HBC vs HBM vs SRAM: Part 2 – A balanced comparison from an architect’s view - Qualcomm

This is the second part of a semiconductor memory technology comparison analysis, presenting an architect's perspective on evaluating HBC (High Bandwidth Cache), HBM (High Bandwidth Memory), and SRAM ... Read original →

From an architectural standpoint, HBC demonstrates a clear edge in AI inference applications, leveraging near-data computing to mitigate trade-offs between bandwidth, power, and cost that traditionally plague HBM and SRAM. This shift is prompting upstream memory vendors to realign their R&D strategies toward high-bandwidth caching solutions. However, HBM remains dominant in high-performance computing, especially in data center and AI training workloads, where its bandwidth advantage is irreplaceable. Geopolitical dynamics, particularly the ongoing tech decoupling between the US and China, are intensifying supply chain risks, compelling firms to reassess autonomy and resilience. Competitors are likely to pursue hybrid architectures combining SRAM and HBM to address diverse use cases. Over the next 12 to 24 months, the demand for low-power, high-density memory in edge AI chips will accelerate HBC innovation, while HBM and SRAM may stabilize in niche markets.