Jiajia Vision Raises 350 Million RMB in Its B2 Funding Round

A familiar scene plays out once again.

Investors have learned that Jijia Vision has announced the completion of another RMB 1 billion B2 round of financing, led by Singapore’s premier cross-border investment firm,狮城资本 (which has consistently increased its investment across multiple rounds), along with the China-Belgium Fund, JianTou Investment, Wanxiang Qianchao, Fosun RuiZheng, Huagai Chuangying, Jin Chuang Tou, Deyi Capital, Huacang Capital, and Yuan Shi Fund. These investors include top-tier state-backed funds, industrial capital, financial institutions, and state-owned platforms. Existing shareholders such as Guozhong Capital, Dacheng Capital, and Turing Asset Management have also significantly increased their stakes.

According to insiders, the investment interest in this round far exceeded the original funding target. Notably, this is the third financing round for Jíjiā Shìjiè within three months, with a total accumulated amount of RMB 3.5 billion.

Thus, Jijia Shijie, along with its behind-the-scenes leader, 90s PhD holder Huang Guan, has created one of this year’s most sensational moments in the venture capital scene. The collective investment backing may signal that the “GPT-3 moment” for physical AGI is on the horizon.

As you can see, nearly all types of leading investment institutions in the market are behind Jiyi Vision.

From its inception, Jijia Vision secured a seed round of tens of millions of yuan, led by Chen Tao Capital. Since then, investors have been lining up to participate—in September 2024, the company completed two consecutive financing rounds, totaling nearly 50 million yuan, with participation from BAIC Industrial Investment, Jiqi Chuangtan, Huamin Investment, Longding Investment, Qingzhi Capital, PKSHA Algorithm Fund, and others.

A year later, Jijia Shijie secured consecutive Pre-A and Pre-A+ funding rounds totaling hundreds of millions of yuan in August 2025. The Pre-A round was led by Guozhong Capital, with participation from Zifeng Capital and existing shareholder PKSHA Algorithm Fund. The Pre-A+ round was led by CICC Capital, Guangzhou Industrial Investment, Yicun Songling, and Huaqiang Capital.

Subsequently, Jijia Vision's fundraising pace intensified. In November of the same year, Jijia Vision completed a new round of A1 financing amounting to hundreds of millions of yuan, led by Huawei Haho and Huakong Fund. One month later, the company completed an A2 financing round of RMB 200 million, led by Dacheng Capital, with existing shareholder Huakong Fund co-leading the round. Leading institutional investors including Shoufa Development Venture Capital, Puyao Xinye, Caixin Capital, Huajin Capital, Zhangke Yaokun, and Fuzhuo Venture Capital participated as follow-on investors, with existing shareholder Hedgong Capital over-subscribing its follow-on investment.

Entering 2026, Jíjiā Shìjiè’s fundraising pace left a deep impression on the venture capital community.

In early March this year, the company completed a Pre-B round of financing totaling nearly RMB 1 billion, with investors including top-tier chip and automotive industry capital such as SMIC Peak, Shanghai Semiconductor Industry Investment Fund, Linxin Capital, Star Source Capital, and Wanlin International, as well as major state-owned platforms and renowned financial institutions including CICC Capital, Suzhou Venture Capital, Huaqiang Capital, Yangtze River Capital, Wuhan Guanggu Industry Investment, Xishan State-owned Investment, Jinyu Maowu, Xinding Capital, Lingyang Investment, Caixin Capital, Zhangke Yaokun, and Chengzhu Investment. Notably, existing shareholders such as CICC Capital, Huaqiang Capital, Caixin Capital, and Zhangke Yaokun continued to significantly increase their investments.

