Nvidia's Kerris: AI is a new compute model
Nvidia media chief Richard Kerris told Runway's New York summit that studios should treat AI as a new compute model to rebuild workflows around, citing Runway Gen-4.5 running near-instantly on Vera Rubin.
Richard Kerris, Nvidia's VP and GM of Media and Entertainment, told the 2026 Runway AI Summit in New York that most of the industry is misusing AI. In his words, companies treat it "like a feature" or "a better version of Google." Runway published a write-up of the talk.
Kerris traced a line from RTX, which made rendering real-time eight years ago, to today's generative tools. "When you can simulate reality, it's the first step in achieving what we're doing today with AI," he said. He said the remaining bottleneck is generation time, and that Nvidia built its Vera Rubin platform to remove it.
Runway was Nvidia's major partner at the Vera Rubin launch. "Runway was able to bring their 4.5 software up in a day on Vera Rubin, and the results blew us all away," Kerris said. According to Runway's account, Gen-4.5 took minutes to generate scenes three months earlier. On Vera Rubin it generates them instantly. Neither company gave benchmark figures.
Kerris argued that AI replaces the usual relay from storyboard to wireframe to animatic. "AI is now the pipeline," he said, which would let teams go back and change earlier stages without starting over. He also pointed to a new Nvidia partnership with Universal Music Group that trains models on UMG's archives so users can query them directly. He did not give terms.
His advice to leaders was to pick one workflow, find where it costs the most time and money, and build agents for that step or work with partners who already have them. He also said governance should stay inside the workflow.