In a landmark series of announcements at CES 2026 in Las Vegas, Nvidia has unveiled two revolutionary advancements poised to reshape the artificial intelligence landscape: the Rubin computing architecture and the Alpamayo AI model family for autonomous vehicles. These developments represent a significant leap in both hardware and software for AI applications.
The Rubin Architecture: A New Benchmark in AI Compute
CEO Jensen Huang officially launched the Rubin computing architecture, named for astronomer Vera Florence Cooper Rubin. The platform, now in full production, is designed to meet the exponentially growing computational demands of AI. It comprises six specialized chips, including the Rubin GPU, enhanced NVLink and Bluefield interconnection systems, new external storage tiers, and the Vera CPU component focused on agentic reasoning. According to Nvidia's internal benchmarks, Rubin delivers 3.5x faster model training and 5x faster inference than its predecessor, Blackwell, while achieving an 8x improvement in inference compute per watt.
The architecture has already secured widespread industry adoption. Major cloud providers, including Amazon Web Services, and leading AI firms like Anthropic and OpenAI, plan to integrate Rubin chips. Research institutions such as Lawrence Berkeley National Lab will incorporate it into next-generation supercomputers like the Doudna. Huang projected that between $3 trillion and $4 trillion will flow into AI infrastructure development over the next five years, underscoring Rubin's strategic importance.
Alpamayo: The "ChatGPT Moment for Physical AI"
Simultaneously, Nvidia introduced Alpamayo, an open-source family of AI models engineered to grant autonomous vehicles human-like reasoning capabilities. The core model, Alpamayo 1, is a 10-billion-parameter Vision-Language-Action (VLA) model that employs a chain-of-thought reasoning process. Unlike conventional systems, it breaks down complex scenarios—like a malfunctioning traffic light—evaluates multiple outcomes, and selects the optimal, explainable action.
Vice President of Automotive Ali Kani stated the model "reasons through every possibility" before acting. CEO Jensen Huang emphasized this shift toward explainable AI (XAI), noting the system can articulate "what action it is about to take, the reasons by which it came about that action"—a critical development for regulatory approval and public trust.
An Open Ecosystem for Acceleration
Nvidia is releasing a complete open ecosystem to foster industry-wide development. This includes the open-source Alpamayo 1 model on Hugging Face, a curated dataset of over 1,700 hours of diverse driving data, the AlpaSim simulation framework on GitHub, and integration with Nvidia's Cosmos generative world models for synthetic data creation. This democratizes access to advanced physical AI, allowing startups and researchers to innovate without requiring billions of miles of real-world data.
Together, these announcements position Nvidia at the forefront of both the hardware and software frontiers of AI, aiming to accelerate innovation in autonomous transportation, robotic logistics, and beyond, while setting new standards for computational efficiency and machine reasoning.
