Jensen Huang talks AI's insatiable appetite for GPUs at the World Economic Forum: 'spot price of GPU rentals is going up, not just the latest generation, but two generation old GPUs'
Jensen Huang has well and truly escaped gaming obscurity these days. The Nvidia CEO just took the stage at the World Economic Forum to talk about AI demand, bubbles, and the impact that artificial intelligence has on the job market now and into the future. Of course, being the man in charge of shipping billions of dollars of AI chips, he has a particularly positive take on the lot.
Chatting to Larry Fink, CEO of Blackrock, who called Nvidia 'nuh-vidia' throughout the conversation, Huang had a few things to say about his booming business.
"One good test on the AI bubble is to recognize that Nvidia now has now millions of Nvidia GPUs in the cloud. We're in every cloud. You know. We're used everywhere.
"And if you try to rent an Nvidia GPU these days, it's so incredibly hard, and the spot price of GPU rentals is going up, not just the latest generation, but two generation old GPUs. The spot price of rentals are going up, and the reason for that is because the number of AI companies that are being created, the number of companies shifting their R&D budget."
Nvidia's datacentre business, the topic of conversation here, has soared to many times its gaming revenue in recent years. It hit a record $51.2 billion through August to September last year. Gaming revenue during the same period was $4.3 billion. Though the demand for datacentre GPUs, such as upcoming Rubin chips, B100/200, and older chips like the H100/H200, do have a knock-on effect in gaming GPUs.
Just look to the price of the RTX 5090—a more than capable card for smaller-scale AI deployment—which launched at $1,999 and is now selling for around $4,000+.
Capacity for production is one factor in all of this, as when AI comes knocking for chips, be those GPUs or memory, there's only so much readily available at any one time. That's why we're seeing RAM prices skyrocket: AI demand outstrips supply and datacenter buildout is more valuable than availability for consumers.
In Huang's words, the fervour to buy more AI chips is "the largest infrastructure build out in human history."
"That's going to create a lot of jobs," he says, citing skilled labor requirements for building out massive datacenters and increasing energy output to meet their demands.
The impact of AI on the job market was one of the main topics of discussion through the talk. Huang, in no uncertain terms, suggested that AI will create a labour shortage, rather than take jobs from humans. He suggests one way to understand his way of thinking is to look beyond the tasks required to do a job and instead determining the purpose of a job.
"Now, the easiest way to think about whether what is the impact of AI on a particular job is to understand what is the purpose of the job and what is the task of the job," Huang says.
"If you just put a camera on the two of us and just watched us, you would probably think the two of us are typists, because I spend all of my time typing and and so if AI could automate so much word prediction and help us type, then we would be out of jobs. But obviously that's not our purpose. And so the question is, what is the purpose of your job? In the case of radiologists and nurses, it's to care for people, and that purpose is enhanced and made more productive because the task has been automated."
I do think there are instances where this is not the case. When AI is entirely replacing the purpose of someone's job—or intrinsically changing it or shifting it further away—such as generative AI and artists or musicians, or closer to home, when AI tools are used to summarise news stories. When the AI is not owned or run by those professions affected by its use, it's not as often beneficial. Jobs may be created in some sectors, but similarly, jobs may move from other sectors entirely as it permeates into the wider market.
Though Huang aims to keep things positive about AI, as you would expect from a CEO in his position signing off with a request for people to "get involved". And if anything is clear from this talk, according to these two, it wouldn't appear there's signs of slowing for the rapid buildout of AI. If anything, they suggest the large-scale investment in AI is only just beginning.