Nvidia(US:NVDA) 国芯网·2025-09-23 13:56Core Viewpoint - OpenAI and NVIDIA have announced a strategic partnership involving an investment of up to $100 billion, aimed at deploying a 10GW AI data center to train next-generation models, significantly enhancing computational capabilities for AI development [1][4]. Group 1: Partnership Details - The partnership will see OpenAI deploy a 10GW capacity data center equipped with 4 to 5 million NVIDIA GPUs, which is equivalent to NVIDIA's total shipment for the year 2025 [1]. - The investment will be phased, with the first 1GW data center expected to be operational in the second half of 2026, based on the Vera Rubin platform, with further investments contingent on progress [1][4]. Group 2: Cost Efficiency and Strategic Implications - For OpenAI, this agreement addresses a critical bottleneck in its transition from a non-profit to a for-profit entity, particularly regarding the high costs of computational power, which currently account for 35% of the total expenses for training GPT-5 level models [4]. - Establishing its own data center could reduce long-term operational costs by 70%, allowing OpenAI to explore advanced capabilities in AGI (Artificial General Intelligence) and automate global logistics through robotic clusters [4]. Group 3: Collaborative Focus - OpenAI will consider NVIDIA as its primary strategic computing and networking partner for its AI factory expansion plans, focusing on optimizing the technology roadmap where OpenAI will concentrate on models and infrastructure software, while NVIDIA will focus on hardware and software ecosystems [5]. - This collaboration strengthens the existing deep partnership between OpenAI and NVIDIA, which includes a network of partners such as Microsoft, Oracle, SoftBank, and the Stargate Alliance, all working towards building advanced AI infrastructure globally [5].