NVIDIA arrived at the ongoing CES 2026 as a very different company from the one people once knew — a manufacturer of graphics cards.
However, over the past few years, it has quietly overhauled itself around artificial intelligence, high-performance computing, and large-scale systems that power everything from data centres to cars and robots.
CES, which is underway now in Las Vegas, gave the company a global stage to show how far that transformation has gone.
Instead of focusing on a single product line, the California-based tech company held a broad set of announcements spanning AI supercomputers, networking, storage, open models, autonomous driving, robotics, gaming, and industrial software.
Many of these updates are not consumer-facing products but building blocks meant for developers, enterprises, and partners shaping the next generation of AI systems.
NVIDIA founder and CEO Jensen Huang took the stage at the Fontainebleau Las Vegas to open CES 2026, declaring that AI is scaling into every domain and every device.
Check out the roundup of what NVIDIA revealed below
Rubin platform built around “extreme codesign”
NVIDIA introduced the Rubin platform, built around “extreme codesign,” meaning the company designed multiple core components together rather than treating them as separate parts.
Rubin combines a Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-6 Ethernet Switch.
NVIDIA says Rubin targets large workloads such as mixture-of-experts models, agentic systems, and long-context reasoning, with claims of up to 10x lower per-token inference cost and fewer GPUs required for some training workloads than Blackwell.
DGX SuperPOD: the “default” design for Rubin-scale deployments
NVIDIA has introduced the DGX SuperPOD as a key setup for Rubin-based systems used in businesses and research.
These DGX Rubin systems integrate computing, networking, and software to facilitate training, inference, and long-context reasoning, claiming to be “up to 10x” cheaper in token costs compared to Blackwell.
DGX Spark + DGX Station: deskside systems for large models
NVIDIA also focused on smaller, local systems: DGX Spark and DGX Station are “deskside AI supercomputers” designed to run open-source and frontier models locally, then scale workloads to the cloud as needed.
The company said DGX Spark can handle 100B-parameter models, and DGX Station can handle up to 1T-parameter models.
BlueField-4 + “context memory storage”
NVIDIA announced a new platform called Inference Context Memory Storage, powered by BlueField-4. This platform addresses the challenge of handling large context data generated by long-context, multi-agent AI models, which don’t fit well in traditional GPU memory.
The company claims that this new platform improves memory capacity and enables sharing of context across large clusters, offering up to five times more tokens per second and five times better power efficiency than traditional storage methods.
BlueField-4 is expected to be available in the second half of 2026, with partnerships including Dell, HPE, IBM, Nutanix, Pure Storage, Supermicro, VAST Data, WEKA, and others.
Enterprise AI Factory update
NVIDIA has added new features to its “Enterprise AI Factory validated design” to enhance security and speed up infrastructure.
The integration now includes platforms like Armis, Check Point, F5, Fortinet, Palo Alto Networks, Rafay, Red Hat OpenShift, Spectro Cloud (PaletteAI), and Trend Micro.
These additions provide benefits such as telemetry via NVIDIA DOCA Argus and improved workload isolation in Kubernetes environments.
Nemotron, Cosmos, Alpamayo, GR00T, Clara
NVIDIA announced a range of new open resources, including models, datasets, and training tools for various fields. They showcased several families of technology: Nemotron (AI agents), Cosmos (physical AI), Alpamayo (self-driving vehicles), Isaac GR00T (robotics), and Clara (biomedical).
They also shared impressive figures for their open data contributions, including 10 trillion language tokens, 500,000 robotics paths, 455,000 protein structures, and 100TB of vehicle sensor data.
Alpamayo for autonomous driving
NVIDIA introduced the Alpamayo family for autonomous driving, which includes AI models, simulation tools, and datasets focusing on rare and complex driving scenarios.
The key components are Alpamayo 1, AlpaSim, and “Physical AI Open Datasets.” Alpamayo is a reasoning model designed to develop autonomous vehicles. Early collaborators on this project include JLR, Lucid, Uber, and Berkeley DeepDrive, all working towards “level 4” autonomy.
DRIVE Hyperion ecosystem
NVIDIA is expanding its DRIVE Hyperion ecosystem and has announced partnerships with major suppliers and sensor companies, including Aeva, Bosch, and Sony.
The DRIVE Hyperion is a ready-to-use architecture that combines computing and sensors, featuring two DRIVE AGX Thor chips that deliver over 2,000 TFLOPS for advanced sensor processing and real-time tasks.
NVIDIA DRIVE AV software goes into production
NVIDIA said its DRIVE AV software will debut in the all-new Mercedes-Benz CLA, starting with an “enhanced level 2” driver-assistance system expected on U.S. roads by the end of this year. The CLA is also described as the first Mercedes model to use the MB platform.OS platform.
The company described its “dual-stack” approach: an end-to-end AI driving stack paired with a classical safety stack built on NVIDIA Halos for redundancy and guardrails.
It also listed capabilities like point-to-point urban navigation, proactive collision avoidance, and automated parking, and pointed to a cloud-to-car pipeline using DGX for training, Omniverse/Cosmos for simulation, and DRIVE AGX + Hyperion in-vehicle compute.
Siemens partnership expansion
Siemens and NVIDIA announced an expanded partnership to bring AI deeper into industrial workflows, including physical AI.
The release states that NVIDIA will provide AI infrastructure, simulation libraries, models, frameworks, and blueprints, while Siemens will commit “hundreds” of industrial AI experts, as well as hardware/software capabilities.
Both sides framed this around digital twins, faster product development, and real-time production adaptation.
RTX AI video generation on PC
NVIDIA showcased its creator tools for local AI video generation using LTX-2, claiming they can produce “up to 20 seconds of 4K video” with built-in audio and multi-keyframe support.
They also collaborated with ComfyUI to improve GPU performance by 40% and added support for NVFP4 and NVFP8 formats, which help reduce speed and VRAM usage on RTX 50 Series graphics cards.
Gaming display and rendering updates
NVIDIA’s gaming update highlighted new rendering and display features. It introduced DLSS 4.5, which includes a “6X” mode that can produce up to 5 additional frames per rendered frame, with availability expected in spring.
Additionally, they announced that G-SYNC Pulsar monitors are now available, featuring over 1,000Hz effective motion clarity and G-SYNC Ambient Adaptive Technology.
