ChatGPT AI will be a boon for Nvidia’s GPUs

artificial intelligence ChatGPT immediately managed to seem capable of answering a large number of very diverse questions and automating a large number of web tasks.

It is also a surprising tool for composing a text or summarizing its essentials, with its good and bad sides since the chatbot has already proven itself to cheat in exams and could become an expert in misinformation and fake news.

Behind these incredible capabilities lie the foundations of an artificial intelligence that requires a lot of hardware to bring it to life. According to TrendForce, ChatGPT will require more than 30,000 Nvidia graphics cards in its final form.

Gigantic revenue potential thanks to AI

This represents a great opportunity for the firm beyond GPUs for desktop and laptop PCs, whose sales are experiencing strong fluctuations (currently falling).

And all the more so since it will be graphics accelerators of the type Nvidia A100 found in supercomputers and costing over $10,000 apiece. After taking advantage of the strong demand for GPUs from the cryptocurrency sector, the company can find a good market there for its professional products and capable of generating $300 million in revenue.

If the Nvidia A100 accelerator uses the Ampère architecture, the firm has already presented its successor, the Nvidia H100 accelerator with Hopper architecture (the equivalent of Ada Lovelace but for professional products), leaving room for future projects (but with a much higher cost of over $30,000).

Question of priorities

The business of Nvidia’s graphics accelerators and artificial intelligence products is growing and should accelerate further with the arrival of new players in the field of conversational artificial intelligence that can be integrated into search engines and applications… and capable of investing hundreds of millions of dollars to equip themselves.

It remains to be seen whether direct competition is not likely to nibble too much market share, while it will also be necessary to monitor the danger of seeing large groups developing their own solutions. Indeed, the Tom’s Hardware site asks the question: can this lead Nvidia to focus its resources on graphics accelerators rather than gaming GPUs?