OpenAI Reportedly Plans to Buy Google TPUs

OpenAI is reportedly planning to buy Google’s Tensor Processing Units (TPUs), according to Reuters, in a move that could deepen ties between the AI rivals. The sources cited did not disclose financial terms or other details. OpenAI is one of NVIDIA’s largest GPU customers, and Reuters reported in June 2025 that it had begun renting Google’s AI chips.

OpenAI has posted a job opening for an engineer to lead work optimizing its most advanced frontier models for Google TPUs, BigGo Finance reported. The report describes the recruitment as a sign that OpenAI is preparing for deeper integration with Google’s chip architecture. Google began offering TPUs for direct purchase to select customers this year as companies seek more cost-effective computing and chips tailored to AI workloads.

OpenAI also has substantial financial and infrastructure ties to NVIDIA. BigGo Finance reported that NVIDIA agreed in August to provide more than $100 billion in credit support for OpenAI’s planned Ohio data center, which is expected to rely exclusively on NVIDIA’s chip architecture. NVIDIA CEO Jensen Huang said in a blog post that OpenAI’s existing and planned deployments could represent approximately $600 billion in NVIDIA compute through 2030.

Separately, Google and Oracle plan to offer TPUs to Oracle Cloud customers, the first time the chips will be available through a cloud platform outside Google’s own. Oracle plans to add them to Oracle Cloud Infrastructure in 2027. Oracle Cloud Infrastructure Senior Vice President Karan Batta described the agreement as a multi-year, multi-gigawatt partnership and said Oracle aims to offer new TPU generations as soon as they become available. Sources cited by BigGo Finance said Oracle will buy the chips directly, with a major customer shared by Google and Oracle expected to use them. Oracle also plans multi-gigawatt deployments of NVIDIA Vera Rubin systems and is adding accelerators from Cerebras and SambaNova.

OpenAI is reportedly planning to buy Google’s Tensor Processing Units (TPUs), according to Reuters, in a move that could deepen ties between the AI rivals. The sources cited did not disclose financial terms or other details. OpenAI is one of NVIDIA’s largest GPU customers, and Reuters reported in June 2025 that it had begun renting Google’s AI chips.

OpenAI has posted a job opening for an engineer to lead work optimizing its most advanced frontier models for Google TPUs, BigGo Finance reported. The report describes the recruitment as a sign that OpenAI is preparing for deeper integration with Google’s chip architecture. Google began offering TPUs for direct purchase to select customers this year as companies seek more cost-effective computing and chips tailored to AI workloads.

OpenAI also has substantial financial and infrastructure ties to NVIDIA. BigGo Finance reported that NVIDIA agreed in August to provide more than $100 billion in credit support for OpenAI’s planned Ohio data center, which is expected to rely exclusively on NVIDIA’s chip architecture. NVIDIA CEO Jensen Huang said in a blog post that OpenAI’s existing and planned deployments could represent approximately $600 billion in NVIDIA compute through 2030.

Separately, Google and Oracle plan to offer TPUs to Oracle Cloud customers, the first time the chips will be available through a cloud platform outside Google’s own. Oracle plans to add them to Oracle Cloud Infrastructure in 2027. Oracle Cloud Infrastructure Senior Vice President Karan Batta described the agreement as a multi-year, multi-gigawatt partnership and said Oracle aims to offer new TPU generations as soon as they become available. Sources cited by BigGo Finance said Oracle will buy the chips directly, with a major customer shared by Google and Oracle expected to use them. Oracle also plans multi-gigawatt deployments of NVIDIA Vera Rubin systems and is adding accelerators from Cerebras and SambaNova.