Google subsidizes world-model access
World Labs
Google’s edge is not that Genie 3 is simply another world model, it is that world generation can be bundled into a much larger consumer and developer package. When a user already pays for Google AI Ultra, Google can treat interactive world access as one feature among many, alongside Gemini and other creation tools, which lowers the need for Genie to stand alone as a profit center and makes distribution much easier than for a startup selling a single product.
-
Genie 3 is already shipped inside Google’s subscription surface. DeepMind says it generates navigable worlds at 720p and 24 fps, and Google says U.S. Google AI Ultra subscribers can use Project Genie. That means customer acquisition can come from an existing paid bundle, not a fresh standalone sale.
-
This is a different economic model from Runway and World Labs. Runway built a large AI video business as a dedicated product, reaching about $90M in estimated revenue by June 2025, while World Labs has raised about $1.23B to build explicit 3D world creation tools. Google does not need Genie to recover those costs directly from world model pricing.
-
Google also has more places to plug a world model in. DeepMind frames Genie as part of a path toward agents and real world action, and Google positions Project Genie with Gemini in an experimental app. That creates room to connect world generation to video, assistant workflows, and robotics research instead of limiting it to one narrow use case.
Going forward, bundling will push standalone world model companies toward sharper specialization. Startups will need to win on things Google’s bundle does not naturally deliver, like editable 3D assets, deterministic camera control, engine exports, and simulation workflows that fit directly into production pipelines.