OpenAI has recently reached a strategically significant long-term partnership with AI chip startup Cerebras, aiming to significantly reduce computing costs and decrease reliance on NVIDIA chips through a diversified hardware strategy. According to informed sources, OpenAI has committed to paying over $2 billion to Cerebras over the next three years for using its chip-powered server clusters. This amount is double the previous market rumors, reflecting OpenAI's strong determination in achieving computational autonomy.

OpenAI, artificial intelligence, AI

As a core part of the agreement, OpenAI will also receive minority stock warrants from Cerebras, with the ownership percentage expected to increase as investment grows. Additionally, OpenAI has agreed to invest about $1 billion to fund Cerebras' construction of high-performance data centers required to run its AI products. Cerebras is known for its unique wafer-scale engine (WSE) technology, which offers much faster memory access speeds than traditional GPUs, which is crucial for OpenAI to optimize the response latency of its next-generation inference models (such as Opus4.7).

This move is seen as a watershed event in the AI infrastructure field. By shifting from a single supplier to a hybrid architecture that includes Cerebras, AMD, and custom chips, OpenAI is attempting to break the current monopoly in the AI computing market. Although OpenAI still needs to prove the stability of emerging chip architectures in handling large-scale, high-complexity customer relationships, the $2 billion deep integration is enough to shake up the existing perceptions of SaaS and underlying computing power structures in the public market.