Meshy's nearly $400 million Series B is less about a flashy AI demo than a plain commercial test: can text-to-3D become part of the daily production stack for games, 3D printing, product design, and a dozen other workflows nobody has listed yet?
A minute and about a dollar. That is the promise Meshy is now selling to game teams, product designers, 3D printing users, and anyone else who used to wait days for a single usable asset. On July 21, 2026, the company said in a PR Newswire release that it had raised nearly $400 million in a Series B round at a $1.5 billion valuation, its first publicly disclosed valuation and what it called the largest funding round to date for a company built specifically for AI 3D.
That is a big claim. It also has numbers under it. Meshy says its annual recurring revenue is growing about 12x year over year, with more than 12 million registered users and over 100 million models created. At the Game Developers Conference in San Francisco in March, the company said it had reached $30 million in ARR after doubling in three months and had crossed 10 million global users.
Those figures tell you why investors came in now. Text-to-3D has spent a long time looking like a demo category, clever enough to share, not reliable enough to build around. Meshy is trying to move it from the feed into the workflow.
The bet is on production, not novelty
Meshy's latest product push is the Meshy 3D Agent, which the company describes as an AI agent for 3D creation. The tool can start from a conversation, a line of text, a photo, or a sketch, then produce a print-ready 3D model in formats including FBX, OBJ, GLB, and STL. It also answers 3D questions through built-in Q&A and exports for Unity, Unreal, Blender, and the rest of the standard pipeline.
No Blender skills required. That is the point.
If you run a small game studio, that changes the calculation quickly. A skilled 3D artist can spend days or weeks on a complex asset, especially once geometry, texture, topology, and engine compatibility all matter. Meshy's own help center currently lists the Pro plan at $20 a month with 1,000 monthly credits, while its credit rules say a full Meshy 6 text-to-3D or image-to-3D generation costs 30 credits when model and texture stages are included. That is roughly 33 full generations before retries, extra credits, or heavier work come into play.
The old draft said 50 full models a month. That was too clean. The current credit table makes the picture a little messier, which is usually where the truth lives.
Meshy also says the 3D Agent reaches a slicer success rate of up to 97% for 3D printing. Keep the phrase "up to" in your head. It matters. A printable model is not just a nice-looking object on screen; it has to survive slicing, part separation, surface repair, and the ordinary ugliness of a physical printer doing exactly what the file tells it to do.
That is why Auto Split is more than a side feature. Meshy says it can split a generated model into printable parts, repair surfaces, place joins along less visible boundaries, and arrange parts on the print bed. For a hobbyist, that saves irritation. For a business trying to test physical designs before tooling, it saves time.
The round shows where AI tools are moving
Dealroom reported that IDG Capital, Matrix Partners China, and Monolith Management led the round, with existing shareholders Granite Asia, HongShan, BAI Capital, and Source Code Capital also participating. Meshy's own release was less specific about investor names, saying the round included global investors and all existing backers. That distinction is worth keeping, because attribution is not decoration. You should know who reported what.
The market logic is straightforward. Generative AI has already pushed through text and image creation. Video is still uneven, but Sora and similar tools have made it serious enough for every media company to watch. 3D is harder because the output has to hold together from more than one angle. A sword, chair, shoe, or robot cannot just look right in one frame. It has to have usable geometry.
Competitors such as Tripo AI and Rodin from Deemos are chasing the same problem. Meshy's advantage, for now, is distribution. Twelve million registered users give it more than a vanity metric; they give it a stream of prompts, failures, retries, downloaded assets, and use cases. Pure compute spending can buy a lot. It cannot instantly buy that history.
Gaming is the obvious beachhead. Meshy Labs, the experimental incubator the company announced at GDC 2026, introduced Black Box: Infinite Arsenal, a title built through the platform. The company also says its customers and partners include Nexon, NetEase Games, 37 Interactive Entertainment, Bambu Lab, Creality, Elegoo, FlashForge, xTool, Hugo Boss, and Sweden's museum of art and design.
That spread matters because 3D is not one market. It's a set of bottlenecks scattered across games, manufacturing, retail visuals, museums, architecture, and 3D printing. If you're a founder building in this space, do not just ask whether the model looks impressive in a launch video. Ask whether a paying customer can take the file, drop it into their existing tools, and ship without hiring another specialist for the next iteration.
Also read: The White House says Moonshot AI trained Kimi K3 on banned Nvidia chips routed through Thailand • The US Army burned through its AI token budget in months and every enterprise should take note • Glow raises $180 million at a $1.2 billion valuation to secure the AI agents that CrowdStrike wasn't built to see
Meshy says the proceeds will go mainly toward research and development and global market expansion. Fair enough. The real test is more practical: whether the next 100 million models are assets people actually use, not files they generate once and forget.
