Nvidia CEO Jensen Huang Unveils New AI Tech in Korea

🔑 Enhanced Key Takeaways

•The Vera Rubin platform is a comprehensive seven-chip, five-rack AI supercomputer architecture specifically engineered for agentic AI workloads, with full production commencing in Q1 2026 and partner availability expected in H2 2026.
•The Jetson Thor robotics platform is a Blackwell-powered edge AI supercomputer delivering 2,070 FP4 TFLOPS of AI compute and 128 GB of memory, designed to accelerate humanoid robot development and physical AI applications, including its integration into the NVIDIA Isaac GR00T reference humanoid robot.
•The RTX Spark AI laptop series features a new superchip combining up to 20 ARM cores (MediaTek Cortex-X925 and Cortex-A725) with an Nvidia Blackwell GPU boasting up to 6,144 cores, connected via NVLink, targeting high-performance, thin, and light Windows laptops for AI, content creation, and gaming, with initial products expected in Fall 2026.
•Nvidia's new AI technology center in South Korea is a strategic R&D facility focused on physical AI, digital twin, and robotics, actively recruiting local engineering talent for collaboration with universities and corporations on projects involving Omniverse, OpenUSD, Cosmos models, and Isaac Sim.
•Jensen Huang's visit to South Korea involved high-level meetings with major conglomerates like SK, LG, Naver, and Hyundai Motor Group, aiming to deepen AI supply chain partnerships, particularly for high-bandwidth memory (HBM) chips, and to integrate Nvidia's physical AI technologies into Korea's manufacturing and robotics industries, building on a previous commitment of over 250,000 GPUs for the nation's AI infrastructure.

📊 Competitor Analysis▸ Show

Feature/Product	Nvidia RTX Spark	Apple MacBook Pro (M5 Max)	Intel/AMD (Data Center CPUs)
Target Market	High-performance thin & light Windows laptops/desktops for AI, creators, gamers	High-performance laptops for creative professionals	Data center server CPUs
CPU Architecture	Up to 20 ARM cores (10 Cortex-X925, 10 Cortex-A725 from MediaTek)	Apple M5 Max (ARM-based custom silicon)	x86 (e.g., Xeon, EPYC)
GPU Architecture	Up to 6,144 Blackwell RTX GPU cores	Integrated GPU with M5 Max	Discrete GPUs (e.g., Nvidia H100/B100) paired with CPUs
Memory	Up to 128GB LPDDR5x unified memory	Up to 128GB unified memory (M5 Max)	DDR5/LPDDR5X, HBM (for GPUs)
Interconnect	NVLink chip-to-chip interconnect	Unified Memory Architecture	PCIe, CXL, NVLink (for GPUs)
AI Performance	Up to 1 Petaflop FP4 AI performance	High AI performance via Neural Engine and GPU	Varies greatly by CPU/GPU combination
Power Draw	45W to 80W range	Efficient power consumption	Varies, typically higher for server CPUs
Operating System	Windows	macOS	Linux, Windows Server
Competitive Stance	Directly challenges MacBook Pro in performance and efficiency for AI/creative tasks	Established leader in creative professional market	Nvidia's Vera CPU aims to displace traditional x86 CPUs in AI server systems by integrating as a core component of Nvidia's unified solution

