The secret to getting more of the highly coveted Nvidia graphics processing units essential for advancing artificial intelligence and machine learning might be yellowtail sashimi.
At a meeting with analysts last week, billionaire Oracle cofounder and chief technology officer Larry Ellison told the audience that he and world’s-richest-man Elon Musk took Nvidia CEO Jensen Huang out to dinner at Nobu Palo Alto and “begged” Huang to give them more GPUs.
“I would describe the dinner as Oracle—me and Elon begging Jensen for GPUs,” Ellison recalled. “Please take our money. Please take our money. By the way, I got dinner. No, no, take more of it. We need you to take more of our money please.”
The outcome was a net positive, said Ellison. “It went ok. I mean, it worked.”
Ellison, whose famed fortune is estimated to be $206 billion, is a prominent figure in the tech industry and serves as executive chairman of enterprise software firm Oracle. He is known for anticipating significant technological shifts, including accurately positioning the $475 billion tech giant to reap the benefits of the internet in its early days. He has further solidified the company by shifting it to capitalize on the surging need for cloud-based enterprise infrastructure. The company has maintained a strong relationship with Nvidia, which controls more than 80% of the AI chip market, and is the prime choice for companies working on AI.
At last week’s meeting Ellison said the competitive landscape won’t hinder Oracle’s growth. The simplest way of looking at it, Ellison replied, is the phrase,“the race goes on.” He likened the current competition in AI to a Formula One driving race.
“You’ve got three people on the podium, but there’s really kind of one winner,” said Ellison. “Someone is going to be better at this than anyone else, and multiple people are trying and there is a race.”
For its part, Oracle is making gargantuan investments in GPU technology, particularly for AI applications. The tech giant this month announced its first quarter results for fiscal 2025 and disclosed revenues were up 7% to $13.3 billion, year-over-year. Profits were $2.9 billion. Oracle also disclosed that it has 162 cloud datacenters in operation or under construction around the world and the largest of its datacenters is 800 megawatts and will house “acres of Nvidia GPU clusters” used to train large-scale AI models. In the first quarter, Oracle signed 42 more cloud GPU contracts, totaling $3 billion.
Similarly, Tesla, where Musk serves as CEO, relies on Nvidia GPUs to power the supercomputers that train its neural networks for self-driving and assisted driving technology.
According to Ellison, being the first to build the most capable neural network in the world “is a big deal.” Given the desire to be at the forefront, many executives in AI are begging Huang for GPUs and building datacenters, he said.
“Does anyone know what it costs to build a frontier model over the next three years?” Ellison was referring to the most advanced, bleeding-edge AI systems.
“One hundred billion,” said Ellison. “That’s going to get you in the game.”
“Not a lot of people and not a lot of companies and not a lot of countries will participate.”
Nvidia and Tesla did not immediately respond to Fortune’s request for comment.
This story was originally featured on Fortune.com
Add Comment