Shortly thereafter in April, the B1 round of financing for Jijia Shijie came to light, led by a renowned tech giant, multiple top-tier state-backed funds, Yili Group’s CVC Jianling Capital, Puhua Capital, Huafu Investment, Yida Capital, the New Industrialization Fund, Shengjing Jiacheng, Turing Asset Management, Kaiyang Capital, Wuhan High-Tech, Guiyang Jin Tou, Shandong Industrial Investment, and other leading state-owned platforms, industrial capital, and dual-currency financial institutions. Existing shareholders including Huakong Capital, Huamin Investment, Yicun Capital, and Lingyang Investment continued to significantly increase their stakes.

At this point, Jiyi Shijie's valuation has surpassed ten billion yuan, becoming China's first unicorn in the world model sector.

Until now, the Series B round of financing has officially been unveiled, meaning that within just three months, Jiyi Vision has raised a total of RMB 3.5 billion, with investors casting their votes with real capital.

A detailed review of Jijia Shijie’s financing journey since its founding reflects the growing confidence of the primary market in the physical AGI sector, and underscores investors’ strong belief in Jijia Shijie’s approach—driving physical AGI through world models combined with production-grade execution capabilities.

The signals conveyed carry profound implications. Moreover, this not only serves as a testament to Jiajia Vision’s accumulated technological expertise but also stands as the most credible endorsement of its leadership in the physical AGI sector and its ability to reshape the industry landscape. It is foreseeable that even more investors will rally behind Jiajia Vision in the future.

As outsiders wonder, why is it Jiayi Vision?

Investing is about investing in people. Behind Jíjiā Shìjiè is Huang Guan, a 90s-born Ph.D. graduate from Tsinghua University. He earned his bachelor’s degree from Huazhong University of Science and Technology, then pursued his master’s degree at the Institute of Automation, Chinese Academy of Sciences, before completing his Ph.D. in the Department of Automation at Tsinghua University. He has also held positions at Horizon Robotics and Jianzhi Robotics, as well as work experience at Microsoft Research Asia and Samsung China Research Institute.

Even more notably, Huang Guan has led or participated in fundraising totaling over RMB 2 billion throughout his career. As such, Huang Guan is one of the rare industry leaders who combines top-tier research experience in physical AI, large-scale engineering expertise, proven commercialization experience, and a track record of serial entrepreneurship.

Led by Huang Guan, this core team has fully experienced the past decade’s advancements in physical AI, consistently delivering outstanding achievements in technological innovation and industrial application across every stage—including computer vision, autonomous driving, embodied intelligence, and world models. This is a rare “hexagonal warrior” team in the industry, possessing top-tier experience and capabilities in the algorithms, data, ontologies, mass production, business, and organization of physical AGI—truly a “dream team” for physical AGI.

If talent has been the driving force behind Jieyi Vision’s rise, then technological innovation is the foundational pillar that establishes its position in the global physical AGI landscape.

It is well known that physical AGI development faces two core bottlenecks: first, data fragmentation, with a lack of high-quality, multidimensional data tailored to physical interaction scenarios; second, language-dominated foundational models are not effective architectures for encoding 3D information, physical causality, and actions, making it difficult for models to understand complex physical laws.

How can these two major issues be resolved? Jiezhiaishijie's solution is to center on a world model while simultaneously building a dual-pyramid system of algorithms and data.

Among these, the data pyramid consists of five layers, progressing from bottom to top as follows: internet video data, human-generated data, world model simulators, simulated synthetic data, and real-world robot data. This five-layer data architecture addresses the key challenges in physical AGI development—insufficient data, low quality, and limited scenarios—providing ample, high-quality "fuel" for training algorithmic models.

The algorithm pyramid is divided into three layers, primarily centered on three core capabilities: world simulation, action alignment, and experience reinforcement. This enables a leap from physical perception to physical execution, and from passive execution to active evolution, endowing physical AGI with learning and adaptive abilities similar to those of humans.

The core value of the "Double Pyramid" system lies in establishing a closed-loop evolution mechanism driven by data and reinforced by algorithms. The data pyramid provides vast quantities of high-quality physical interaction data to support the training and optimization of algorithmic models; in turn, the iterative advancement of the algorithm pyramid enhances the precision of data collection and the realism of simulated data, thereby enriching the content of the data pyramid.

