductor performance improvements were mainly driven by transistor scaling, but as the cost and challenges of further shrinking transistors rise, the focus is shifting toward custom silicon solutions.

Marvell’s custom platform strategy leverages system and semiconductor design expertise, advanced manufacturing processes, and a comprehensive portfolio of semiconductor platforms and IP to help customers transform infrastructure performance, efficiency, and value.

Related Topics

AI infrastructure

Subscribe to NG.ai News for real-time AI insights, personalized updates, and expert analysis—delivered straight to your inbox.