OpenAI's Costly Battle with ChatGPT
OpenAI's ChatGPT has become incredibly popular, but its maintenance costs are nothing short of staggering. According to a report by The Information, OpenAI spends up to $700,000 a day to keep its infrastructure running smoothly. This cost is primarily due to the expensive servers required by the AI, which are incredibly power-hungry and specialized. Dylan Patel, Chief Analyst at SemiAnalysis, explains that this cost estimate was based on the previous model, GPT-3, and that the newer GPT-4 model would cost even more.
The issue is not unique to ChatGPT, as conversational AIs are incredibly expensive to maintain. A new proprietary AI chip called "Athena," which is being developed by Microsoft, could potentially help reduce the cost per chip by a third when compared with Nvidia's current offerings. This could result in significant savings for OpenAI, which has invested billions of dollars in the Microsoft project.
The potential savings from Athena couldn't come soon enough for OpenAI. The newer GPT-4 model may already be approaching the limit of practical scalability, according to OpenAI's own analysis. While larger models have generally meant more power and greater capabilities for an AI, they will drive up costs in the long run, as Patel has pointed out.
Despite the hefty costs, OpenAI is probably not hurting for money, thanks to the success of ChatGPT. It remains to be seen how Athena will impact the future of conversational AIs, and whether it will provide the cost savings that are needed to keep these technologies affordable and accessible to all.
