PIMIC debuts game-changing silicon to drive AI efficiency at the edge
PIMIC, an AI semiconductor startup, launched to develop advanced AI inferencing silicon for voice-activated edge devices.
PIMIC aims to make AI more accessible for everyday applications. The company’s Jetstreme Processing-in-Memory (PIM) architecture enhances AI computing efficiency with low power and compact design.
Jetstreme technology offers a 10x advantage in power, performance, and area over existing near-memory solutions. PIMIC’s solutions support dynamic AI model updates and low-energy model reconfiguration for battery-operated devices.
“We founded PIMIC to empower AI devices at the edge with the kind of efficiency and performance that has, until now, only been possible in large, power-intensive environments,” says Subi Krishnamurthy, CEO/CTO of PIMIC. “We recognized a need for AI to be more accessible and practical, not just for cutting-edge technology users, but for everyday applications that can enhance lives.”
The AI industry is shifting towards edge computing, with IDC forecasting $41 billion in edge AI processor revenue by 2028.
“Endpoint devices are increasingly adding on-device AI capabilities that require very small die size and very low power consumption,” says Phil Solis, Research Director at IDC. “PIMIC’s solutions uniquely address this and are small enough to fit inside MEMS sensor devices, making them easy to integrate into existing solutions, and are also scalable enough to run AI models up to billions of parameters while eliminating the memory interface bottleneck.”
PIMIC’s technology enables on-device AI capabilities, reducing reliance on cloud infrastructure and enhancing integration into existing solutions. Products based on Jetstreme technology are expected to be available in early 2026, with PIMIC showcasing at CES 2025.
Article Topics
AI | edge devices | PIMIC | semiconductor
Comments