OpenAI has happened without a shadow of a doubt one of the most important technologies in the artificial intelligence ecosystem. In recent years, it has already gained quite a lot of popularity thanks to its GPT language model and DALL-E image creation model. However, if there is a service that made it a trendy technology, it was ChatGPT. Since its launch at the end of last year, its relevance (both as a service and as a company) has not stopped growing.

During this year we attended the presentation of GPT-4 in mid-March and more recently DALL-E 3 in mid-last month. In both cases, Microsoft was quick to confirm that these new models had been added to Bing (which, remember, is based on OpenAI technology, albeit with a few tweaks from Microsoft). We also learned over a week ago that Spotify will be using Whisper (which is also from OpenAI) to dub its podcasts… and that’s just a small part of the huge demand for the company’s services.

So, without a doubt, we can confirm that, in general, the company’s services are extremely successful, which is reflected in the already huge and growing infrastructure needs provide such services. But there is an important problem here, which is that the enormous demand for specialized chips for AI tasks is not fully met by the current production volume, and that the main supplier, NVIDIA, has put a lot of effort into making its production volume as high as possible.

So it seems that the tech company has taken a drastic decision in this regard because according to Reuters OpenAI has been studying the possibility of creating its own AI chips for some time, among a set of possible solutions to fully cover your integrated demand. According to this information, no firm decision has yet been made in this direction, but the company’s executives have evaluated (and are still doing so today) the possibility of acquiring a company dedicated to the production of integrated devices, which would be devoted exclusively to its own chips.

This option, although really challenging, would be much more economical (both in time and money) than starting your own manufacturing plant from scratch, so we understand that this option, if not excluded, is at the bottom of the list. Furthermore, your intention is probably, at least in the short and medium term, to use your own integrated but also stay on NVIDIA’s customer listgiven the great reliability and performance demonstrated by its A100 and H100.