News
OpenAI被曝将向谷歌采购TPU芯片，算力供应链格局再变 10月28日消息，据媒体报道，作为全球人工智能领域的头部玩家，OpenAI正计划向谷歌采购人工智能芯片
2 min read
Source: Telegram AI频道
OpenAI was revealed to be purchasing TPU chips from Google, and the computing power supply chain pattern has changed again. According to media reports on October 28, as a leading player in the global artificial intelligence field, OpenAI is planning to purchase artificial intelligence chips from Google. For a long time, OpenAI has been one of the most important buyers of NVIDIA graphics processing units (GPUs), mainly relying on its chips to complete the training of massive AI models and complex inference operations. As industry demand for computing power continues to soar, companies are actively seeking alternatives that are more cost-effective and suitable for running AI models. Google's self-developed TPU (Tensor Processor) has gradually come to prominence in recent years and began direct sales to some external customers this year. Looking back at the evolution of the cooperation between the two parties, a clear "three jumps" have been completed in the past 16 months: In June 2025, OpenAI began renting Google's computing power for the first time; in July of the same year, it publicly stated that it was conducting early testing of Google TPU, but there was no large-scale deployment plan; and now, both parties have officially moved towards a substantial procurement plan. In addition, OpenAI is currently actively recruiting engineers responsible for adapting and optimizing its models on TPU, and the underlying adaptation of the model layer is usually an important leading indicator for large-scale procurement. Looking back on its computing power expansion trajectory, Google TPU played an important role in the early days of OpenAI. Since then, OpenAI has fully shifted to the NVIDIA ecosystem and has successively connected to chips from multiple suppliers including AMD, Cerebras and AWS. For the industry landscape, this shift sends a clear signal: The exclusivity of Nvidia's orders is gradually waning as its most important customers forge new branches of computing power beyond the single GPU route. According to data from the authoritative organization TrendForce, NVIDIA's share of the AI server processor market is expected to decline from 69.6% in 2025 to 64% in 2026. The overall industry trend is highly consistent with the current diversified layout of technology and supply chain. via AI News (author: AI Base)