Introducing Newton: The Future of Robotics Simulation Made Simple

 

Meet Blue: The AI-Powered Robot

At the GTC 2025 AI conference, NVIDIA CEO Jensen Huang introduced Blue, an adorable AI-powered robot developed in collaboration with Disney Research and Google DeepMind. Inspired by Star Wars, Blue made its way onto the stage and engaged in a lively, real-time interaction with Huang.

“Hi Blue!” Huang greeted the robot, showcasing its advanced AI capabilities. Blue is powered by two NVIDIA computers housed within its compact frame, demonstrating how cutting-edge technology can bring robots to life.

“This is how we are going to train robots,” Huang explained, highlighting Blue’s role in showcasing the future of robotics. Blue is a perfect example of how Newton’s simulation technology can be used to create intelligent, interactive robots that feel almost human.

Introducing Newton: The Future of Robotics Simulation Made Simple

Imagine a world where robots can learn, adapt, and interact with their surroundings just like humans do. Sounds like science fiction, right? Well, thanks to Newton, a new open-source physics engine developed by NVIDIA, Google DeepMind, and Disney Research, this future is closer than ever.

Newton is designed to make robotics simulation faster, more accurate, and accessible to everyone—whether you’re a researcher, developer, or just someone curious about the future of robotics. Let’s break it down in simple terms.

---

Why Do We Need Robotics Simulation?

Before robots can be deployed in the real world, they need to be trained and tested. But testing robots in real-life scenarios can be expensive, time-consuming, and sometimes dangerous. That’s where simulation comes in.

Simulation allows developers to:

• Train robots in virtual environments that mimic the real world.
• Test algorithms safely without risking damage to hardware.
• Speed up development by running multiple simulations at once.

However, there’s a catch. Many simulators struggle to perfectly replicate real-world physics, creating a gap known as the “sim-to-real” problem. Newton aims to solve this by offering a more realistic and flexible simulation platform.

---

What Makes Newton Special?

Newton isn’t just another physics engine—it’s a game-changer. Here’s why:

1. It’s Open Source
   Newton is free to use, modify, and share. This means anyone, from big companies to individual developers, can use it to build and test their robots.
2. Powered by NVIDIA GPUs
   Built on NVIDIA Warp, Newton uses the power of NVIDIA GPUs to run simulations at lightning speed. This makes it perfect for training complex AI models and running large-scale experiments.
3. Works with MuJoCo-Warp
   Newton integrates seamlessly with MuJoCo-Warp, a high-performance simulator developed by Google DeepMind. This integration allows developers to achieve incredible speedups—up to 100x faster for tasks like in-hand manipulation.
4. Differentiable Physics
   Newton supports differentiable simulations, a fancy term for its ability to calculate gradients for optimization. This makes it easier to train robots using machine learning techniques.
5. Highly Customizable
   Whether you’re simulating rigid objects, soft materials, or even complex interactions like sand or cloth, Newton can handle it. Developers can also add custom solvers to simulate unique behaviors.
6. Built on OpenUSD
   Newton uses OpenUSD (Universal Scene Description), a framework that makes it easy to create detailed and realistic environments. Think of it as a universal language for describing robots, objects, and their interactions.

---

Real-World Applications

Newton isn’t just for researchers—it’s already being used to create real-world innovations:

• Disney Research is using Newton to develop next-generation entertainment robots, like the Star Wars-inspired BDX droids. These robots are designed to be more expressive and interactive, bringing characters to life in ways we’ve never seen before.
• Google DeepMind is leveraging Newton to advance its robotics research, particularly in areas like humanoid locomotion and dexterous manipulation.

---

A Collaborative Effort

Newton is the result of a unique collaboration between NVIDIA, Google DeepMind, and Disney Research. Together, these organizations are setting a new standard for robotics simulation.

They’re also working on an OpenUSD asset structure for robotics, which will make it easier to share and reuse robotic models and data. This means developers won’t have to start from scratch every time they build a new robot.

---

What’s Next for Newton?

The first version of Newton is expected to be released later this year. In the meantime, developers can explore the technologies behind it:

• NVIDIA Warp: A framework for GPU-accelerated simulations.
• NVIDIA Isaac Lab: An open-source platform for robot learning.

---

Why Should You Care?

Newton isn’t just for robotics experts—it’s for anyone excited about the future of technology. Whether you’re a student, a hobbyist, or a professional developer, Newton offers the tools you need to bring your ideas to life.

So, get ready to dive into the world of robotics simulation. With Newton, the future is in your hands.

For more information, visit the official NVIDIA Robotics page.

 

NVIDIA’s New Blackwell Ultra: Supercharging AI to Think and Solve Problems Like Never Before

 

At its annual GTC 2025 conference, NVIDIA announced Blackwell Ultra—a groundbreaking AI platform designed to power advanced AI reasoning. Here’s a simplified breakdown of what this means and why it matters.

---

What Is Blackwell Ultra?

Blackwell Ultra is NVIDIA’s latest AI platform, acting like a supercharged engine for two key tasks:

1. Training AI: Teaching models using massive datasets.
2. AI Reasoning: Enabling AI to solve problems step-by-step, akin to human logic.

This marks a shift from AI that learns (e.g., recognizing patterns) to AI that thinks (e.g., planning, analyzing).

---

Why Is This Important?

1. Smarter AI Assistants and Robots
Blackwell Ultra supports agentic AI (AI that acts autonomously). Imagine:

• Logistics AI rerouting deliveries during a storm.
• Robots fixing machinery without human input.

2. Faster, More Accurate Responses
The HGX B300 NVL16 system delivers 11x faster inference for large language models (LLMs) like Llama Nemotron Reason, improving chatbots and medical AIs.

3. Cost Efficiency
NVIDIA claims Blackwell Ultra reduces costs while boosting performance, democratizing access to advanced AI.

---

How Does It Work?

The platform includes two key products:

• GB300 NVL72: A rack-scale “AI factory” combining 72 Blackwell GPUs and 36 Grace CPUs for massive computing power.
• HGX B300 NVL16: A compact system optimized for trillion-parameter models.

Both leverage NVIDIA’s Blackwell architecture, offering 4x more memory than the prior Hopper generation.

---

Cool Uses for Blackwell Ultra

• Physical AI: Generate photorealistic 3D worlds with tools like NVIDIA Omniverse to train robots and self-driving cars.
• Cloud AI: Faster, cheaper services on AWS, Google Cloud, and Microsoft Azure.

---

Networking and Software Upgrades

• Spectrum-X 800G Ethernet: NVIDIA’s high-speed networking platform reduces delays for distributed AI.
• Dynamo Software: Open-source inference framework that splits tasks across GPUs to cut costs.

---

When Can You Use It?

Starting late 2025, partners like Dell, HPE, and Lenovo will offer Blackwell Ultra systems. Cloud providers like CoreWeave and Lambda will host instances.

---

Why Should You Care?

• Better Everyday AI: Smarter chatbots, faster translations, reliable self-driving tech.
• Cheaper Innovation: Startups can leverage NVIDIA AI Enterprise for scalable solutions.
• Future Tech: Foundations for AI scientists, advanced robotics, and more.

---

The Bottom Line

Blackwell Ultra isn’t just an upgrade—it’s a leap toward AI that thinks and solves problems. Dive deeper by watching NVIDIA’s GTC 2024 keynote.

For developers, explore tools like CUDA-X libraries and NIM microservices to build next-gen AI.