OpenAI moves toward integration of Google TPU chips
OpenAI has begun recruiting engineers to lead the optimization of its AI models for Google's TPU hardware, signaling a shift in infrastructure strategy.
OpenAI is preparing to incorporate Google’s Tensor Processing Unit (TPU) chips into its AI infrastructure. The company has initiated a recruitment drive for engineers tasked with optimizing its large-scale models to function on Google’s hardware, according to reports from The Information.
This move represents a return to Google’s ecosystem, which served as a foundation for OpenAI during its early development. In recent years, the firm has primarily relied on Nvidia’s hardware, while subsequently branching out to include AMD, AWS, and proprietary chip developments, such as the Jalapeño project.
While OpenAI confirmed in July 2025 that it was conducting early-stage testing on Google’s TPUs, it noted at the time that no large-scale deployment was planned. The shift in focus to formal optimization indicates that the company is actively pursuing hardware diversification to meet increasing computational demands.