It is reported that OpenAI is considering leasing NVIDIA chips, which could save 10%-15% in expenses

Zhitong
2025.09.24 14:03
portai
I'm PortAI, I can summarize articles.

OpenAI is discussing the feasibility of leasing artificial intelligence chips with NVIDIA, expecting to save 10%-15% in cost expenditures. Through the leasing model, OpenAI does not need to raise funds for chip procurement and can obtain chips more quickly. This collaboration may set a five-year term, against the backdrop of NVIDIA planning to invest up to $100 billion in OpenAI to build AI data centers. Altman emphasized that having sufficient computing power will drive the development of important applications and plans to disclose more collaboration details in the coming months

According to reports, OpenAI is exploring the feasibility of leasing artificial intelligence chips from NVIDIA (NVDA.US) to reduce cost pressures in their large-scale collaboration on new data centers.

Insiders revealed that by leasing server chips instead of purchasing them directly, OpenAI could save 10% to 15% on cost expenditures. With the GPU leasing model, OpenAI does not need to raise additional funds for chip procurement and can obtain chips more quickly.

Currently, the ChatGPT developer led by Sam Altman has leased NVIDIA chips from cloud service providers such as Microsoft (MSFT.US) and Oracle (ORCL.US), and the new leasing agreement with NVIDIA is expected to set a five-year term, similar to OpenAI's current leasing agreement with Oracle.

The background of this leasing collaboration can be traced back to NVIDIA's recent strategic plan: it intends to gradually invest up to $100 billion in Microsoft-backed OpenAI to build and deploy AI data centers with a scale of at least 10 gigawatts based on NVIDIA systems. The first phase of the project is expected to start in the second half of 2026, deploying on NVIDIA's Vera Rubin platform.

Regarding this massive project, Altman elaborated on its profound significance in a blog post. He emphasized that if there is 10 gigawatts of computing power, artificial intelligence might find a cure for cancer or provide personalized tutoring for every student worldwide.

"If limited by the scale of computing power, we would be forced to make choices among many important applications, and no one wants to see that, so we must break through the limitations with full force." He further envisioned establishing a capacity to add 1 GW of AI infrastructure weekly to continuously expand computing power.

Altman added that more details about plans and partners will be disclosed in the coming months. He specifically pointed out, "Later this year, we will share specific financing plans; given that the expansion of computing power is directly related to revenue growth, we have some interesting new ideas."

Source: Zhitong The copyright of this article belongs to the original author/organization.
The views expressed herein are solely those of the author and do not reflect the stance of the platform. The content is intended for investment reference purposes only and shall not be considered as investment advice. Please contact us if you have any questions or suggestions regarding the content services provided by the platform.