Nvidia’s Strategic Expansion into Physical AI: A New Era for Industrial Automation

Nvidia CEO Jensen Huang is aggressively positioning the company to lead the “Physical AI” market by securing critical supply chain partnerships across South Korea. By integrating generative AI into robotics, manufacturing, and autonomous systems, Nvidia is moving beyond pure data center processing to control the hardware foundation of future industrial automation. These strategic moves, which involve major players like SK Hynix, LG, and Doosan, signal a shift toward AI-driven physical infrastructure that could redefine global industrial standards.

Securing the Hardware Backbone: HBM4 and Memory Supply

The core of Nvidia’s physical AI strategy relies on high-bandwidth memory (HBM), the bottleneck for modern AI training and inference. According to official company disclosures, Nvidia has deepened its multi-year partnership with SK Hynix to secure a consistent supply of HBM4 memory.

SK Hynix currently commands a significant share of the global HBM market, and the agreement focuses on the delivery of 12-layer HBM4E DRAM. These components are essential for Nvidia’s upcoming Vera Rubin platform and Jetson Thor robotics architecture. While Nvidia warns that global memory shortages may persist for years, SK Hynix has committed to doubling its production capacity by 2030 to meet this demand, utilizing Nvidia’s CUDA-X and digital twin software to optimize its semiconductor manufacturing processes.

Integrating AI into Robotics and Manufacturing

Nvidia is embedding its software stack directly into the hardware of industrial giants to accelerate the adoption of autonomous robotics. Through partnerships with the LG Group and Doosan, the company is targeting specific verticals within the “Physical AI” landscape:

* LG Electronics: The company is developing a dedicated “AI factory” utilizing Nvidia Cosmos to streamline production. Additionally, LG Innotek will integrate vision-sensing modules into Nvidia-powered robotic systems.
* Doosan Robotics: By integrating Nvidia Isaac Sim and the Jetson Thor platform into its proprietary robot operating system, Doosan aims to launch commercially viable humanoid robots by 2028.
* Energy Infrastructure: Doosan Enerbility is exploring the use of small modular reactors (SMRs) to provide the massive, consistent power required by future AI-heavy data centers.

The Shift Toward Gigawatt-Scale Infrastructure

The rapid deployment of physical AI systems is driving an unprecedented need for cloud and data center capacity. South Korean firms Naver and SK Telecom are building out “gigawatt-scale” AI clouds to support these workloads.

Naver recently announced plans to expand its Gak-Sejong data center to 55 megawatts by 2027, with long-term goals to establish overseas facilities capable of gigawatt-level output. This shift highlights a departure from traditional, smaller-scale data infrastructure toward massive, centralized power and compute hubs that can handle the real-time processing demands of physical AI, such as autonomous factory management and robotic fleet coordination.

Market Challenges and Sovereign AI Concerns

Despite the rapid expansion of Nvidia’s ecosystem, the company faces scrutiny regarding its influence on local markets. Observers have noted that smaller, domestic South Korean AI semiconductor startups—such as FuriosaAI and Rebellions—were notably absent from recent high-level industry receptions hosted by Nvidia.

This exclusion has fueled national debates regarding “Sovereign AI.” Analysts suggest that while South Korea remains a critical hardware partner, there is growing pressure to develop local software and data capabilities to avoid becoming a “data colony” in an AI-dominated landscape. As Nvidia projects its total addressable market for these technologies to reach one trillion euros by 2027, the tension between global hardware integration and local technological autonomy will remain a central theme in the industrial sector.

Key Takeaways

• Physical AI Focus: Nvidia is pivoting from pure software and GPU sales to integrating AI into physical robotics and manufacturing via the Jetson Thor and Isaac Sim platforms.
• Supply Chain Control: The partnership with SK Hynix for HBM4 memory is the foundational move to ensure hardware availability for next-generation AI factories.
• Industrial Partnerships: LG and Doosan are providing the physical hardware and energy solutions necessary to scale AI-driven automation globally.
• Sovereign AI Debate: The dominance of Nvidia’s ecosystem is prompting South Korean firms to weigh the benefits of integration against the need for national data and AI independence.