ZeroPoint Technologies Introduces AI-MX for Enhanced Memory Optimization

ZeroPoint Technologies has unveiled a new hardware-accelerated memory optimization product called AI-MX, which promises to increase the addressable memory of foundational models by 50% announced in a press release. This innovative solution is designed to enhance the performance of enterprise and hyperscale datacenters by enabling a 1.5 times increase in memory capacity, bandwidth, and tokens served per second for applications relying on large foundational models.

The AI-MX product is set to be delivered to initial customers and partners in the second half of 2025. It operates with low nanosecond latencies, making it significantly faster than traditional compression algorithms. The product is compatible with various memory types, including HBM, LPDDR, GDDR, and DDR, ensuring broad applicability across AI acceleration use cases.

ZeroPoint Technologies' CEO, Klas Moreau, highlighted the potential cost savings for companies operating large-scale datacenters, as AI-MX addresses the growing demand for memory capacity, power, and bandwidth. The company aims to further enhance the capacity and performance of AI-MX in future iterations, addressing the critical needs of today's hyperscale and enterprise data center operators.