Nvidia Plans LPU-Based SRAM Inference Chip at GTC, Options Strategy Suggested
Morgan Stanley projects Nvidia will unveil an inference chip at its GPU Technology Conference featuring LPU architecture with large on-chip SRAM to boost sequential inference speed and alleviate HBM supply-chain bottlenecks. J.P. Morgan recommends options strategies to capitalize on Nvidia’s recent rebound following its stock sell-off last week.
1. Upcoming LPU-Based Inference Chip
Morgan Stanley anticipates that Nvidia will introduce a new inference chip at the upcoming GPU Technology Conference built around a Language Processing Unit architecture. This design emphasizes large on-chip SRAM for hot-path execution to accelerate sequential inference and bypass current supply-chain constraints in HBM and advanced packaging.
2. Market Strategies and Stock Rebound
Following a recent sell-off, Nvidia shares have rebounded, prompting J.P. Morgan analysts to outline options strategies aimed at capturing upside potential. These derivative plays are structured to leverage anticipated demand for Nvidia’s forthcoming inference chip and the broader momentum in AI workloads.