Jibang Consulting: 30,000 A100 graphics cards are needed for ChatGPT to run, and Nvidia can earn US $300 million from OpenAI.

CTOnews.com March 7 news, market research firm TrendForce Jibang Consulting released the latest report that if calculated based on the processing capacity of the Nvidia A100 graphics card, running ChatGPT will need to use 30000 Nvidia GPU.

TrendForce, an investigative agency, pointed out in a report that it is estimated that 20000 graphics cards are needed for ChatGPT to process training data. As OpenAI continues to deploy ChatGPT and the company's generative GPT model commercially, more than 30000 graphics cards may be required.

Nvidia A100 graphics cards currently sell for between $10000 and $15000, depending on how many graphics cards OpenAI needs and whether Nvidia will offer order discounts to AI. TrendForce estimates that Nvidia could earn more than $300m (CTOnews.com Note: currently about 2.082 billion yuan).

Nvidia also sells the A100 as part of the DGX A100 system, which has eight accelerators and sells for up to $199000. Given the scale of OpenAI's operations, the company may buy A100 separately and stack them into clusters.