CES 2026  Report

On January 6, 2026,at the opening keynote of CES 2026 in the early hours of Beijing time, NVIDIA founder and CEO Jensen Huang unveiled to the world the strategic blueprint of the chip giant’s transformation from a “GPU provider” to a “full-stack AI systems architect.” The presentation revealed not only NVIDIA’s deep advancements in frontier fields like physical AI, robotics, and autonomous driving but also signaled its unprecedented commitment to building next-generation AI infrastructure through an open ecosystem and hardware innovation.、

Signal of Transformation: From GPU to “Full-Stack AI Architect”

“We are no longer merely a provider of computing hardware, but an AI systems architect integrating computing, networking, storage, and software orchestration.” Huang’s statement echoed industry analysts’ views. Creative Strategists noted that even as hyperscalers shift towards custom silicon, the tightly integrated platform advantage makes NVIDIA “increasingly difficult to replace.”

In robotics, NVIDIA released a suite of new foundational models, simulation tools, and edge hardware, all open-sourced on Hugging Face. Data from the platform shows robotics has become the fastest-growing AI application category, with NVIDIA’s related models leading in downloads, indicating the early success of its “Android-like” ecosystem strategy.

1. Product Breakthroughs: Three Hardware Innovations Redefining AI Boundaries
2. Autonomous Driving: Debut of End-to-End Inference Model Alpamayo
   The “inference” model Alpamayo, designed specifically for L4 autonomous driving, was officially launched. Optimized on the DRIVE Hyperion platform, it supports real-time decision-making and response to unexpected road conditions. Huang revealed that based on collaboration with Mercedes-Benz, the first vehicles equipped with this technology will hit U.S. roads in Q1 2026.
3. Storage Revolution:BlueField-4 Reshapes AI Memory Architecture
   The newly launched BlueField-4 Data Processor significantly expands GPU memory capacity to meet the needs of long-context, multi-turn AI reasoning, with energy efficiency improved by up to 5x compared to traditional solutions. “AI is evolving from single-turn conversations to collaborative systems with long-term memory and physical understanding,” Huang emphasized. “BlueField-4 is built for this.”
4. Desktop Supercomputing: DGX Series Drives Localized AI Development
   Two desktop AI supercomputers for developers – the DGX Spark (supporting 10-billion parameter models) and DGX Station (supporting 1-trillion parameter models) – were officially unveiled. Based on the Grace Blackwell architecture, they enable local development and seamless cloud scaling for major models like DeepSeek, Llama, and GPT.

Chip Leap: Vera Rubin in Full Production, Cost Reduced to One-Tenth of Blackwell

Huang announced that the next-generation AI supercomputing chip platform, Vera Rubin, has entered full production and is scheduled for customer delivery within 2026. The platform reduces the cost of running AI models to one-tenth of the current flagship Blackwell and requires approximately 75% fewer chips to train models of similar scale.

“The pace of Moore’s Law has slowed significantly. Relying solely on transistor growth cannot meet AI computing demands,” Huang stated. “Only through full-stack redesign and radical engineering can we achieve breakthroughs simultaneously across chips, networking, and software.”

[Simultaneous Focus] Intel Launches 3rd Gen Core Ultra: AI PC Enters the 18A Era

On the same stage, Intel CEO Lip Bu Tan introduced the 3rd Gen Intel Core Ultra processor – the first AI PC platform built on the Intel 18A process node. It will power over 200 device designs globally, becoming the company’s broadest AI PC lineup ever.

The performance of the Panther Lake architecture, designed for gaming handhelds, was particularly notable: the integrated Arc B390 graphics delivered a up to 77% gaming performance increase over the previous generation, surpassing comparable AMD products and the NVIDIA RTX 4050 mobile GPU. Intel simultaneously announced partnerships with MSI, Acer, Microsoft, and others to build a “handheld ecosystem” to drive the large-scale adoption of high-performance handhelds.

The flagship 3rd Gen Core Ultra model features a 16-core CPU, 12 Xe-core graphics, and 50 TOPS NPU performance, delivering a 60% multi-threaded performance boost and support for up to 27 hours of battery life. Versions for edge computing have been launched simultaneously. The first laptops featuring these processors will be available for pre-order starting January 6, with global availability beginning January 27.

Industry Insight: Two Giants Diverge in Strategy, AI Race Enters “Full-Stack Competition & Cooperation” Era

NVIDIA is deepening vertical integration from chips to robotics and autonomous driving, while Intel is expanding horizontal scenario coverage from PCs to handhelds and edge computing. The former lowers development barriers with an open model ecosystem; the latter enhances end-user experience with process node advances and efficiency gains. Despite differing paths, both are accelerating toward the same goal: making AI omnipresent and accessible.

About the Consumer Electronics Show (CES)
Organized by the Consumer Technology Association (CTA), CES is one of the world’s most influential technology events. Coinciding with CTA’s 100th anniversary in 2026, this year’s CES focuses on four key trends: Artificial Intelligence, Sustainable Technology, Digital Health, and Immersive Entertainment, bringing together global tech leaders and innovators.