AI today runs on massive, power-hungry GPUs — but a new kind of memory could change everything. This video reveals how resistive memory performs computation directly inside memory, unlocking a faster, cheaper, and far more efficient future for AI. References: [1] O. Mutlu, [2] IBM Research, [3] O. Mutlu, S. Ghose, J. G´. omez-Luna, R. Ausavarungnirun, M. Sadrosadati, G. F. Oliveira, [4] M. Jiang, Yichun Xu, Zefan Li, Can Li, Current Opinion in Solid Stat and Materials Science Volume 37, pages 101226 (2025), [5] W. Wan et. al. Nature volume 608, pages 504–512 (2022) [6] M. L. Gallow et. al. Nature Electronics volume 6, pages 680–693 (2023) [7] Q. Huo et. al. Nature Electronics volume 5, pages 469–477 (2022) [8] H. Jiang, W. Li, S. Huang and S. Yu, 2022 IEEE Symposium on VLSI Technology and Circuits (VLSI Technology and Circuits), Honolulu, HI, USA, pp. 266-267 (2022), doi: [9] S. Ambrosio et. al. Nature volume 620, pages 768–775 (2023) [10] Q. Liu et al., 2020 IEEE International Solid-State Circuits Conference - (ISSCC), San Francisco, CA, USA, pp. 500-502 (2020), doi: [11] W. -H. Huang et al., 2023 IEEE International Solid-State Circuits Conference (ISSCC), San Francisco, CA, USA, pp. 15-17 (2023), doi: [12] T. -H. Wen et al., 2024 IEEE International Solid-State Circuits Conference (ISSCC), San Francisco, CA, USA, pp. 580-582 (2024), doi: [13] V. K. Sahu, A. K Das, RS Ajimsha, P. Misra, Journal of Physics D: Applied Physics, Volume 51, pages 215101 (2018). Credits: All the animations are created using the open source library Manim, maintained by Manim Community. Chapters: 0:00 Intro 0:38 Vector Matrix Multiplication in AI 1:16 VMM in GPU/CPU 2:26 The Memory Wall 3:36 Resistive Memory 4:32 Crossbar Structure 4:56 VMM in Resistive Memories 5:49 Memory That Computes 6:33 The Limitations of Resistive Memories 7:11 Outro