OpenAI appears to be considering making its own chips
In the relatively near future in the semiconductor market, we could see another, new player as the company OpenAI is reportedly considering making its own chips. The creator of the popular ChatGPT bot is currently investigating this possibility and has reportedly already evaluated a possible acquisition of the chip company, Reuters writes.
The CEO of this startup, Sam Altman, previously blamed the GPU shortage for user complaints about the speed and reliability of the company’s API, which is why he reportedly made the acquisition of AI chips a priority.
In this way, OpenAI could solve the lack of GPUs it is struggling with, and by using its own chips, it could also reduce the costs associated with running its products. However, the company is still in the consideration phase and has not taken an official step in this direction.
Each ChatGPT query reportedly costs the company about four cents, and the bot’s services reached 100 million monthly users in the first two months. This further means millions of queries per day, according to Bernstein Research analysis. If ChatGPT queries reach a tenth of what Google gets, it will initially need $48.1 billion worth of GPUs, and in the future OpenAI would spend $16 billion a year on chips alone.
At the moment, Nvidia is the company that dominates the chip market for AI applications, and other technology companies use thousands of its GPU products for various services. For example, Microsoft’s supercomputer, which used OpenAI to develop its technology, uses 10,000 Nvidia GPUs.
This is exactly why other companies, bigger players in the technology industry, have decided to start developing their own chips. The aforementioned Microsoft, which of all companies invested the most in the OpenAI startup, has been working on its own AI chip since 2019. That product is codenamed Athena, and ChatGPT’s creator is reportedly testing its technology.
Still, even if one of the most well-known startups, OpenAI, decides to go ahead with its chip plans, it could be years before the company becomes independent from commercial providers like Nvidia and AMD. Therefore, in order to use its own chips to power its products, this company, in addition to a large amount of money, will also need more time.