Nvidia Unveils Two Game-Changers: Mass Production of the Rubin Computing Powerhouse and Alpamayo Open Source to Usher in the Physical AI Era

At the 2026 CES venue in Las Vegas, NVIDIA CEO Jensen Huang’s 90-minute speech served as a "bellwether" for the tech industry. When he held up the Rubin platform integrating six new chips and announced its mass production plan as well as the decision to open-source the Alpamayo 1 autonomous driving inference model, the audience erupted in thunderous applause. This was not just a hardware upgrade and technological opening-up, but NVIDIA’s official declaration of the "era of Physical AI": "The ChatGPT moment for the physical world has arrived, and embodied intelligence will reshape tens of millions of factories and 200,000 warehouses worldwide."

The launch of the Rubin platform marks a shift in the AI computing power race from "scale competition" to an "efficiency revolution". Named after astronomer Vera Rubin, this supercomputing platform adopts a six-chip collaborative design, integrating core components such as the Rubin GPU, Vera CPU, and NVLink 6 switch to build a full-stack optimized computing infrastructure. Its core metrics are nothing short of revolutionary: it delivers 50 PFLOPS of inference computing power, five times that of the previous-generation Blackwell platform; its HBM4 bandwidth exceeds 22TB/s, the number of transistors has increased to 336 billion, while inference costs have been cut by a factor of ten. Jensen Huang emphasized that this "ultra-optimized collaborative design" is not just about stacking components, but about addressing the core bottleneck to large-scale AI deployment through deep hardware-software integration. Against the backdrop of AI computing demand growing tenfold annually, Rubin transforms large-model inference from a "laboratory luxury" into an "industrial necessity". Microsoft has already confirmed it will adopt the Rubin platform to build its next-generation Fairwater AI super factory, and cloud service providers such as Coreweave have simultaneously announced plans to deploy the system in the first wave.

If the Rubin platform is the "computing engine" for Physical AI, then Alpamayo 1, the industry’s first open-source visual-language-action (VLA) integrated autonomous driving inference model, breaks through the limitations of traditional autonomous driving systems that separate "perception" and "decision-making". It can understand the causal relationships of the physical world like humans do, making precise decisions in complex traffic conditions while also explaining "why such operations are carried out". This capability stems from its deep learning of physical laws—by integrating years of NVIDIA’s simulation data and real-world road testing experience, the model can handle extreme scenarios such as heavy rain and sudden obstacles, providing critical technical support for L4/L5 autonomous driving. What is even more disruptive to the industry is its open-source strategy: any enterprise can conduct secondary development based on this model, which will significantly lower the R&D threshold for advanced autonomous driving and accelerate technology popularization. In his speech, Jensen Huang specifically highlighted the importance of open-source power, citing the breakthroughs of China’s DeepSeek R1 model as an example and stating that "open-source ensures that no one is left behind in the intelligence revolution".

Behind these two core launches lies NVIDIA’s profound insight into industry trends. Jensen Huang defined the current industry phase as a "once-in-a-decade platform reset period": the computing paradigm is shifting from "programming software" to "training software", and every layer of the technology stack is being reinvented. Physical AI, as the core direction of this transformation, aims to move AI beyond the screen and into real-world scenarios such as factories, warehouses, and roads. In manufacturing, embodied intelligent robots can leverage Rubin’s computing power and Alpamayo’s inference capabilities to perform complex tasks such as precision assembly and flexible sorting; in logistics, intelligent warehousing systems can achieve fully unmanned end-to-end operations; in the autonomous driving sector, open-source models will drive the industry to evolve from "isolated breakthroughs" to "ecosystem collaboration". Practices by enterprises such as Pangqi Technology have already demonstrated that embodied intelligent robots can boost production line efficiency by five times and expand warehouse storage capacity by 120%, and this productivity revolution is emerging as the core driver of global industrial upgrading.

Notably, NVIDIA’s strategy has evolved from being a "chip supplier" to a "full-stack AI ecosystem builder". The Rubin platform not only provides hardware support, but also optimizes the AI software stack through collaboration with Red Hat; Alpamayo 1 opens up model weights and development tools, complementing NVIDIA’s NeMo ecosystem. This three-in-one "hardware + software + open-source" model will build a strong technological moat while accelerating the formation of industry standards. As Jensen Huang put it, the essence of the AI revolution is to "make computing ubiquitous", and the combination of Rubin and Alpamayo is the key piece of the puzzle to realize this vision.

In the second half of 2026, the Rubin platform will be officially delivered, with initial applications concentrated in autonomous driving, intelligent manufacturing, and cloud computing. As computing costs drop sharply and open-source models gain widespread adoption, Physical AI will move from high-end scenarios to large-scale deployment, and an unprecedented technological transformation is on the horizon. NVIDIA’s dual launches have not only reshaped the competitive landscape of the AI industry, but also defined the core direction of technological development for the next decade. When computing power is no longer scarce and intelligence is open-source, the way humans interact with the physical world will undergo a revolutionary transformation.