Story perspectives

Revolutionary In-Memory Computing Boosts AI Performance and Efficiency

10/17/2025

23 2

Summary

Story summary

Advances in in-memory computing use ferroelectric capacitors and spintronic neurons to create non-volatile memory systems that boost energy efficiency and performance for machine learning.
Researchers are pursuing architectures such as crossbar arrays and hybrid spin-CMOS logic gates to enhance computational capabilities.
A 28 nm STT-MRAM macro achieves high performance for AI inference.
Capacitive crossbar arrays demonstrate effective in-memory computing and support AI edge processing.

1 of 1

Story summary

Advances in in-memory computing use ferroelectric capacitors and spintronic neurons to create non-volatile memory systems that boost energy efficiency and performance for machine learning.
Researchers are pursuing architectures such as crossbar arrays and hybrid spin-CMOS logic gates to enhance computational capabilities.
A 28 nm STT-MRAM macro achieves high performance for AI inference.
Capacitive crossbar arrays demonstrate effective in-memory computing and support AI edge processing.