Meshy Hits $100M ARR, Growing 100x in Under Two Years

After GPT-6 Astra's release, the most-watched question in the 3D world became delicate: now that general models can model, do we still need specialized AI 3D models? The question was raised because Astra demonstrated the ability to build houses in Blender and then import them into Unreal Engine 5 to form walkable scenes, and some people on social media even used Blender MCP to make it build a hospital corridor from a text prompt alone.

But between the impressiveness of a demo and the usability of an asset, there is a crack that only specific tasks can expose. When the same blonde swordsman reference image was given to Astra and Meshy respectively, Meshy's output had hair strands clearly separated, facial features distinct, and the cloak, belt and armor retained obvious layering, with the leaf patterns on the shield also restored; Astra's result was more abstract and cartoonish, with hair like a single helmet block, a round-eyed cartoon face, and a body that looked like several geometric pieces stuck together.

This comparison is not enough to dismiss general models, but it shows that being able to operate Blender, turn requirements into scripts, and judge object structure and position is a different capability from learning from a large number of real assets what shapes and details should look like. The former excels at tasks that can be broken down into geometric relationships; the latter is what characters and complex props—objects with higher demands on proportions, structure and detail—actually need.

According to Meshy's disclosure, its ARR grew from $1 million to $100 million in under two years. ARR, or annual recurring revenue, is the current sustainable, repeatable subscription or contract revenue of a business annualized; one-time revenue is usually not counted. It measures not how many people are watching, but how large a scale of continuous paying revenue has already formed.

This speed needs a reference frame to be understood. Bessemer Venture Partners summarized in "The State of AI 2025" that nova-type AI companies reach $100 million ARR in about 4 years on average and maintain roughly 60% gross margins, while supernova-type companies reach $100 million ARR in about 1.5 years on average, but such companies are rarer. Judging by the time scale alone, Meshy's 100x ARR expansion in under two years already shows explosive power close to supernova-style growth.

Take representative companies of the AI era as further reference: ElevenLabs, a voice generation company valued at $11 billion, took about 20 months from its first product launch to $100 million ARR; AI video company HeyGen took 29 months to grow from $1 million ARR to $100 million. Meshy's under-two-year time is faster than HeyGen's. And 3D happens to be the hardest of these modalities to turn into a business—generated sound and video are basically usable right away, while 3D assets still have to pass through topology, texturing and dimensions before they can enter a game engine or go onto a printer.

It must be emphasized that the above ARR figures, benchmark scores, customer lists and efficiency data all come from Meshy's disclosure or the article's retelling, and lack independent third-party verification. The article does not explain Meshy's ARR statistical basis, whether one-time revenue is included, or the specific cutoff date.