World models are drawing surging attention and capital across the technology industry, fueled by expectations that artificial intelligence is poised to move beyond screens and into the physical world of cars, robots, and other machines. A wave of startups is now building these systems, which are typically trained on video data rather than the text-based large language models behind products such as OpenAI's GPT and Anthropic's Claude.

Momentum has accelerated in recent weeks. AMD announced plans to acquire World Labs, the spatial intelligence startup co-founded by AI researcher Fei-Fei Li, for $8.2 billion in stock, and video AI startup Runway, best known for its generative video tools, unveiled the first open-weight version of its world model.

Just as generative AI spawned an entire industry to train and run large language models, the rise of world models is catalyzing a new generation of infrastructure companies. On Monday, San Francisco-based Reactor announced that Nvidia's venture capital arm, NVentures, and Sapphire Ventures have joined its cap table, extending a Series A round first disclosed in May and lifting total funding to $74 million.

Reactor, which emerged from stealth in May, builds software designed to run world models faster and more cheaply, and sells a cloud service that manages the underlying computing capacity.

"Once the model exists, how do you actually run it?" co-founder CEO Alberto Taiuti told Fortune, describing the question that prompted him and co-founder Bryce Schmidtchen to start the company. Both previously worked on Apple's Vision Pro, and Taiuti was also co-founder and chief technology officer of AI video generator startup Luma AI.

According to Reactor, Hollywood studios, advertising platforms, and streaming services currently run active projects on its platform. But the company views robotics as a major area of focus, and whether that focus broadens its customer base is a development to watch as the new capital is deployed.

Building conventional simulations for robot testing is time-consuming and costly, Taiuti said, requiring developers to create three-dimensional objects and configure scenes manually. World models could cut much of that work by generating environments directly from prompts, he said. Reactor says it can run world models that respond to robot actions, allowing developers to test their software inside those generated environments.

Hosting the models in the cloud gives developers access to greater computing power without adding hardware to each individual robot, Taiuti said. It also lets operators upgrade the computers running models without replacing equipment across an entire fleet. He argued the company can adapt technology developed for interactive media to robotics because "similar models and similar workloads can also be run on the same core engine."

Reactor plans to use the new capital to secure additional computing capacity, expand its team, and buy robot hardware to test its systems, Taiuti said.

For Nvidia, the investment through NVentures adds a stake in the infrastructure layer for running interactive models. The chipmaker has developed its own Cosmos world models, alongside Isaac tools for robot simulation and Jetson computers that run AI on robots — a stack it groups under its broader "physical AI" push. Nvidia has used NVentures to assemble a broad portfolio of AI investments in recent years, spanning cloud providers and model developers, making Reactor a bet on the emerging run-time layer of that ecosystem.

This story was originally featured on Fortune.com.