According to Reuters reported today, sources revealed that OpenAI is exploring the manufacture of their own AI chip, and has begun to assess the “potential” acquisition target.
People familiar with the situation said that since at least last year, OpenAI has been discussing various options to solve the problem of its AI chip shortage. These options include developing its own AI chips, working more closely with chipmakers such as NVIDIA, and diversifying suppliers.
However, according to recent “internal discussions” by Reuters, OpenAI has not yet decided to move forward. At the same time, OpenAI also declined to comment on the matter.
Prior to this, OpenAI CEO Sam Altman in the first half of the year publicly complained about the “acute shortage” of GPUs on which it relies heavily, which has also led to the reliability and speed of its APIs being criticized by OpenAI customers. In Altman’s view, OpenAI faces a shortage of advanced processors and high hardware operating costs.
According to Bernstein analyst Stacy Rasgon’s analysis, at 4 cents per query, if ChatGPT’s search volume reaches 1/10 of Google’s, OpenAI will initially need GPUs worth a total of about 48.1 billion U.S. dollars (currently about RMB 351.611 billion), and about 16 billion U.S. dollars worth of chips per year (currently about 116.96 billion yuan) to maintain operation. The OpenGPUs are worth about $48.1 billion (currently about RMB 351.61 billion), and the chips worth about $16 billion (currently about 116.96 billion yuan) are needed every year to maintain operation.
Industry veterans analyze that OpenAI’s self-developed chips will be a “major strategic move” and will cost hundreds of millions of dollars per year. But even that is no guarantee of success.