More importantly, after three years of refinement, Jiexian Shijie has developed a dual-model system: "World Generation-Action." The World Action Model translates the understanding and prediction of the world model into robotic action strategies—GigaBrain-0: a self-developed embodied VLA large model driven by a world model, which achieved a 51.67% task success rate and secured the global championship in RoboChallenge, the world’s largest real-robot evaluation benchmark.

GigaBrain-0.5M*: The world's first native paradigm for physical agents centered on "world model-driven experiential learning," achieving near 100% success rate in high-difficulty, long-duration tasks through the integration of "world models + reinforcement learning."

GigaWorld Policy: A world-action model that breaks the impossible triangle of "speed-performance-efficiency," achieving 10x faster inference, 10x improved training efficiency, and approximately a 30-percentage-point increase in task success rate. It has defeated Nvidia GR00T N1.5, PI0.5, and others to claim the top rank on RoboCasa365, the globally authoritative evaluation platform for home-based mobile manipulation tasks, becoming the first world-action model to reach number one on the leaderboard.

The world generation model understands, simulates, and generates the physical world, providing data and simulation infrastructure along with pre-trained parameters for action models—GigaWorld-0: A landmark achievement, the first to validate that data generated by world models can effectively enhance real-world robot performance; released and open-sourced in December 2025, with over 1.5k stars on GitHub.

GigaWorld-1: An Action-Conditioned World Model (AC-WM) that achieved a composite score of 62.34 on the authoritative benchmark WorldArena, outperforming top international models from Google, NVIDIA, Alibaba, and others to claim the global number-one ranking—it is the first model ever to surpass a score of 60 on this leaderboard.

DriveDreamer: The world's first autonomous driving world model designed for the real physical world, invited for an NVIDIA oral presentation and recognized as one of the most influential papers at ECCV 2024,率先实现世界模型的大规模产业落地.

Undoubtedly, the world action model and the world generation model are both indispensable, forming a synergistic, mutually reinforcing relationship that together constitute the foundational models for physical AGI, thereby accelerating physical AGI toward its “GPT-3 moment.” To some extent, Jiyi Vision has charted a novel path that is gradually being validated.

A new watershed moment in the AI era has arrived.

Over the past few years, digital AGI has focused on information processing and virtual interaction, leveraging large language models and multimodal generative models to enable functions such as text creation, image design, and code generation—essentially optimizing and enhancing "information productivity."

The limitations are also evident. Although digital AGI has greatly improved the efficiency of information dissemination, content creation, and data processing, it can never transcend the boundary between virtual and reality. As "Godmother of AI" Fei-Fei Li said, large language models are still "wordsmiths in the dark"—eloquent but lacking experience, knowledgeable but not grounded.

Therefore, in the view of the Jiyi Vision team, AGI should not remain confined to screens. The core value of physical AGI lies in physical execution and transformation of the physical world—understanding physical laws through world models, perceiving the physical environment through multimodal sensing, and executing physical actions through mechanical bodies.

Undoubtedly, GPT-3 is widely recognized as the pivotal moment when scaling laws first demonstrated emergent capabilities on the path to digital AGI. Today, after three years of continuous breakthroughs in algorithms and data systems, Jiyi Vision has observed a clear trend toward convergence in the physical AGI approach, suggesting that the "GPT-3 moment" for physical AGI may be imminent.

According to reports, GigaBrain-1 will be released in the third quarter of this year. As the world’s first physical AGI foundational model built on the “Double Pyramid” architecture, GigaBrain-1 will deliver three key breakthroughs: vision-native understanding (using vision as the primary channel for state perception), high-level language planning (leveraging language for task decomposition at the strategic level), and physical law alignment (systematically expanding training data across all types and scales).

