TECH

ChatGPT could be a good boost for Nvidia and AMD

Trendforce says in its report that Nvidia (and not only Nvidia) could see considerable success thanks to increased interest in artificial intelligence. The successes of systems such as ChatGPT could contribute to increased demand for its accelerators.

Chatbots have been a big topic in recent weeks, but it's not just about the software itself. Powerful hardware also plays a significant role here, which, according to Trendforce, could be a significant boost for companies involved in the development and production of hardware for AI applications. The company Nvidia, whose latest financial results were not the best, was mentioned above all. According to Trendforce, the GPT system on which ChatGPT is built grew from 120 million to 180 billion training parameters from 2018 to 2020. Trendforce estimates that around 20,000 GPUs were needed for this three years ago.

Today, although card performance is increasing, an increased number of accelerators will still be needed, so it is estimated that GPT training now requires over 30 Nvidia A100 cards. And it is the increased interest in AI applications that pushes up interest in the hardware on which this AI will be trained, which could greatly help Nvidia. In addition, according to the company, it could mean a boost for competitor AMD with the MI200 and MI300 accelerators, and the companies that produce these chips, which is primarily TSMC, should also benefit from it. But good times could also await other companies that develop AI chips for various applications, such as GUC, AIchip, Faraday Technology and eMemory.

by: Milan Šurkala