Better AI Inference Stock to Own: Nvidia or Cerebras?

While large language model (LLM) training dominated the first phase of artificial intelligence (AI), inference is eventually expected to become the much larger market.While LLM training is compute-heavy and more technically challenging, inference tends to be memory-centric and needs to be more cost-efficient given that it's an ongoing process. Traditionally, graphics processing units (GPUs) and other AI accelerators are packaged with high-bandwidth memory (HBM) to help optimize their performance in this area.However, Nvidia (NASDAQ: NVDA), through its recent "acquisition" of Groq, and Cerebras Systems (NASDAQ: CBRS) are now looking toward on-chip SRAM (static random-access memory) to speed up AI workloads for inference. This is a new approach, and both companies are using SRAM in a much different way. While using SRAM can dramatically increase inference speeds, it is physically bulky, which creates some trade-offs between chip size, memory capacity, and the data center infrastructure required to power and cool the chips. Continue readingWeiter zum vollständigen Artikel bei MotleyFool Weiter zum vollständigen Artikel bei MotleyFool