After this, GigaBrain-2 and GigaBrain-3 will be launched in succession. GigaBrain-3 will be trained on 10 million hours of video data and 1 million hours of world-action data, aiming for the “GPT-3 moment” of physical AGI.

Of course, technology must ultimately deliver real industrial value.

JiJia Shijie has taken a unique approach: targeting C-end households and B-end factories simultaneously. In the industry, very few embodied AI companies have secured household orders, as real-world home environments present far more complex and diverse demands compared to standardized industrial settings.

Nevertheless, Jíjiā Shìjiè pressed forward, recently launching its home scenario sub-brand “Shíguāng SeeLight” and introducing its first general-purpose humanoid robot designed for real homes, the Shíguāng S1, which has already secured orders for 100 units for real-world home deployment. The robots will be率先 deployed in the Guanggu Zhiyu community, with large-scale operations set to begin in the third quarter; the next-generation general-purpose home robot, Shíguāng S2, will also be released in the third quarter.

Thus, Jiayi Shijie has once again broken new ground by securing the industry’s most scarce resource: real-home device data. Aligned with the product roadmap of the Shiguang S2/S3, this marks the physical AGI equivalent of ChatGPT’s breakthrough moment—enabling everyday skills to be widely applied in real home environments.

On the B-side, Jiayi Vision is transitioning from single-point validation to large-scale production in industrial manufacturing scenarios. In April this year, Jiayi Vision launched Maker H01, its fully self-developed physical AGI-native general-purpose robot, and, in collaboration with FAW Mold and Alibaba Cloud, successfully implemented an end-to-end solution for embodied intelligent robots in real-world industrial manufacturing environments, reducing the traditional automation scenario adaptation cycle from months to weeks.

Meanwhile, Jijia Vision announced this month that it plans to deploy 1,000 general-purpose robots equipped with Jijia Vision’s World Model embodied brain and Maker series in Wuxi in collaboration with Longsheng Technology over the next three years—marking the world’s first large-scale industrial deployment of general-purpose robots powered by a physical intelligence foundation model, signaling that China’s embodied intelligence has fully moved beyond pilot programs and entered a phase of mass industrial production.

On the other hand, Jiyi Vision has already established its DriveDreamer series of world models as an industry benchmark—this next-generation driving simulator, centered on world models, has secured contracts and mass production partnerships with multiple leading domestic automakers, overseas and joint-venture automakers, as well as major AI chip and Tier 1 suppliers, serving over 30 top automakers and autonomous driving companies worldwide.

In summary, the B-end strategy represented by industrial product lines aligns with the Claude Code moment for physical AGI—breakthroughs in advanced skills within productivity scenarios.

More importantly, the practical implementation of Excellent Vision achieves dual-track operation, enabling the continuous accumulation of real-world data and cash flow to further reinforce the data foundation of the “Dual Pyramid,” driving the “Scenario–Data–Model–Product–Ecosystem” flywheel.

In Huang Guan’s view, the GPT-3 era marked the emergence of intelligent capabilities in models; the ChatGPT era brought productivity benefits to every ordinary person; and the Claude Code era elevated digital intelligent models to the level of professional experts.

As a leading domestic enterprise that was among the first to invest in physical AGI, Jiajia Vision believes that physical AGI will also undergo similar stages, with the key difference being that physical AGI will directly interact with the real physical world. It will not only enhance information efficiency but also reshape production and lifestyle, resulting in a more profound impact on the economy and society.

Throughout the history of human civilization, every major leap in productivity has been driven by groundbreaking advancements in core technologies. Therefore, when AI truly transcends digital boundaries and enters the vast realm of the physical world, it will inevitably spark a new productivity revolution, unleashing limitless physical productive capacity.

This is precisely the ultimate vision painted by JiJia ShiJie: the era when physical AGI serves every individual will gradually unfold in real households.

Perhaps this moment is about to arrive.

This article is from the WeChat public account "Investment Times" (ID: pedaily2012), authored by Liu Bo.