Qualcomm

AI inference workloads place different demands on memory. Explore how HBC, HBM, and SRAM compare across common inference scenarios and the tradeoffs architects may consider when designing AI infrastructure.

As AI systems scale, memory architecture plays an increasingly important role. Explore the characteristics, tradeoffs, and design considerations behind HBC, HBM, and SRAM.

As AI workloads grow, memory and data movement become increasingly important design considerations. Explore how near-memory computing may help address these challenges.

Heading to AI Infra Summit? Stop by Booth #206 to see how Qualcomm Dragonfly is powering the next generation of AI infrastructure. See you in Santa Clara.

Everyone's racing for peak compute, but AI inference is often limited by memory. When compute has to wait on data, efficiency suffers. Qualcomm High Bandwidth Compute (HBC) takes a different approach by moving compute closer to the data, helping reduce bottlenecks and improve efficiency while delivering an estimated 6x increase in bandwidth per watt vs. HBM and 200x increase in capacity per watt vs. SRAM. Read more: https://bit.ly/44Tlzt7

Today we announced a multi-generation collaboration with Amazon Web Services (AWS) to help build the next generation of AI data center infrastructure. Together we are working across customized silicon and AI inference at scale. This is a significant validation of our data center strategy, with a potential revenue opportunity of up to $60 billion over time. Amazon choosing Qualcomm is a strong endorsement of our technology, our roadmap and our ability to deliver data center solutions at scale. AI infrastructure is entering a new phase where efficiency matters as much as performance. For decades we have delivered industry-leading compute performance within tight power envelopes, and we are now applying that expertise to the data center. Read the announcement: https://lnkd.in/e27N8i4f Read the 8-K: https://lnkd.in/ej5tCBiX