Marvell introduces industry’s first 2nm custom SRAM for AI and cloud data centers

Marvell Technology launched the industry’s first 2nm custom Static Random Access Memory (SRAM) for custom XPUs and devices used in cloud data centers and AI clusters. The Marvell custom SRAM, built on a 2nm process with custom circuitry and software, provides up to 6 gigabits of high-speed memory at frequencies up to 3.75 GHz, while reducing on-chip memory standby power by up to 66% and die area by up to 15% compared to standard SRAM at equivalent density.
The SRAM allows chip designers to reclaim up to 15% of a 2nm design’s total area, enabling the addition of compute cores, expanded memory, reduced device size, or optimization of performance, power, and total cost of ownership. This follows Marvell’s earlier developments, including CXL technology for integrating terabytes of memory and compute capacity into custom silicon for cloud servers and custom high-bandwidth memory (HBM) technology, which increases memory capacity by up to 33% while reducing space and power requirements.
The custom SRAM is part of Marvell’s custom technology platform, which includes intellectual property such as serializer/deserializers (SerDes), die-to-die interconnects for 2D and 3D devices, silicon photonics, co-packaged copper, custom HBM, system-on-chip fabrics, optical IO, and compute fabric interfaces like PCIe Gen 7. This platform aims to address customer-specific infrastructure needs through tailored semiconductor designs and manufacturing processes.