The company has not accessed the investment-grade bond market since raising $5 billion 2021, a source familiar with the matter told Reuters in a report published Monday (June 15).
The bond is comprised of seven tranches of notes, maturing as late as 2056, the report added, citing a term sheet seen by Reuters. A Nvidia spokesperson told the news outlet that the company wants to use the proceeds for general corporate purposes, such as the repayment and refinancing of outstanding notes.
The report also pointed out that Big Tech firms have shown no plans to slow their spending on AI projects, with total investments expected to exceed $700 billion for 2026, compared to around $400 billion in 2025.
According to the report, Meta filed in October for its largest bond offering of up to $30 billion, while Google recently revealed plans to sell Japanese yen-denominated bonds for the first time.
Reuters noted that although Nvidia has not been constructing large-scale data centers, it does make the chips used in those centers, and is thus witnessing strong demand from companies seeking to train and run increasingly advanced AI models.
To keep up with the fast-moving industry, the report added, Nvidia has spent heavily to develop the most advanced processors, rolling out a new family of chips every year, each with greater AI capabilities than its predecessor.
In other Nvidia news, PYMNTS wrote last week about the launch of Cosmos 3, an open world foundation model for physical AI trained using 20 trillion tokens of multimodal data — including almost one billion images, 400 million real and synthetic videos, audio and action data from humans and robots.
The model is designed for machines that need to understand the physical world before acting in it. Nvidia founder and CEO Jensen Huang noted during the product’s launch that “the big bang of physical AI is just around the corner thanks to breakthroughs in multimodal reasoning language, vision and world models.”
“The distinction matters for anyone building or deploying physical AI,” PYMNTS wrote. “An LLM learns from text. A world foundation model learns from physical environments: how objects move, collide, fall and interact over time.”
Cosmos is different from a video generator, the report added, in that it doesn’t only generate realistic scenes, “it predicts what a robot or vehicle should do next within them.”
