OpenAIa San Francisco company known for creating the ChatGPT chatbot, would be evaluating the possibility of manufacturing its own chips independently for artificial intelligence, designed specifically for its needs. This was reported by Reuters, citing sources close to the company.
After all, OpenAI, which has Microsoft among its main financiers, finds itself facing the scarcity and high cost of the chips needed to develop its technologies of AI, in particular its generative applications. The CEO Sam Altman has in fact declared that it has a growing need to acquire more and more graphics processing units (GPUs). a market dominated by Californian Nvidia. The latter, we remember, had a surge in the stock market following the boom in AI technologies.
Why does OpenAI want to produce its own chips?
OpenAI currently uses a supercomputer built by Microsoft, which it employs 10 thousand Nvidia GPUs. However, the biggest cost to society is ChatGPT. In fact, every time the chatbot receives a request from users, it consumes around 4 cents in GPU. According to an analysis by Bernstein Stacy Rasgonif requests to ChatGPT reach even just a tenth of those made to Google, OpenAI would need about $48.1 billion worth of GPUs and $16 billion worth of chips per year to keep your chatbot operational.
For this reason, OpenAI is reportedly considering the idea of acquiring a company that produces chips for artificial intelligence, or starting to produce them independently. This way, the company could customize its chips to its specific needs and reduce its dependence on commercial suppliers. This strategy has already been adopted in the past by other technology giants such as Alphabet e Amazon.
However, according to experts interviewed by Reuters, This path is neither easy nor cheap. Developing your own chips for artificial intelligence would indeed require investments of hundreds of millions of dollars per year. It also wouldn’t completely eliminate the reliance on Nvidia GPUs.