In-memory computing, which seeks to reduce power consumption by processing data where it's stored, is seeing a boost in interest fueled by the AI energy crisis
and even inspiring some geopolitical maneuvering https://www.bloomberg.com/... via @BW
Context & Ripple Effects
This sits in a broader push to cut the energy cost of moving data, not just performing computation. IBM had already demonstrated that direction with its NorthPole prototype designed to avoid frequent external-memory access.
Later coverage shows why memory-centric efficiency has become strategically relevant: SoftBank and Intel formed Saimemory to pursue stacked DRAM for AI targeting roughly half the power of current HBM, while AI demand has also tightened the memory supply picture.
First-order effects
- In-memory computing receives greater attention from AI hardware developers and governments because it targets a major source of energy use: data movement between processing and storage.
- The geopolitical dimension raises the stakes for control of the architectures and memory technologies that could lower AI infrastructure power needs.
Second-order effects
- Chip and memory suppliers face added pressure to differentiate on energy efficiency alongside performance and capacity, particularly as AI systems become more memory-intensive.
- Alternative low-energy architectures gain a clearer commercial rationale, reflected in later funding for spatial-dataflow AI chips designed to minimize energy consumption.
Third-order effects
- If these approaches mature, memory placement, packaging, and data-movement efficiency could become durable AI-infrastructure bottlenecks rather than secondary chip-design details.
- The competition for AI advantage may increasingly extend from access to compute into control of power-efficient memory and system architectures, though broad adoption still depends on production-scale execution.
The trend: AI’s energy constraint is shifting hardware competition toward architectures that reduce data movement and make memory efficiency a strategic capability.