← Back
NVIDIA Blog

Korea Advances Physical AI Across Industries With NVIDIA

•4 min read•
#compute#agents#deployment
Level:Advanced
For:AI Engineers
✦TL;DR

Samsung Electronics leveraged NVIDIA’s cuISS, cuDSS, cuEST, cuEquivariance and PhysicsNeMo to accelerate semiconductor material and process simulations, achieving up to 5× faster quantum‑chemical runs and a ≈1,600× speedup (18 h → 41 s) for AI‑homogenized thermal‑mechanical analysis. Kakao Mobility will integrate NVIDIA Hyperion on DRIVE AGX Thor and NVIDIA Halos to build a Level‑4 autonomous‑driving stack, while LG Electronics demonstrated an end‑to‑end physical‑AI workflow using Isaac GR00T, NVIDIA Cosmos, and Jetson AGX Thor for robotic manipulation in a live factory. These collaborations illustrate how NVIDIA’s accelerated‑computing stack can move physical‑AI from research to production‑scale deployment across chips, vehicles, and robots.

⚡ Key Takeaways

  • Samsung achieved a ≈1,600× speedup (18 h → 41 s) for AI‑homogenized thermal/mechanical analysis using GPU‑accelerated coarse‑grid solves.
  • NVIDIA cuISS and cuDSS accelerate sparse linear system solves in semiconductor CAD; cuEST speeds quantum‑chemical calculations 5×; cuEquivariance speeds tensor ops for atomistic AI models.
  • The physical‑AI data flywheel combines real‑world data, NVIDIA Cosmos augmentation, GR00T model training, and Jetson AGX Thor inference for repeatable robot learning.
  • NVIDIA Agent Toolkit is proposed to orchestrate simulation tools and AI models into engineering agents that can evaluate design alternatives automatically.
  • GPU‑based acceleration (e.g., DRIVE AGX Thor, Hyperion) is required; CPU‑only pipelines will see orders‑of‑magnitude slower runtimes.
💡 Why It Matters

By swapping CPU‑bound simulation kernels for NVIDIA’s GPU‑optimized libraries, engineers can shrink design‑cycle times from hours to seconds, making AI‑guided optimization feasible for high‑volume manufacturing, autonomous‑vehicle validation, and factory robotics.

✅ Practical Steps

  1. Replace CPU sparse‑linear solvers in your semiconductor CAD flow with NVIDIA cuISS and cuDSS to gain GPU acceleration.
  2. Integrate NVIDIA cuEST into your quantum‑chemical calculation pipeline to achieve up to 5× faster end‑to‑end runs.
  3. Build a physical‑AI data flywheel by feeding real robot sensor data into NVIDIA Cosmos for augmentation, training an Isaac GR00T model, and deploying inference on Jetson AGX Thor.

Want the full story? Read the original article.

Read on NVIDIA Blog ↗

More like this

Introducing workload recovery and expanded infrastructure checks on Amazon SageMaker HyperPod

AWS ML Blog•#deployment

Introducing Unity Gateway Insights

Databricks Blog•#deployment

Introducing Gemini 4 Neon

Google DeepMind•#llm

DaisyRose-27B: Pushing the Frontier of Post-Training Efficiency

Hugging Face Blog•#llm

EXPLORE AI NEWS

Daily hand-picked stories on LLMs, RAG, agents and production AI — curated for engineers who ship.

BROWSE NEWS

GET THE WEEKLY DIGEST

Join engineers getting the Monday signal-over-noise AI breakdown. No spam, unsubscribe anytime.

LEARN AI ENGINEERING

Curated courses, research papers, repos and tutorials built for engineers leveling up in AI.

START LEARNING