The top three AI 3D players raised over 8 billion yuan in half a year; Meshy's ARR exceeds $100 million

Text | FoST Future Narrative, Author | Congcong

A founder of a leading AI 3D company once explained to us the impact of AI 3D this way:

AI 3D changes the entire relations of production; you can think of it as a phone camera, or an input method—its core significance is enabling the general public to participate in content creation.

Now, his words are increasingly being confirmed.

The leading AI 3D companies are not only competing neck and neck at the model layer, but are also continuously optimizing the experience at the product usage level: from AI 3D editing features to Agent capabilities, products are continuously extending toward a complete creative workflow.

On October 16, leading AI 3D company Meshy released Meshy Edit. Meshy said this is a brand-new way to edit 3D models, capable of adding, removing, or replacing local parts of a model without opening Blender.

Meshy Edit

On October 20, another 3D company VAST released and open-sourced the world model EdenLive, pushing its previously released model Project Eden from the "state" level to the "rendering" level. EdenLive only needs reference-provided structure and motion, and the image can be generated in real time and continuously updated.

EdenLive

Behind the frequent "new releases," the AI 3D industry has become a new money-making tool. As of the end of September, Meshy's ARR had surpassed 100 million USD, and this is just the tip of the iceberg of AI 3D's money-making.

Besides Meshy and VAST, there is also the earlier entrant in AI 3D, Yingmou. These three companies founded by post-90s entrepreneurs are becoming darlings of the capital market, raising over 8 billion RMB from the market within half a year.

In what way will AI 3D disrupt the future content industry?

01 AI 3D startups compete on application scenarios

Compared with text-to-image, AI 3D not only has to generate visual appearance, but also has to handle issues such as three-dimensional geometric structure, textures, and topology. If the goal is to obtain a model that can continue to be edited, rigged, and enter the production pipeline, post-generation repair and optimization will also increase costs.

So, can a 3D model obtained from a single generation be edited, reused, and rendered to achieve longer-term economics and reduce the cost of generating and modifying 3D assets?

This is exactly what VAST, Meshy, and Yingmou have all been doing over the past year or so.

In September, VAST officially released the Tripo P2.0 (P2.0) model. P2.0 introduced smart UV mapping, improved textures, and restructured the skeletal rigging and animation tools, capable of generating native quad meshes with face counts between 500 and 25,000. At the same time, Tripo quietly wrote a line in its official description:

The mesh editing feature allows you to refine specific areas without regenerating the entire model. Pro, Max, and Team users only need to spend 100 credits to generate up to four versions with different face counts in a single run.

More than a month later, Meshy went further by specifically launching Meshy Edit, a brand-new way for Meshy to edit 3D models, in which the "local editing" feature only changes the part you select and keeps the rest of the model unchanged.

Meshy directly gives an example in its official description:

You generated a model, and the result is almost right. The pose is right, the proportions are right, but one part is wrong: there is an extra piece, a missing accessory, or a weapon you want to replace.

Now, you have only two choices: regenerate the entire model and pray that the other parts stay consistent; or export to Blender and fix it manually.

Meshy Edit gives you a third choice: click Edit on the model to enter Edit Mode.

Details that might have taken hours to modify manually in Blender in the past—Meshy Edit hopes to help users complete them within minutes.

Behind this is a change in the thinking behind AI 3D products: model generation is no longer the end point of creation, but a starting point that can be adjusted repeatedly.

In June this year, when Meshy released Agent, it also made a similar statement:

Meshy 3D Agent is an AI 3D model generator you can talk to.

Without using Blender or Maya, users can complete more 3D creation tasks through conversation. Built for 3D printing manufacturers, indie game developers, 3D artists, and designers.

Meshy is not the first company to add editing features to 3D modeling. As early as the beginning of this year, Yingmou Technology announced the Rodin Gen-2 Edit (text/image-to-3D) model, billed as the industry's first commercial product that integrates "3D generation" and "3D editing" into a complete workflow.

Rodin Gen-2 Edit supports generating 3D content directly on the platform and using prompts to make local modifications to the result; it can also import any existing model (including third-party AI-generated models or historical assets) and edit based on the original model.

The founder of Yingmou once said in an interview that there is a huge gap between what is developed in the lab and what can actually be used. To close this gap, the key is that the generated models must meet usage standards and that users can control the 3D generation results.

