Nvidia VP Says GPUs Are the “Currency” of Today’s AI Researchers

In a revealing discussion about the inner workings of AI innovation at Nvidia, Jonathan Cohen, Vice President of Applied Research, described graphics processing units (GPUs) as the most valuable asset for AI researchers—comparing them to currency in the fast-paced world of machine learning.

Speaking with Nvidia Developer, Cohen highlighted how critical compute power has become in accelerating breakthroughs, noting that the availability of GPUs directly impacts how fast AI models can be developed.

“These days, the currency in any AI researcher is how many GPUs they get access to,” Cohen said. “And that’s just as true at Nvidia as anywhere else.”

Inside the Birth of Llama Nemotron

Cohen led the development of Llama Nemotron, Nvidia’s latest line of AI models launched in March 2025. What’s remarkable isn’t just what the models can do, but how quickly they came together—within just one to two months.

This rapid progress was made possible through internal collaboration across Nvidia, with teams voluntarily reallocating their compute resources to support the model training.

“Researchers willingly gave up their GPU time so we could get these models trained quickly,” he explained.

The Power of Swarming and Cross-Team Collaboration

Cohen credited Nvidia’s flat organizational structure and “swarm” approach for enabling swift progress. When a priority project like Llama Nemotron arises, managers across departments are encouraged to assess whether their teams can contribute, even if the work falls outside their usual scope.

“We don’t rely on formal structures. If something is important, people come together from across the company to make it happen,” Cohen said.

The result was a cross-functional initiative where staff from various domains united behind a shared mission—driven not by hierarchy but by collective focus.

Sacrifices and Culture of Support

Nvidia’s open, mission-first culture allowed researchers to put aside individual priorities for a greater cause. Cohen described it as a moment of “egoless decision-making” that demonstrated strong leadership and team spirit.

“There were a lot of sacrifices—of compute power, of time—but the team rallied, and that’s what made it possible.”

A Glimpse Into Nvidia’s AI Engine Room

While Nvidia did not issue an official comment prior to publication, Cohen’s interview offers rare insight into the collaborative engine behind one of the most influential AI companies in the world.

As AI models grow in complexity, access to powerful GPUs and a culture of adaptability will likely define who stays ahead.https://www.youtube.com/watch?v=xS9juMlXVlY