Runway built its name in the creative industry with AI tools that generate and edit visuals. Now the New York-based company sees a broader path for the same core technology: helping robotics and self-driving car companies train systems through simulation.
The shift is not a full departure from entertainment. Instead, Runway is treating its visual-generating world models as a foundation that can serve more than one market.
Why robotics is entering Runway's roadmap
Runway has spent the past seven years developing tools for visual generation. Its technology includes video and photo generation AI world models, described as large language models that create a simulated version of the real world.
The company released Gen-4, its video-generating model, in March. It followed with Runway Aleph, its video-editing model, in July.
As those models became more realistic, Runway started hearing from companies outside its original creative customer base. Anastasis Germanidis, Runway co-founder and CTO, told TechCrunch that robotics and self-driving car companies began reaching out because they saw potential in the company's ability to simulate the world.
Germanidis said the company did not initially imagine robotics and self-driving cars as a target market when Runway launched back in 2018. The opportunity became clearer after companies in robotics and other industries contacted Runway about broader uses for its models.
What simulation changes for training
According to Germanidis, robotics companies are using Runway's technology for training simulations. The logic is straightforward: training robots and self-driving cars in real-world settings can be expensive, slow, and difficult to scale.
Runway is not claiming that simulated training can replace real-world training. Germanidis made clear that the company does not see that as the goal. The value, as he described it, is in giving companies another way to run tests that would be hard to repeat cleanly in the physical world.
Simulation can help teams isolate specific variables. In a real-world scenario, changing one action while keeping every other part of the environment identical is difficult. In a model-based simulation, companies can test a particular decision or movement while holding the rest of the situation steady.
That matters for robotics and self-driving car development because these systems must interact with the real world. If a company wants to understand how a robot or car might behave under a narrow set of conditions, a simulation can make that testing more controlled and repeatable.
The business case for Runway
Runway's interest in robotics also reflects a wider view of what AI world models can become. The company is still focused on entertainment, which Germanidis described as a growing and large area for Runway. But he also said the ability to simulate the world is useful beyond that market.
For Runway, the revenue opportunity is not limited to releasing consumer-facing creative tools. If robotics and self-driving car companies use its models for simulation, the same technical foundation could support a different class of customers.
The company does not expect to launch a completely separate line of models for robotics and self-driving car customers. Germanidis said Runway plans to fine-tune its existing models to better serve those industries. It is also building a dedicated robotics team.
That approach suggests Runway sees robotics as an extension of its core model work rather than a separate product universe. The company is adapting its technology to a new use case while keeping the same broad premise: better representations of the world can support different kinds of work.
Investors and competition
Germanidis said robotics and self-driving cars were not part of Runway's original investor pitch. Even so, he said investors support the expansion.
Runway has raised more than $500 million from investors including Nvidia, Google, and General Atlantic. The company has a $3 billion valuation.
Runway is not alone in connecting world models with robotics. Nvidia released the latest version of its Cosmos world models, along with other robot training infrastructure, earlier this month.
The presence of another major company in the same area shows that simulation is becoming an important theme for robotics training. For Runway, the question is whether technology first built for visual generation can become useful infrastructure for machines that need to act in physical environments.
A broader definition of generative AI
The most important part of Runway's move may be how it frames generative AI. The company is not presenting its models only as tools for making media. It is presenting them as systems for building increasingly useful simulations of the world.
Germanidis described Runway as being built around a principle rather than a single market. That principle is simulation: creating better representations of the world and then applying those models across industries.
For creative professionals, that has meant generating and editing images and video. For robotics and self-driving car companies, it could mean testing actions, outcomes, and edge cases in controlled simulated environments.
The expansion does not erase the need for real-world data or real-world testing. But it gives Runway a reason to look beyond the creative industry as its models improve. If world models can make simulation more specific, scalable, and cost-effective, robotics may become one of the places where visual AI finds a larger commercial role.