NVIDIA Bets $100 Billion on OpenAI: The Dawn of an AI Power Shift

A $100 Billion Bet to Build an ‘AI Power Grid’

NVIDIA is set to invest up to $100 billion in OpenAI to jointly build data center infrastructure with a capacity of 10 gigawatts (GW). This project, equivalent to the output of ten nuclear power plants, marks an unprecedented scale of computing power dedicated to artificial intelligence.

The investment will be injected in stages, with the first phase featuring NVIDIA’s next-generation AI chip, Vera Rubin. The two companies are preparing to finalize a strategic partnership agreement, aiming to bring the new infrastructure online in the second half of 2026.

OpenAI Accelerates Model Development with Supercomputing

For OpenAI, the surge in computing demand triggered by the global expansion of ChatGPT has made large-scale infrastructure an urgent necessity.

Currently, ChatGPT boasts nearly 700 million weekly active users, while simultaneously powering a wide range of AI services. CEO Sam Altman emphasized, “Computing infrastructure will be the foundation of the future economy. Our collaboration with NVIDIA lays the groundwork for the next wave of AI innovation.”

Through this investment, OpenAI will pursue both in-house AI chip development and massive data center construction — strengthening its independence while expanding its growth capacity.

NVIDIA Sharpens Its Competitive Edge in AI Chips

NVIDIA, meanwhile, is leveraging this alliance to consolidate its dominance in the AI chip market. CEO Jensen Huang explained, “10GW equates to roughly 4 to 5 million GPUs - comparable to NVIDIA’s total shipments for this year.”

By ensuring OpenAI adopts its chips at scale, NVIDIA effectively narrows the entry points for rivals such as AMD and Broadcom, while shaping the cloud and AI service ecosystem around its own technological standards.

Speculation has also arisen that NVIDIA may acquire a stake in OpenAI, signaling a deeper integration of hardware and software strategies.

Microsoft, Oracle, SoftBank...The AI Infrastructure Arms Race

OpenAI is already bound by strategic partnerships with Microsoft, and is also driving the “Stargate” data center project in collaboration with Oracle and SoftBank. Additionally, it is reportedly co-developing custom AI chips with Broadcom.

NVIDIA’s entry into this network strengthens OpenAI’s ecosystem by adding a formidable compute partner, complementing existing alliances.

Altman highlighted the division of roles among partners, saying, “Microsoft and NVIDIA are not just passive investors - they are our most important collaborators, each with a distinct role in this shared mission.”

By Ju-baek Shin, KMJ /  jbshin@kmjournal.net