On September 22, Yingmou Hyper3D officially launched Agentic Mode, integrating Hyper3D Rodin with LLM multimodal context analysis capabilities into the same creative workflow: first understanding and automatically processing on demand the images, dimension drawings, and other materials submitted by users, autonomously determining the most appropriate 3D modeling method, and keeping the post-generation editing and processing within the same pipeline.

Behind VAST, Meshy, and Yingmou all laying out 3D editing features is the need to solve the unavoidable revision process in 3D generation. And when creators can achieve their goals through a platform, the platform not only gains user retention but even has the chance to become users' long-term work entry point.

02 AI 3D, rapid development, "money-making" at the speed of light

Behind AI 3D entering more creators' field of view, a new "money-making" industry is taking shape.

On one hand, AI 3D is attracting more and more capital. Within half a year, VAST, Meshy, and Yingmou attracted over 8 billion yuan.

In September 2026, VAST announced the completion of Series B and B+ financing, totaling 3 billion yuan. Previously, from March to July, VAST conducted multiple rounds of financing, with amounts ranging from 50 million USD to 200 million USD. Roughly calculated, VAST completed over 5 billion yuan in financing within half a year.

Behind VAST stands a young founder from the SenseTime circle: Song Yachen.

Song Yachen was born in 1997 and holds dual bachelor's degrees in International Relations and Economics from Johns Hopkins University in the United States. In 2019, Song Yachen joined SenseTime Technology and began working with AIGC 3D-related technologies.

In 2023, Song Yachen founded VAST. From 2023 to 2025, it gradually grew at a pace of one financing round per year. In 2026, VAST completed cumulative financing of over 5 billion yuan, with a valuation reaching 2 billion USD.

Not only VAST, but Meshy also completed an unprecedented 400 million USD financing in July 2026. It was jointly invested by IDG Capital, Matrix Partners China, and Monolith Management, with existing shareholders including Granite Asia, Sequoia China, BAI Capital, and Source Code Capital following on. The post-investment valuation was 1.5 billion USD, over 10 billion yuan.

Meshy founder Hu Yuanming was born in 1994. His early career can be described as smooth sailing: recommended to Tsinghua's Yao Class, earned a PhD from MIT in three years under graphics authority Frédo Durand.

In 2021, Hu Yuanming returned to China to start a business and founded Taichi Graphics. With the halo of Yao Class and an MIT PhD, Sequoia China invested in the angel round. But Hu Yuanming's early entrepreneurship was not smooth; at one point, due to unclear positioning, going head-to-head with established local software led to a dead end, until the flash of inspiration that became Meshy.

What Meshy wants to build is an AI tool that generates 3D models from text and images. Meshy's initial version was also not smooth: the models were not detailed enough, the rendering looked eerie, and internally it was called "four faces." The first developer to use it was someone making horror games.

Meshy thus completed its cold start.

Since its founding, Meshy has completed only three rounds of financing. Apart from the latest 400 million USD, the previous round was a 36 million USD Series A completed in 2024.

Earlier than Meshy and VAST to enter the field was Yingmou Technology. Founder Wu Di was born in 1997 and was in the second undergraduate cohort of ShanghaiTech University. Yingmou Technology was founded in 2020 and is a startup incubated by the university.

Wu Di speaks at the ShanghaiTech University 10th Anniversary Conference on Science-Education Integration and Innovation Development. Image source: ShanghaiTech University official website

In 2022, Yingmou Technology received tens of millions of yuan in investment from Sequoia China and MiraclePlus. In 2025, Yingmou Technology successively completed Series A and A+ financing. In addition to investment institutions such as Sequoia China, internet giants Meituan and ByteDance also appeared on the investor list. In June 2026, Yingmou Technology completed a Series B financing of several hundred million yuan.

One trend is that since 2025, AI 3D has begun to receive more and more attention. VAST, Meshy, and Yingmou Technology all received their largest investments since founding within the past year or so.

Behind this, the AI 3D industry is running into the commercialization channel. The Quiet Lion Report's August AIGC site payment ranking shows that five AI 3D sites had a combined payment intent increment of over 190,000, ranking second only to AI video in number of entries.

At the end of September 2026, Meshy announced that ARR (annual recurring revenue) exceeded 100 million USD, and the ARR of VAST and Yingmou Technology had also reached tens of millions of USD.

After achieving the 100 million USD milestone, Hu Yuanming wrote on his personal WeChat public account (Hu Yuanming Ethan):

We support paying employees at the highest market salaries. Besides because we believe in "being kind to others" and hope colleagues working hard together can live a little better, another reason is that our business has relatively high gross profit, which supports us in doing this.

