Being the trendsetter, ChatGPT, the natural language processing tool driven by AI, has been used worldwide for getting a response about simple to complex topics and tasks. As a result, people worldwide started using OpenAI’s AI chatbot for content creation, translation, coding, debugging and many more. 

The exceptional ability to understand and respond to a wide range of language inputs, learn and improve over time, answer complex queries, etc., has made the AI chatbot dearest to people around the globe. The latest version GPT4 has also caused an enormous wave in the industry with its multimodality. 

According to the Business Insider, the ChatGPT that can be used to do tremendous tasks for you could cost over $700,000 per day due to its costly tech infrastructure on which the AI runs, said Dylan Patel, the chief analyst of SemiAnalysis, a semiconductor research firm. 

The cost surge is mainly due to the requirement of a humungous amount of computing power to calculate responses according to each user’s prompts. “Most of this cost is based around the expensive services they require”, Patel added. He mentioned that his current estimate is based on OpenAI’s GPT-3 model, and GPT-4 is expected to be even more expensive. The costs to infer ChatGPT exceed the training costs every week”, said another analyst at SemiAnalysis. 

Providing a solution for the cost surge of generative AI models, Microsoft is reportedly developing an AI chip called “Athena”. The project, which was initially started in 2019, comes to the forefront after the tech behemoth’s $1 billion deal with OpenAI that asked OpenAI to run its models exclusively on their Azure cloud services. 

According to the report, the idea behind the chip was two-folded. As Microsoft felt like falling behind Google and Amazon in its efforts to build their own chips, they started looking for cheaper alternatives. The AI models of Microsoft were run on NVIDIA’s chips called graphics processing units, and they decided to develop less costly chips. 

According to The Information, more than 300 employees are reportedly working on the chip. The chip is expected to release for internal use by Microsoft and OpenAI next year.