Nvidia adds generative AI and Unity to Omniverse

Nvidia just brought generative AI and the Unity engine directly into its Omniverse simulation platform, alongside a push to put the tool on millions of creators' laptops.

Announced ahead of CES 2023, the update turns Nvidia's heavyweight 3D collaboration suite into something much more accessible, while signing up high-profile enterprise customers like Mercedes-Benz, Dentsu, and Zaha Hadid Architects.

A factory built twice

The marquee showcase this time around comes from the automotive world. Mercedes-Benz revealed plans to build a complete digital twin of a German vehicle plant inside Omniverse before breaking ground in the real world.

Engineers will simulate and optimize the production floor virtually to catch layout flaws early. Once the physical facility is built based on that simulation, live sensor data will feed back into the digital twin to refine operations on the fly.

Behind the scenes, Nvidia is doubling down on Pixar's open-source Universal Scene Description (USD) format, allowing teams using completely different software suites to work together without messy file conversions. To prove the point, Nvidia demoed six artists scattered across the globe tweaking a complex 3D scene simultaneously in real time.

On the hardware front, the platform now hooks into Nvidia's Ada Lovelace architecture. Tapping third-gen RTX tech and DLSS 3, enterprise systems running chips like the RTX 6000 Ada, L40, and OVX can see up to a 3x leap in rendering performance.

AI search and Studio laptops

The update also takes dead aim at workflow friction. A standout feature is Omniverse DeepSearch, an AI service that is now generally available. Instead of digging through messy, untagged 3D backlots, artists can find models using plain English prompts or a rough 2D reference image.

Ecosystem bridges have widened as well. Beyond new ties to Unity, Nvidia added connectors for Adobe Substance 3D Painter, Autodesk Alias, PTC Creo, Kitware Paraview, and export support for Siemens NX.

Perhaps the biggest surprise is distribution. Nvidia Omniverse VP Richard Kerris confirmed that the Omniverse Launcher will come preinstalled on new Nvidia Studio laptops. That puts the simulation stack one click away for millions of individual creators, backed by Omniverse Cloud for users in over 100 countries on standard machines.

Why it matters

While the hype around cartoon metaverses and virtual land has mostly evaporated, Nvidia is showing where virtual worlds actually pay the bills: industrial engineering and high-stakes simulation.

For enterprise teams, the return on investment is dead simple. Running autonomous driving edge cases or testing robotic assembly lines virtually costs pennies compared to physical prototypes or retrofitting a multi-million-dollar factory floor. DeepSearch alone solves a massive time sink for digital artists managing decades of disorganized assets.

We think Nvidia's platform playbook here is brilliant, if predictable. By adopting the open USD standard and throwing in cloud access, Nvidia looks like a benevolent ecosystem unifier. But in reality, the heavier and more accurate these physics-based simulations become, the more Ada-generation GPUs and OVX enterprise nodes customers will inevitably need to buy.

The real test will be creator adoption. Preinstalling Omniverse on RTX laptops sounds great on paper, but complex industrial simulation isn't something an indie artist picks up overnight. Unless Nvidia can make the daily user experience as frictionless as standard game engines, Omniverse risks remaining an elite enterprise playground.