03 The "AI 3D Big Three": different business visions, different business paths

VAST founder Song Yachen has mentioned Douyin many times, hoping to build a 3D version of Douyin. He said:

"3D Douyin" is a brand-new content format, more of an ecosystem of "strong interaction + light content," which will allow users to generate and share 3D content.

To realize this vision, the threshold for generating 3D content needs to be low enough. Similarly, Jianying greatly lowered the threshold for video editing and fostered the short video content ecosystem. Song Yachen once revealed to us:

In the next 3 to 5 years, VAST will continuously launch new products, including "3D Jianying" and "3D Douyin."

To realize such a long-term vision, VAST has not only laid out an AI 3D work platform like Tripo Studio, providing models such as Tripo H3.1 and Tripo P1.0/P2.0, but also explored pushing 3D assets toward becoming "the foundation of the next-generation UGC interactive platform and 3D content ecosystem."

On October 20, VAST released open-source EdenLive, attempting to advance 3D asset generation into a real-time, interactive three-dimensional world. VAST said this is the real-time generative renderer of the Project Eden series. Project Eden is a world model released by VAST in May this year; it supports multi-player concurrency and has native persistence.

In terms of ARR, VAST does not stand out among these three companies; the figure announced in August 2025 was $12 million. But in its latest funding round, VAST's post-money valuation has reached $2 billion, surpassing Meshy's ten billion RMB.

According to a report by the STAR Market Daily, VAST's Tripo series models currently have 20 million users. In addition, a March report by 36Kr mentioned that VAST's ecosystem partnerships cover companies such as Tencent, ByteDance, and Bambu Lab as well as more than 90,000 developers, and its Tripo Studio platform has gathered more than 6.5 million creators, who have cumulatively generated nearly 100 million 3D models.

The Fun here is not the shallow entertainment of scrolling short videos, nor is it killing time—experiencing nearly infinite experiences within a finite life can itself be a kind of meaning.

Meshy founder Hu Yuanming believes that when productivity is greatly liberated, people's core question will shift from "how to work more efficiently" to "how to create, express, connect, and gain a sense of meaning."

Meshy calls the form that takes on this task "real-time interactive multimodal content experiences."

Meshy's revenue model is subscription plus pay-per-use: consumer creators subscribe monthly for generation quotas, while B2B game studios and 3D printing customers can also obtain services through API licensing. Its growth engine likewise comes from the two major scenarios of gaming and 3D printing.

According to Meshy's disclosures, it has more than 12 million registered users, and the platform has cumulatively generated more than 100 million models. Meshy is already directly targeting a broader range of creators: a user throws out a prompt or a 2D image, and Meshy can directly spit out a 3D model that can go into a game engine.

Unlike the visions of VAST and Meshy, Yingmou's goal from the start has been to serve professionals. To this end, Yingmou's core founders personally use 3D software such as Blender, edit videos and do rendering themselves, to experience what difficulties may be encountered in the process.

In the view of the Yingmou team, in the next 3-5 years, 3D generation will still mainly be used by professional users. Therefore, Yingmou emphasizes entering enterprise customers and real production workflows with professional-grade 3D generation capabilities. Yingmou Technology CEO Wu Di once said:

3D generation is a foundational capability that gaming, film and television, industrial design, and even the intelligentization of the physical world cannot do without.

We continue to explore scene generation technology, and the research result CAST can already directly benefit fields such as virtual scene experiences, game concept design, and embodied intelligence simulation environment training.

Although VAST, Meshy, and Yingmou have already established certain advantages and are highly favored by capital, beyond startups, many major tech giants are also accelerating their layouts:

In April 2026, ByteDance released Seed 3D 2.0, claiming that it has reached industry-leading levels on both core metrics of geometry generation and texture/material generation.

In the same period, Tencent Hunyuan released and open-sourced the Hunyuan 3D World Model 2.0 (HY-World2.0). The model supports multiple input methods such as text, images, and videos, and can generate a complete 3D space containing people, objects, and scenery with one click.

In August, Tencent launched the 3D open-world generation framework WorldClaw; users input a single open-ended text description, and WorldClaw can generate a large-scale 3D world that can be freely explored and further edited.

The giants have also joined this AI 3D boom, which certainly affirms the post-90s entrepreneurs who bet earlier. But it also means they will face greater challenges.