Canva's Three Layer AI Stack

This reveals Canva is trying to own the whole AI creation stack, not just add a chatbot to a design editor. The model layer gives Canva its own design tuned generation engine through Leonardo and Phoenix. The app layer turns that engine into everyday tools for writing, image making, and video creation inside the editor. The ecosystem layer fills product gaps fast with outside apps, so Canva can offer specialist features without building every tool itself.1 2 3 4 5

• The foundation model matters because Canva is generating editable design files, not just flat images. Its design model and Leonardo acquisition push AI output directly into templates, layouts, and brand assets that teams can keep changing inside Canva, which is much closer to replacing real creative work than a one off image generator.2 3 4
• The native product layer is where monetization happens. Canva had reached $2.55B ARR by September 2024, then $4B ARR by the end of 2025, while folding AI into docs, presentations, websites, spreadsheets, and code generation inside Visual Suite 2.0. That lets one subscription cover more daily workflows and supports higher team pricing.2 3 5
• The apps layer is Canva's speed advantage against both Adobe and AI point tools. Canva documents hundreds of marketplace apps and separate third party AI allowances, while category research shows Canva using partners like HeyGen, D-ID, and VEED to cover avatars, voice, and video niches that would be slow to build in house.6 7 8

Going forward, this stack should push Canva further from being a lightweight graphics tool toward being the system where marketing teams research, generate, edit, approve, and publish visual work. As foundation models get cheaper and more interchangeable, control of the workflow, the template graph, and the distribution inside a 200M plus user product becomes the more durable advantage.3 5 8