ComfyUI turns prompt control into a 500M dollar company
ComfyUI raised 30 million led by Craft Ventures at a 500 million valuation. Before that it was a 19 million Series A from Chemistry Ventures in late 2024, with Cursor Capital and Vercel's Guillermo Rauch on the cap table. The product is a node-based workflow tool that gives creators granular control over image, video and audio from diffusion models. 4 million users. The word ComfyUI artist is now a job title.
What makes this editorially interesting is not the check size. It's the counter-positioning. Every mainstream AI image tool is racing toward single-prompt magic. Midjourney, Sora, Nano Banana β type a sentence, get a result. ComfyUI went the other way: exposing every node, every sampler, every condition. CEO Yoland Yan puts it bluntly. Prompt tools deliver 60 to 80 percent of what you want. The last 20 percent is where work actually happens, and it cannot be refined into existence by writing more words.
This lands on agent street in two places. First, the production pipeline for agent-generated creative is not going to run on prompt boxes. Studios shipping AI-assisted VFX and ads have already made their choice β they use ComfyUI workflows because they need to hit spec, not vibes. Second, a node graph is the most agent-operable surface in creative AI. You can have an agent read a ComfyUI JSON, modify three nodes, re-run, compare. Try doing that with a Midjourney prompt history.
The four-million-user traction and the 25x revenue-multiple-implied valuation both matter. Open source started the project in 2023 and it never lost that vibe, which is exactly why the professional tier keeps getting bigger. Open source plus a production-grade node runtime plus investor support equals the underlayer for creative AI agents, not yet another wrapper.
https://techcrunch.com/2026/04/24/comfyui-hits-500m-valuation-as-creators-seek-more-control-over-ai-generated-media/
← Back to all articles
What makes this editorially interesting is not the check size. It's the counter-positioning. Every mainstream AI image tool is racing toward single-prompt magic. Midjourney, Sora, Nano Banana β type a sentence, get a result. ComfyUI went the other way: exposing every node, every sampler, every condition. CEO Yoland Yan puts it bluntly. Prompt tools deliver 60 to 80 percent of what you want. The last 20 percent is where work actually happens, and it cannot be refined into existence by writing more words.
This lands on agent street in two places. First, the production pipeline for agent-generated creative is not going to run on prompt boxes. Studios shipping AI-assisted VFX and ads have already made their choice β they use ComfyUI workflows because they need to hit spec, not vibes. Second, a node graph is the most agent-operable surface in creative AI. You can have an agent read a ComfyUI JSON, modify three nodes, re-run, compare. Try doing that with a Midjourney prompt history.
The four-million-user traction and the 25x revenue-multiple-implied valuation both matter. Open source started the project in 2023 and it never lost that vibe, which is exactly why the professional tier keeps getting bigger. Open source plus a production-grade node runtime plus investor support equals the underlayer for creative AI agents, not yet another wrapper.
https://techcrunch.com/2026/04/24/comfyui-hits-500m-valuation-as-creators-seek-more-control-over-ai-generated-media/
Comments