Meet Blackwell Ultra, Vera Rubin, and Feynman: Nvidia’s Next-Gen GPUs Revolutionizing Performance
At the recent Nvidia GTC 2025 conference in San Jose, CEO Jensen Huang unveiled a series of groundbreaking new GPUs that will soon hit the market. These innovations are set to transform the landscape of artificial intelligence (AI) and computational performance, particularly with the introduction of the highly anticipated Vera Rubin GPU.
Introducing the Vera Rubin GPU
The Vera Rubin GPU, slated for release in the second half of 2026, is expected to revolutionize AI inferencing and training tasks. This powerful GPU will feature:
- Tens of gigabytes of memory
- A custom Nvidia-designed CPU known as Vera
- Significant performance enhancements compared to its predecessor, Grace Blackwell
Performance Capabilities of Vera Rubin
Nvidia claims that when used in conjunction with the Vera CPU, the Vera Rubin can achieve:
- Up to 50 petaflops of performance during inference, more than double the 20 petaflops offered by the current Blackwell chips.
- A CPU performance that is twice as fast as that found in the Grace Blackwell GPU.
Upcoming Releases: Rubin Ultra and Blackwell Ultra
Following the Vera Rubin, Nvidia plans to release the Rubin Ultra in the second half of 2027. This advanced collection will consist of four GPUs in a single package, boasting an impressive:
- 100 petaflops of performance
In the near term, Nvidia will launch the Blackwell Ultra in H2 2025. The Blackwell Ultra will come in multiple configurations, with a single Ultra chip providing:
- 20 petaflops of AI performance
- 288GB of memory, an increase from the 192GB found in the standard Blackwell.
The Future of Feynman GPUs
Looking ahead, Nvidia is also developing the Feynman GPUs, named after the renowned physicist Richard Feynman. Although details about the architecture remain sparse, Huang confirmed that Feynman will utilize a Vera CPU and is expected to launch in 2028.
For more information on Nvidia’s latest advancements, visit the official Nvidia website. Stay tuned for more updates on these exciting developments in GPU technology!
Note: An earlier version of this article incorrectly stated that Vera Rubin would have “tens of terabytes” of memory. It will actually feature “tens of gigabytes” of memory. We apologize for any confusion.
