Recently a Reuters article claimed that OpenAI, which is a company behind ChatGPT, is for the first time eschewing Nvidia’s AI hardware and rather, the AI behemoth has begun leasing TPU chips of Google to power its ChatGPT and other products.
For OpenAI, one of the largest buyers of Nvidia GPUs for AI training and inference computing—the process of using AI models to solve practical problems—this is a significant shift. In order to satisfy its quickly increasing computer needs, the business is now also using Google’s infrastructure.
An unexpected alliance between two bitter competitors in the AI sector was hinted at earlier this month when Reuters reported that OpenAI intended to include Google Cloud services into its operations.
The action is a component of Google’s larger initiative to open up its specially created tensor processing units (TPUs) to external customers. TPUs, which were previously mostly utilized for internal projects, are now assisting Google in luring a diverse clientele, which includes tech behemoth Apple and AI start-ups Anthropic and Safe Superintelligence, both of which were established by former OpenAI officials.
Furthermore, OpenAI recently shifted away from using data centers controlled by its largest investor, Microsoft, alongside its decision to start renting TPU chips of Google.
This could make TPUs a more affordable option than Nvidia’s GPUs, according to The Information, which broke the story first. Additionally, according to the study, OpenAI hopes to lower the high expenses related to inference by utilizing these Google chips.
Nevertheless, it has been stated that Google is not providing OpenAI with its most sophisticated TPUs, even with this recent partnership. According to The Information, which cited a Google Cloud employee, the tech giant is preventing a direct rival in the escalating AI competition from using its top-tier CPUs.
Loading ...
