Numem has announced its AI Memory Engine, a fully synthesizable memory subsystem IP designed to address traditional memory performance limitations for artificial intelligence (AI) workloads in data centers and edge environments. According to Numem, the AI Memory Engine significantly improves power efficiency, performance, and endurance for emerging memory technologies, including Magnetoresistive RAM (MRAM), Resistive RAM (RRAM), Phase Change RAM (PCRAM), and Flash memory.
Specifically optimized for embedded applications, the AI Memory Engine integrates with Numem’s patented MRAM architecture, enabling SRAM-level performance while achieving densities up to 2.5 times higher and standby power consumption 100 times lower than traditional memory solutions. Numem says its MRAM supports die densities up to 1 GB, providing scalability and integration flexibility in standard foundry environments.
“Every week, I hear the same thing from customers: their memory can’t keep up,” said Max Simmons, CEO of Numem. “Not enough performance, not enough density, and way too much power consumption. AI workloads are pushing existing architectures to the limit – especially in areas like automotive, where in-vehicle infotainment (IVI) systems now rely on multiple cameras and real-time AI. DRAM just isn’t cutting it. It’s too slow to boot, consumes too much power, and simply can’t meet the performance demands of modern systems.”
Numem claims the AI Memory Engine achieves between 30% and 50% power savings compared to existing high-bandwidth memory technologies, translating into reduced operating costs and lower energy usage. Its flexible power management architecture supports multiple power modes, enabling seamless integration into data center and edge infrastructure without extensive hardware redesign.
The company’s approach leverages a fabless model, offering foundry-ready IP and silicon to simplify deployment for a variety of AI-driven use cases, including automotive and custom system-on-chip (SoC) designs. According to Numem, this positions the company strategically within the emerging MRAM market, projected by Polaris Market Research to reach a total addressable market of $25.1 billion by 2030.
Source: Numem
