• NVIDIA to invest up to $100 billion to power OpenAI's deployment of 10 gigawatts of datacenter capacity.
  • The first 1 GW phase, representing a $10 billion investment, is slated for the second half of 2026.
  • Bernstein analysts view the deal as a multi-year growth catalyst for NVIDIA, reinforcing its AI leadership.

NVIDIA's strategic partnership with OpenAI, announced late Thursday, represents one of the largest private infrastructure commitments in the technology sector, positioning the chipmaker at the center of the next wave of AI expansion. The agreement will see NVIDIA provide the hardware backbone—comprising millions of its latest GPUs—for OpenAI's ambitious plan to build out 10 GW of dedicated AI datacenter capacity.

According to people familiar with the terms, the capital commitment is tiered and contingent on build-out milestones. The initial phase, a 1 GW datacenter cluster powered by NVIDIA's forthcoming Vera Rubin platform, is already in advanced planning stages and is expected to come online in 2H 2026. This first step alone is valued at approximately $10 billion.

"This isn't just a supplier relationship; it's a deep, strategic alignment," said one person involved in the negotiations, who asked not to be identified because the details are private. "The scale is unprecedented." The deal effectively ensures NVIDIA's technology will be the foundation for OpenAI's next-generation AI models for the foreseeable future.

In a note to clients, Bernstein analyst Stacy Rasgon reiterated an Outperform rating on NVIDIA with a $225 price target, calling the partnership a "positive catalyst" that supports the company's growth trajectory well into 2026 and beyond. Rasgon noted that while the news caused some short-term volatility for peers like Broadcom (AVGO), the underlying trend of massive AI infrastructure spending is a tailwind for the entire high-quality semiconductor ecosystem.

Efforts to reach NVIDIA and OpenAI for additional comment were not immediately successful. The partnership underscores a significant shift in how leading AI labs are securing compute capacity, moving beyond traditional cloud service contracts to direct, large-scale infrastructure ownership and partnerships. This move could pressure other cloud providers to accelerate their own AI hardware roadmaps.

The 10 GW power requirement highlights the immense energy demands of advanced AI. For context, 10 gigawatts is roughly the output of ten large nuclear power plants, signaling that energy sourcing and efficiency will be critical constraints and competitive advantages in the race for AI supremacy. The deal is likely to draw increased regulatory and political scrutiny around the concentration of AI resources and their environmental impact.

Correction: An earlier version of this article misstated the codename of NVIDIA's upcoming platform. It is the Vera Rubin platform.