ChatGPT Chatbot has become popular in the upstream chip field from the downstream AI application. As a natural language AI application, ChatGPT has a huge demand for computing power, and its development cannot be separated from the support of high-performance computing chips, such as GPU.

In addition, memory technology is also important in AI computing, such as HBM technology, which is moving from the background to the front thanks to ChatGPT's popularity.

High Bandwidth Memory (HBM) is a kind of high value-added DRAM that can achieve high bandwidth. Compared with ordinary DRAM, HBM achieves higher bandwidth by stacking. Bandwidth refers to the amount of data that can be transmitted in a specific unit of time. HBMS are mainly applied in high-performance computing scenarios, such as supercomputers, AI accelerators, and high-performance servers.

According to Korean media, HBM orders from Samsung and SK Hynix have increased rapidly since the start of 2023, and the price of HBM3 DRAM has increased five times recently.

HBMS can improve machine learning and computing performance when working with cpus and Gpus. The ChatGPT boom is already benefiting GPU vendors like Nvidia -- ChatGPT uses more than 10,000 Nvidia "A100" Gpus to learn massive document data. The HBM can be installed in the GPU to improve the ChatGTP service experience through high-speed data processing. The Nvidia "A100" is equipped with up to 80GB HBM2 memory.

In addition, media reports say Nvidia has asked storage vendors to produce the latest HBM3 RAM to pair with Nvidia's A100 GPU for ChatGPT.