Artificial intelligence (AI) has revolutionized multiple sectors in the last decade. However, the cost of maintaining and operating these advanced systems is staggeringly high. One of the clearest examples is OpenAI’s ChatGPT, whose maintenance represents a daily expense of up to $700,000.
Let’s take a look at the financial situation OpenAI is facing and how Microsoft is developing solutions to reduce the costs associated with the use of AI.
Investment and spending on infrastructure
ChatGPT has been a huge success since its launch, but this success comes at a price. Figures provided by SemiAnalysis suggest that OpenAI pays up to $700,000 a day to maintain the infrastructure necessary to run the system. These figures come from the costs associated with the high-end servers used to run the GPT-4 model, the most advanced version of the AI.
Microsoft, one of the main investors in OpenAI, is working on developing its own AI chip to deal with these rising costs. The chip, called “Athena,” has been in development since 2019 and could cut the cost per chip by a third compared to solutions from Nvidia, the current provider of graphics processing units (GPUs) for AI infrastructure.
What does the Athena chip entail?
The Athena chip is an attempt by Microsoft to create a more efficient and cheaper solution for AI systems. If Athena manages to compete with Nvidia’s offerings, it could mark a major milestone in lowering costs for maintaining advanced AI models like ChatGPT.
Although Microsoft is looking to develop its own AI chip, this does not imply a complete break in the relationship between Microsoft and Nvidia. The two companies have recently agreed to a long-term collaboration in the AI ​​space, suggesting that the Athena chip is just one of many solutions being explored to address the AI ​​infrastructure cost challenge.
Creating and maintaining advanced AI models like ChatGPT pose significant economic and technological challenges. As these models approach a point of diminishing returns due to their sheer size and maintenance costs, it is critical that companies develop new solutions to address these challenges.
Microsoft’s Athena chip is one example of how innovation can address the AI ​​cost problem, but it’s probably not the only solution on the horizon. As AI continues to evolve, we are likely to see more technological developments that seek to reduce the costs associated with running and maintaining these systems.
At the moment what seems to be clear is that for OpenAI 700,000 dollars a day is a price worth paying, in fact it is surely already earning more than that with the subscriptions of users and services that use ChatGPt on a daily basis.
