Recently, Haozhe Technology made an official announcement stating that its VVT300 in-memory computing AI SoC chip, specifically designed for edge devices, has been successfully taped out. This milestone significantly accelerates its path towards commercialization. Built upon the company's proprietary "Bai Ze" distributed in-memory computing architecture, the VVT300 chip overcomes the constraints inherent in the traditional von Neumann architecture. Through the integration of SRAM as a storage medium and a distributed computing unit design, the chip enables highly efficient in-memory computing capabilities.

Measured data reveals that the VVT300 chip boasts an impressive AI computing power efficiency exceeding 20TOPS/W. Additionally, it achieves a remarkable reduction in system power consumption by over 40%, coupled with a latency decrease of up to 10 times. These advancements collectively pave the way for the widespread commercialization of in-memory computing technology.