OpenAI starts using Google AI chips to cut costs and rely less on Microsoft, Nvidia: Report

India Today

2 days ago

OpenAI has reportedly begun using Google's artificial intelligence chips in a move aimed at cutting costs and reducing its dependence on long-time partners Microsoft and Nvidia, according to a report by Reuters. Reportedly, OpenAI has recently started renting Google's tensor processing units (TPUs), using them to support its AI applications, including ChatGPT. The TPUs are accessed through Google Cloud, and the move is part of OpenAI's broader effort to manage the soaring cost of running large AI models. The development was first reported by The Information, which also noted that OpenAI is hoping the TPUs will help bring down the cost of inference, a critical part of maintaining and scaling its AI products.advertisementThe decision to move to Google chips marks the first meaningful use of non-Nvidia chips by OpenAI, a major shift for the company that is among the largest buyers of Nvidia's powerful GPUs. These chips have traditionally powered both the training of its AI models and inference computing, which is basically the process of using trained models to generate outputs such as text responses in ChatGPT.This new arrangement also highlights a surprising collaboration between two direct rivals in the AI space. Google, which has developed its own AI models like Gemini, typically reserves its TPUs for internal projects. However, it has recently started opening up access to external clients, including big names such as Apple and AI startups like Anthropic and Safe Superintelligence, both of which were founded by former OpenAI leaders.advertisement
Despite the new partnership, Google has reportedly chosen not to offer OpenAI access to its most advanced TPUs. A Google Cloud employee reportedly told The Information that the company is limiting OpenAI's use of its highest-end chips, likely to maintain a competitive edge. The shift to Google's hardware comes at a time when OpenAI's relationship with Microsoft is showing signs of strain. Microsoft has been OpenAI's biggest investor. The company put in $1 billion in 2019 and integrated OpenAI's technology into its products such as Microsoft 365 and GitHub Copilot.Meanwhile, The Wall Street Journal recently reported that OpenAI executives had discussed whether to accuse Microsoft of anticompetitive conduct and potentially seek a regulatory review of their contract. Though both companies issued a joint statement expressing optimism about their continued collaboration, behind-the-scenes negotiations over financial terms and operational control reportedly remain unresolved. One of the biggest sticking points is apparently around revenue sharing. OpenAI is reportedly planning to reduce the share of revenue it pays Microsoft, from the current 20 percent to just 10 percent by 2030.Additionally, the two firms have not yet agreed on how to structure Microsoft's equity stake in a reconfigured unit of OpenAI, nor on the rights to future profits.OpenAI has also asked to revisit an exclusivity clause that gives Microsoft sole rights to host its models on the Azure cloud. With its growing computing needs and new reliance on Google Cloud's TPUs, the company is looking to create more flexibility in its cloud infrastructure strategy.Another reported source of friction is OpenAI's $3 billion acquisition of Windsurf, a startup focused on AI-powered coding. Microsoft is said to be pushing for access to Windsurf's intellectual property, potentially to strengthen its own GitHub Copilot product. OpenAI, however, has apparently resisted these efforts. - Ends