MaxLinear to Showcase Panther Storage Accelerator for AI Workloads at FMS 2026

MaxLinear will showcase its Panther Storage Accelerator platform at the Future of Memory and Storage conference, taking place August 4-6 in Santa Clara, California.

 

The technology addresses a challenge emerging as AI shifts from single-model deployments toward agentic AI, where systems run many autonomous AI agents at once. As infrastructure needs shift from optimizing individual model performance to supporting agent scalability, AI clusters increasingly need to handle millions of concurrent agents, which drives sharp growth in memory capacity and bandwidth requirements. As the number of agents grows, high-bandwidth memory (HBM) and DDR memory resources come under more pressure, which reduces how efficiently GPUs and CPUs can operate. Panther addresses this through hardware-accelerated compression, achieving up to a 3x compression ratio on AI data. That efficiency gain can translate into up to 2.2 times more effective capacity for KV cache, the temporary memory used to store intermediate AI computation results, which in turn lets systems support up to 2.2 times more AI agents using the same existing hardware.

 

Vikas Choudhary, senior vice president and general manager of connectivity and storage at MaxLinear, said that as AI adoption accelerates, organizations are rethinking how they store, move, and manage data at s...