Nvidia Unveils Next-Gen GPUs: Blackwell Ultra, Vera Rubin, and Feynman

Nvidia CEO Jensen Huang announced the upcoming release of new GPUs, including Vera Rubin, Blackwell Ultra, and Feynman, at the GTC 2025 conference.

At the GTC 2025 conference in San Jose, Nvidia’s CEO Jensen Huang didn’t just introduce new GPUs—he basically dropped the mic with a lineup that’s set to redefine performance. The Vera Rubin GPU, landing in late 2026, isn’t playing around with its tens of terabytes of memory and a custom Nvidia-designed CPU (yep, also named Vera). This beast is designed to leave its predecessor, Grace Blackwell, in the dust, especially when it comes to AI inferencing and training. Pair it with the Vera CPU, and you’re looking at up to 50 petaflops in AI model inference—more than double what Blackwell can muster. Talk about an upgrade.

But wait, there’s more. The Rubin Ultra, slated for late 2027, packs four GPUs into a single package, pushing the envelope to a staggering 100 petaflops. And if you’re thinking that’s a ways off, the Blackwell Ultra GPU is coming much sooner (late 2025, to be exact), with configurations boasting up to 288GB of memory and a solid 20 petaflops AI performance boost over the original Blackwell. Not too shabby, right?

And for those who like to plan way ahead, Nvidia’s already teasing the Feynman GPUs by 2028. Named after the legendary Richard Feynman (because why not aim for genius-level performance?), details are still under wraps. But we do know it’ll include the Vera CPU and is expected to take the baton from Vera Rubin, pushing GPU tech into the next stratosphere. So, buckle up—the future of computing is looking ridiculously fast.