At the GTC 2025 conference in San Jose, Nvidia’s CEO Jensen Huang revealed the company’s ambitious lineup of next-generation GPUs designed to supercharge AI performance and data processing. The announcement signals a bold leap forward, with new GPUs named Blackwell Ultra, Vera Rubin, and Feynman set to roll out over the next few years.
Vera Rubin GPU Series: Nvidia’s AI Powerhouse for 2026
Slated for release in the second half of 2026, the Vera Rubin GPU is poised to be Nvidia’s most powerful AI chip yet. This dual-GPU package comes loaded with tens of gigabytes of memory and a custom-built Nvidia CPU, Vera, engineered specifically for AI workloads.
Nvidia claims that Vera Rubin will significantly outperform the current Grace Blackwell series, especially in AI training and inferencing tasks. When combined with the Vera CPU, this powerhouse can hit up to 50 petaflops of AI inferencing performance — more than double the 20 petaflops delivered by Blackwell GPUs. Notably, the Vera CPU itself runs twice as fast as the processor used in Grace Blackwell, pushing the performance boundaries even further.
Rubin Ultra: Scaling AI Capabilities to New Heights by 2027
The Rubin Ultra GPU, expected in the second half of 2027, takes things a step further. Designed as a four-GPU package, Rubin Ultra will offer an astonishing 100 petaflops of AI inferencing performance. This massive boost is expected to power the most demanding AI models, making it a perfect fit for industries pushing the frontiers of generative AI and large language models.
Blackwell Ultra Drops in 2025 with Expanded Memory
Before Rubin’s debut, Nvidia plans to launch the Blackwell Ultra GPU in the second half of 2025. While Blackwell Ultra maintains the 20 petaflops AI performance of the original Blackwell chips, it packs a serious upgrade in memory capacity.
Each Blackwell Ultra chip comes with 288GB of memory, a major jump from the 192GB offered by standard Blackwell GPUs. This memory expansion targets AI developers and enterprises needing larger model capacity and faster processing speeds for complex workloads.
Feynman GPUs: Nvidia’s Mysterious Future Chipset Arriving in 2028
Looking even further ahead, Nvidia teased the arrival of the Feynman GPUs, named after the legendary physicist Richard Feynman. Details remain scarce, but Huang confirmed that the Feynman architecture will also utilize the powerful Vera CPU.
Expected sometime in 2028, Feynman GPUs are positioned as the successor to the Rubin series, promising groundbreaking advancements that could reshape AI computing as we know it.
Nvidia’s Next GPUs Are Set to Transform AI Performance
With Blackwell Ultra, Vera Rubin, Rubin Ultra, and Feynman GPUs in the pipeline, Nvidia is signaling its dominance in the AI chip race. These powerful GPUs are designed to meet the growing demand for faster, more efficient AI inferencing and training.
As industries increasingly rely on AI to solve complex problems, Nvidia’s next GPUs will likely play a central role in shaping the future of computing.
