$100 billion investment in OpenAI from NVIDIA: Data centers will use the energy of 10 nuclear power plants

NVIDIA recently announced its intention to invest $100 billion in OpenAI. The investment will be made as part of a major deal for new data centers to launch AI models and train them.

The companies have agreed to deploy the NVIDIA system for OpenAI's new infrastructure with an approximate capacity that can be provided by 10 nuclear reactors or 10 GW.

Before the collaboration with NVIDIA, Microsoft was the company's largest investor. In January of this year, Microsoft changed the terms of the partnership, thus allowing OpenAI to build partnerships with other investors. Since then, OpenAI has also been collaborating with SoftBank and Oracle.

Currently, the first phase of the infrastructure is planned to begin operating in the second half of 2026 on the NVIDIA Vera Rubin platform, and there is also the possibility of Rubin Ultra. Powerful accelerators with 76 TB of HBM4 memory will be available, which should ensure the execution of FP4 logical conclusions at a speed of 3.6 exaflops. TSMC factories are already producing Vera central processors and Rubin graphics processors.

The scale of the new project is simply enormous, with 10 GW of power capable of running 4-5 million graphics processors. For example, modern data centers use approximately 50-100 MW. The planned capacity can even be compared to the energy consumption needs of several cities.

Financial costs of a large-scale project

In terms of financial investment, building a 1 GW data center can cost an average of $50 billion, more than half of which will be spent on NVIDIA systems and chips. Therefore, investment in 10 GW of infrastructure could exceed $500 billion. It is currently unknown what form NVIDIA's investment will take — financial, chips, or cloud credits. 

The need for large-scale infrastructure is growing rapidly, with OpenAI currently having more than 700 million weekly active users. The new partnership has also boosted NVIDIA's stock price.

The partnership between the companies could be the largest project in AI. Users will potentially have access to new services and improved AI tool performance. At the same time, such a scale could negatively impact environmental sustainability and raise many other challenges for the environment.