NewsBite analysis · AI, Hardware
Toms Hardware
Qualcomm reveals HBC near-memory AI architecture, AI250 and AI350 accelerators published 25 June 26
Qualcomm has officially revealed its HBC near-memory architecture designed for AI applications, introducing the AI250 and AI350 accelerators. The new technology is engineered to provide six times the bandwidth efficiency per watt compared to traditional HBM, offering 200 times more capacity than conventional on-chip SRAM. This development positions Qualcomm at the forefront of AI hardware innovation, enabling faster and more efficient processing for AI workloads.
Act nowConfidence: HighShort term horizon
Published 25 Jun 2026, 10:45 UTC1 min read
Visit original article ↗Lead image — Toms Hardware
Summary
TL;DR
- Introduction of HBC near-memory AI architecture by Qualcomm
- Launch of AI250 and AI350 accelerators
- Achieves 6x higher bandwidth per watt than HBM
- Offers 200x more capacity compared to on-chip SRAM
- Enhances efficiency for AI workloads
Interesting facts
- The AI250 and AI350 are aimed specifically at addressing the growing demands of AI applications.
- The architecture represents a significant shift from traditional memory hierarchies to near-memory processing.
Coverage gaps