NUPIAO reporter |
NUPIAO editor | Wen Shuqi
The 3D generation world just got another bombshell funding announcement.
On June 23, Yingmo Technology, the company behind the hot new 3D generation models, announced they’ve closed a new funding round worth hundreds of millions of RMB. Sources tell us that Kaifu Fund and Shanghai Guotou XianDao led the round, with existing investors doubling down, while Guangyuan Capital served as the exclusive financial advisor.
You might not have heard much about Yingmo Technology until now—that’s because they haven’t been chasing “casual” users. But in the niche world of 3D generation, this young team has been a veteran player for years.
Founder and CEO Wu Di started his journey at ShanghaiTech University back in 2015. By 2016, he was deep in Prof. Yu Jingyi’s lab focusing on computational photography and graphics, eventually leading the charge to build the university’s largest MARS Lab (AI & Graphics). When Yingmo officially launched in 2020, they were among the very first teams to define the underlying framework for 3D generation. Interestingly, Shanghai Guotou XianDao is once again showing up big in this latest investment lineup.
Earlier this year at CES, Nvidia CEO Jensen Huang showed off a Keynote workflow that relied heavily on Yingmo’s Hyper3D Rodin for the 3D asset generation part. That moment really put Yingmo under the global spotlight.
But here’s the real kicker: Yingmo has hit a crucial inflection point for product commercialization.
On January 28, 2026, they rolled out Rodin Gen-2 Edit, effectively merging “3D generation” with “3D editing.” Their flagship product, Hyper3D, runs on the powerful Rodin large model engine. It’s built specifically for real industrial workflows—delivering high-quality, controllable, production-grade 3D assets. Simply put, hop onto the Hyper3D.AI platform, drop in an image or type a prompt, and boom—you get a 3D model.
Coinciding with the funding news, Yingmo also unveiled their latest Gen-2.5 model. They claim it’s the world’s first 3D model capable of generating tens of millions of polygon details. For the first time, they’ve injected that “think-first, then generate” logic from Large Language Models into the 3D space. The precision can reach tens of millions of polygons, and they can whip out a million-polygon model in just 4 seconds. Zhang Qixuan, Co-founder and CTO, told NUPIAO and other media outlets with confidence that their model adapts better than any other tool on the market across different use cases.
As AI technology races forward, the 3D generation track is heating up globally. Startups like World Labs and VAST have all secured huge investments this year. But for Yingmo, turning early tech breakthroughs into a viable business model took some serious grinding.
Back in 2022, we interviewed Wu Di when the “Metaverse” buzz was at its peak. At that time, Yingmo was locked in on the digital human赛道, using their Dome Light Field products for high-precision face scanning. That system was the only sub-micron level facial scanning setup in Asia—a core technology the founders bet everything on early on. When the Metaverse hype cooled down, Yingmo pivoted back to traditional graphics industries like gaming and VFX, expanding their业务范围 from just face generation to full-spectrum 3D assets.
That pivot wasn’t easy. Once they stopped focusing solely on digital humans, they faced a life-or-death technical decision in 2023. Most 3D generation tools on the market were following the mainstream “2D-to-3D” route, which had obvious quality ceilings. With only a few million RMB left in the bank, the team had to choose: stick to the old path to cash in quickly, or gamble on a new road? The new road promised better quality, but who knew if results would show up anytime soon?
They chose the native 3D path. They innovated at the algorithm and framework foundation levels to solve some stubborn industry pain points at the root. In 2024, Yingmo released their native 3D large model framework, CLAY, which was even nominated for Best Paper at SIGGRAPH, the top conference in computer graphics. Building on CLAY, they launched their native 3D generation product, Hyper3D.AI Rodin, mid-2024.
Wu Di explained that what drove this decision was a core belief in the team: research must meet “Production-Ready” standards. This term comes from the movie VFX industry, meaning “ready to use straight out of the box.” In the early days, the team spent nearly a year aligning data from dome light field scans for film companies before it could fit into the industry’s production pipeline. The lesson learned? “No matter how cool your tech is, if the client can’t use it, it equals zero.”
Their main customers are top-tier game studios, many of whom operate overseas. Zhang Qixuan believes gaming’s inherent real-time interactivity makes it a perfect match for 3D generation. While 3D generation won’t fully replace human artists for high-end games, the models are now deeply embedded in the design pipelines. He gave some examples: some studios use Hyper3D.AI to render front views first, then generate consistent, ready-to-process orthographic views and models. Others generate “white box” versions that better match the original concept art before outsourcing them as references.
In this segment, Yingmo has already grabbed a significant chunk of the market. The company revealed that major domestic and international game studios have all purchased Hyper3D.AI accounts, with some opting for private deployments.
Currently, Yingmo has achieved tens of millions of dollars in ARR (Annual Recurring Revenue). Since the launch of Hyper3D Rodin Gen-2.5, C-end user share has jumped to 60%. In the first month after the new model dropped, both subscription growth and ARR growth exceeded 400% month-over-month.
Looking at the broader industry, B-end (enterprise) clients will remain Yingmo’s primary battleground. Wu Di estimates that most AI 3D users right now are actually professionals with specific modeling needs—a “prosumer” crowd. “Our positioning feels a bit like Anthropic,” he noted. “B-end business is our key growth direction.” Today, they serve clients in e-commerce, gaming, film, and industrial design. Additionally, 3D generation models can fill gaps in simulation and training data for embodied intelligence, opening up plenty of collaboration opportunities.
In 2025, Yingmo secured two massive funding rounds. The first was led by ByteDance and Meituan Dragon Ball, with Sequoia Capital and Mirach Academy joining in. Then in August 2025, they closed another multi-million dollar round led by BlueRun Ventures, with old shareholders like ByteDance and Sequoia China Seed Fund continuing to invest.
Having big tech giants in the shareholder camp proves these giants are watching closely. In the AI application space, big tech’s strategy is usually “spray and pray,” and 3D generation is no exception. Last year, Tencent open-sourced their Hunyuan 3D World Model.
When asked about competition from big tech, Zhang Qixuan offered a potentially friendlier take for startups: 3D generation isn’t on the main path to AGI (Artificial General Intelligence). While it might be influenced by AGI trends, it’s not something language model development naturally covers. So, compared to image generation, this 3D track might actually be where startups can shine brighter.
Focusing on the startup race, everyone faces the same hurdles: data and model capabilities. The next step is commercialization. Zhang explains that while latecomers can catch up on raw 3D generation quality given enough time, what’s truly hard to replicate are the capabilities triggered by specific applications that require model-side adaptations. He believes the real core competency lies in bottom-up architectural innovation.
Yingmo is a classic AI-native startup—mostly Gen Z researchers, with a total headcount of just around 60 including interns. Now that this funding round is done, they’ll ramp up market expansion and push their product capabilities into new areas. Wu Di said they plan to accelerate commercialization full speed ahead, “taking advantage of our current model lead to sprint out there and grab users.”
