Nvidia unveils five platforms to expand enterprise AI infrastructure

Nvidia announced five interconnected enterprise AI platforms at CES 2026, and said the Rubin inference platform has entered full production with partner deployments expected in the second half of 2026.

Nvidia presented five interconnected platform launches during Jensen Huang's CES 2026 keynote, positioning the company as an end-to-end infrastructure provider for enterprise AI. The announcements cover data center inference economics, autonomous vehicle reasoning, industrial manufacturing, desktop AI development, and consumer gaming. The company emphasized a shift from discrete chip sales toward integrated systems designed to address enterprise infrastructure bottlenecks. Organizations evaluating these offerings face questions about deployment timing, integration complexity, and total cost of ownership. Key announcements: - Rubin platform: Rubin has entered full production and is described as an "extreme codesign" of six silicon components (Rubin GPU, Vera CPU, NVLink 6, ConnectX-9, BlueField-4 DPU and Spectrum-6 switch); Nvidia reports 50 petaflops of NVFP4 inference compute, HBM4 memory bandwidth up to 22 TB/s and 288 GB per GPU, uses adaptive 4-bit precision with reported minimal accuracy loss, and partner deployments are expected in the second half of 2026 with CoreWeave among the early cloud adopters. - Alpamayo and DRIVE: Alpamayo is an open portfolio of vision-language-action models for level 4 autonomous vehicle development, with a first 10-billion parameter model that outputs driving trajectories and reasoning traces; Mercedes‑Benz is set to launch Alpamayo in the all-new CLA and Nvidia expanded its DRIVE Hyperion ecosystem to include tier‑1 suppliers and sensor partners for real‑time perception and in‑vehicle deployment. - Siemens industrial partnership: Nvidia and Siemens announced an expanded collaboration to build an industrial AI operating system, with Siemens planning the first fully AI‑driven adaptive manufacturing site in 2026 at its Erlangen electronics factory; the partnership includes GPU acceleration for Siemens simulation tools using CUDA‑X and PhysicsNeMo, targeting 2–10× speedups in some verification and layout workflows. - DGX Spark desktop system: Nvidia reported software and performance optimizations for DGX Spark that can deliver up to 2.6× faster large language model performance across several libraries; the compact system is described as offering 1 petaflop of FP4 compute, 128 GB of unified CPU‑GPU memory and the ability to run models up to 200 billion parameters locally. - DLSS 4.5 and G‑SYNC Pulsar: Nvidia introduced DLSS 4.5 with a second‑generation transformer for super resolution and dynamic multi‑frame generation supporting up to six‑times frame multiplication, and launched G‑SYNC Pulsar monitors from Acer, AOC, ASUS and MSI (27‑inch 1440p IPS models starting at $599). Summary: Nvidia's five platform announcements span data center inference, autonomous vehicles, industrial manufacturing, desktop development and gaming, and reflect a strategic emphasis on integrated systems rather than standalone chips. Rubin is reported as entering production with partner rollouts expected in the second half of 2026, Siemens plans an AI‑driven factory deployment in 2026, and several software and hardware products were presented as part of a coordinated enterprise roadmap. Undetermined at this time: the broader market adoption timeline beyond the stated partner and pilot deployments.