NVIDIA Vera Rubin: AI Revolution Unleashed
Unlock 10x cheaper AI inference with NVIDIA Vera Rubin, arriving 2026.
6 gen 2026 (Aggiornato il 16 feb 2026) - Scritto da Lorenzo Pellegrini
NVIDIA, the NVIDIA logo, GeForce, Quadro, and Tesla are trademarks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries.
Lorenzo Pellegrini
6 gen 2026 (Aggiornato il 16 feb 2026)
NVIDIA Vera Rubin: The Next AI Supercomputer Revolution Unveiled at CES 2026
NVIDIA has launched the Vera Rubin platform, its groundbreaking next-generation AI infrastructure named after pioneering astronomer Vera Florence Cooper Rubin. This rack-scale supercomputer promises dramatic efficiency gains, addressing the surging computational demands for advanced AI as highlighted by CEO Jensen Huang at CES 2026.
What is the NVIDIA Vera Rubin Platform?
The Vera Rubin platform represents NVIDIA's latest leap in AI hardware and software codesign. It delivers up to 10 times reduction in inference token cost and 4 times fewer GPUs needed for training mixture-of-experts models compared to the Blackwell platform. Key components include the NVIDIA Vera Rubin NVL72 rack-scale solution and the NVIDIA HGX Rubin NVL8 system, engineered for agentic AI, advanced reasoning, and massive-scale workloads.
Key Innovations Powering Vera Rubin
Vera Rubin introduces five major breakthroughs that set it apart in AI computing.
- NVIDIA Vera CPU: Features 88 custom Olympus cores with full Arm compatibility, optimized for next-generation AI factories.
- NVIDIA Rubin GPU: Delivers high-performance AI compute with HBM4 memory and an enhanced Transformer Engine.
- NVIDIA NVLink 6: Provides 3.6 TB/s GPU-to-GPU bandwidth via sixth-generation interconnect technology.
- NVIDIA ConnectX-9: Offers high-throughput, low-latency networking for scale-out AI clusters.
- NVIDIA BlueField-4 DPU: A dual-die data processing unit enhancing efficiency and security.
Additional advancements include Confidential Computing, RAS Engine for reliability, and Spectrum-X Ethernet Photonics switches, which boost power efficiency by 5 times and improve uptime.
Early Adopters and Deployment Timeline
Major cloud providers are lining up to deploy Vera Rubin systems starting in 2026. AWS, Google Cloud, Microsoft, OCI, and NVIDIA Cloud Partners like CoreWeave, Lambda, Nebius, and Nscale will integrate these instances.
- Microsoft plans Vera Rubin NVL72 in next-generation AI data centers and Fairwater superfactories for enterprise and research applications.
- CoreWeave will add Rubin-based systems in the second half of 2026, supporting training, inference, and agentic workloads.
- Nebius will offer Vera Rubin NVL72 in US and Europe data centers from H2 2026 via Nebius AI Cloud and Token Factory, targeting agentic and reasoning AI.
Supermicro also announced support for Vera Rubin NVL72 and HGX Rubin NVL8 with expanded liquid-cooled rack-scale manufacturing.
Jensen Huang's Vision for AI's Future
At CES 2026, NVIDIA founder and CEO Jensen Huang emphasized the explosive growth in AI computational needs. He positioned Vera Rubin as the blueprint for scaling AI into agentic systems capable of multistep problem-solving with long token sequences at the lowest cost per token. This platform obsoletes current AI infrastructure, accelerating innovation across industries.
Conclusion
The Vera Rubin platform marks a pivotal advancement in AI supercomputing, unifying high-performance chips, interconnects, and software for unprecedented scale and efficiency.
As AI evolves toward reasoning and agentic capabilities, Vera Rubin's deployments in 2026 will empower developers and enterprises to push boundaries. Stay tuned for how this technology transforms real-world applications.
While Vera Rubin promises unprecedented rack-scale efficiency for agentic AI, its extreme codesign risks entrenching NVIDIA's dominance into a de facto AI infrastructure monopoly, forcing hyperscalers into proprietary lock-in just as open standards could democratize the next computing era.