🛠️ Technical Deep Dive

NVIDIA Vera Rubin Platform:
- Comprises seven co-designed chips across five rack-scale systems: Rubin GPU, Vera CPU, Groq 3 LPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-6 Switch.
- Rubin GPU: Built on TSMC's 3nm process with a dual-die design, featuring 336 billion transistors, 288 GB HBM4 memory with 22 TB/s bandwidth, and delivering 50 PFLOPS NVFP4 inference performance.
- Vera CPU: Features 88 custom Arm-based Olympus cores (Armv9.2) with spatial multi-threading for 176 effective threads, up to 1.5 TB LPDDR5X memory with 1.2 TB/s bandwidth, and 1.8 TB/s coherent bandwidth via NVLink-C2C to GPUs. It is designed for agentic AI orchestration and data processing.
- Vera Rubin NVL72 Rack: Houses 72 Rubin GPUs and 36 Vera CPUs in a single liquid-cooled rack, providing 3.6 EFLOPS of NVFP4 inference and 2.5 EFLOPS of training compute.
NVIDIA Jetson Thor Robotics Platform:
- Powered by NVIDIA's Blackwell architecture, delivering up to 2070 FP4 TFLOPS of AI compute.
- Features a 14-core Arm Neoverse-V3AE 64-bit CPU.
- Equipped with 128 GB 256-bit LPDDR5X memory, offering 273 GB/s bandwidth.
- Power consumption is configurable from 40W to 130W.
- Supports high-speed sensor integration with 4x 25 GbE networking, a camera offload engine, and NVIDIA Holoscan Sensor Bridge.
- Not pin-compatible with previous Jetson Orin modules.
NVIDIA RTX Spark AI Laptop Series:
- A superchip design integrating up to 20 ARM cores (10 Cortex-X925 and 10 Cortex-A725 from MediaTek) and an NVIDIA Blackwell GPU with up to 6,144 CUDA cores and fifth-generation Tensor Cores.
- Utilizes an NVLink chip-to-chip interconnect.
- Configurable with up to 128GB of LPDDR5x unified memory.
- Targets a power draw in the 45W to 80W range.
- Delivers up to 1 Petaflop of FP4 AI performance.

🔮 Future ImplicationsAI analysis grounded in cited sources

Nvidia's integrated hardware and software stack will significantly accelerate the development and deployment of agentic AI and humanoid robotics.

The Vera Rubin platform is explicitly designed for agentic AI workloads, and the Jetson Thor, coupled with the Isaac GR00T open development platform, provides a comprehensive, unified solution for humanoid robot development, from training to real-time inference.

Nvidia's deepened engagement in South Korea will solidify its supply chain for critical AI components like HBM and foster advanced collaboration in physical AI and manufacturing.

Jensen Huang's direct meetings with major Korean conglomerates and the establishment of a dedicated AI technology center indicate a strategic effort to secure HBM supply from key manufacturers and integrate Nvidia's physical AI technologies into Korea's robust manufacturing and robotics industries.

The RTX Spark series will intensify competition in the high-performance laptop market, particularly for AI-centric workloads, posing a direct challenge to established players like Apple.

RTX Spark combines powerful ARM CPUs and Blackwell GPUs with substantial unified memory, targeting thin and light Windows laptops for creators and AI developers, directly positioning itself against Apple's M-series chips in a market segment focused on performance and efficiency.

⏳ Timeline

2025-10

Nvidia announces partnership with South Korea for over 250,000 GPUs and a $3 billion investment with Hyundai Motor Group for a physical AI cluster and AI Technology Center.

2026-01

NVIDIA's Rubin platform enters full production.

2026-03

Detailed specifications and system architecture for the Vera Rubin platform are finalized at GTC 2026.

2026-06-01

NVIDIA unveils the Vera CPU, designed for agentic AI and as the host CPU for Vera Rubin platforms.

2026-06-01

NVIDIA announces the Isaac GR00T Reference Humanoid Robot, built on the Jetson Thor platform.

2026-06-01

CoreWeave announces the industry-first bring-up and validation of NVIDIA Vera Rubin NVL72.

2026-06-05

Nvidia CEO Jensen Huang unveils new AI tech in South Korea, including Vera Rubin, RTX Spark, and Jetson Thor, and confirms the establishment of an AI technology center.

Nvidia CEO Jensen Huang Unveils New AI Tech in Korea

⚡ 30-Second TL;DR

Key Points

🧠 Deep Insight

🔑 Enhanced Key Takeaways

🛠️ Technical Deep Dive

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline

📎 Sources (28)

👉Related Updates

Trina Solar Returns to Profitability Driven by Energy Storage

Chinese robot maker AgiBot pursues Hong Kong IPO

Insta360 Plans 2 Billion RMB Tech Innovation Bond Issuance