From Magnetic Storage to Magnetic Computing: CRAM for the AI Era and Beyond

The HDD industry has spent decades advancing our ability to store, move, and reliably access enormous amounts of digital information. Artificial intelligence is now creating a new challenge: the energy and performance bottleneck is increasingly not the computation itself, but the movement of data between storage, memory, and processors. This emerging "memory wall" is becoming one of the fundamental limitations of AI systems. This talk will discuss how advances rooted in magnetic materials, spintronics, and MRAM can enable a new computing architecture called Computational Random Access Memory (CRAM). Rather than continuously moving data from memory to a separate processor, CRAM transforms memory itself into a programmable digital computing fabric, allowing logic and arithmetic operations to be performed directly where the data resides. CRAM retains random-access memory functionality while adding reconfigurable digital computation, creating a fundamentally different approach to reducing data movement. I will review the evolution from magnetic recording and MRAM toward compute-in-memory, the device and architectural principles behind CRAM, and the path from laboratory demonstrations to foundry-compatible silicon. Particular attention will be given to opportunities relevant to the HDD and storage industry, including intelligent HDD/SSD controllers, near-storage computing, data processing close to storage, and increasingly AI-driven storage systems. Finally, I will discuss how the same technology can scale from low-power edge and storage applications to large-scale AI inference, and why expertise developed by the HDD industry in magnetic materials, devices, reliability, signal processing, and high-volume manufacturing could play an important role in the next transition--from magnetic storage to magnetic computing.

「AI for Science」時代の次世代計算基盤：富岳NEXTの概要と階層化ストレージ戦略

「富岳NEXT」は、2030年頃の運用開始を目指し、AIとシミュレーションの融合（AI for Science）を核に設計されています。 ノード構成はCPUとGPUのヘテロ構成を採用し、ストレージは高速なSSDによる第1階層と、大容量の第2階層からなる2階層アーキテクチャを採用し、ゼッタスケール時代の膨大なデータ需要に対し、コストと容量の最適化を図る戦略についての現状を説明。

後日お知らせいたします

IBMが考える量子コンピューターの未来

量子コンピューターの現状と、今後どのように進化していくのかをIBMの量子コンピューターを例に説明します。また、どのようなユースケースがあり、またEcoシステムがどのようになっているかも説明致します。

閉会挨拶