Raised 2 billion yuan in half a year, the youngest embodied unicorn is born.

At the beginning of 2026, the venture capital circle experienced a "barrage of news". Dozens of embodied intelligence companies successively announced new rounds of financing. In just two months, the total disclosed financing amount was close to 15 billion RMB, and the single - round financing amount often started at one billion.

However, if you take a closer look at these companies, you'll find that almost all of them were established before 2024. This also confirms a conclusion in a previous article on China Venture Capital News: As resources become more concentrated, the opportunities for new entrants in the field of embodied intelligence are getting smaller and smaller. But the interesting thing is that even though the general trend is like this, you'll always expect "newcomers" to break through the old barriers, disrupt the old pattern, and create a brand - new narrative.

The pattern of embodied intelligence is far from being set, and the "newcomers" are already here. China Venture Capital News exclusively learned that Simplexity Robotics completed 5 rounds of financing in less than half a year, with a total financing amount of 2 billion RMB. After the last round of financing, its post - investment valuation exceeded 1 billion US dollars. Thus, Simplexity Robotics, founded in the second half of 2025, has become the youngest unicorn in the current embodied intelligence track.

The so - called "youngest" means firstly, it has a short establishment time. According to the business registration information, it was registered at the end of July 2025. As of now, it's only been a little over 8 months, less than a year old. Secondly, it became a unicorn at a fast pace. China Venture Capital News learned that these 5 completed rounds of financing only took less than half a year. The rapid progress has also led to a strong consensus among institutions. Finally, from a time perspective, although Simplexity Robotics is a "newcomer", it's not inexperienced. Its founding team comes from the core group of Li Auto, and they have already proven themselves in terms of technology, products, and business.

Looking at the investors in these 5 rounds of financing, they are all well - known market - oriented institutions in China: including leading investment institutions such as Vision Plus Capital, BlueRun Ventures, Sequoia China, Legend Capital, CAS Star Ventures, and Banyan Capital, as well as strategic investors from the technology ecosystem like Tencent and Alibaba (part of the information is from CVSource). Moreover, according to public information, most of the old shareholders, such as Sequoia China, BlueRun Ventures, and Legend Capital, have continuously increased their investment in multiple rounds since the first - round investment.

What do these keywords - less than half a year, 5 rounds, 2 billion RMB, unicorn, well - known institutions, and multiple - round investment increases - make you think of? What comes to my mind is that the long - lost sense of hunger has finally emerged in the venture capital circle. And the fact that this long - lost "hunger" appears in a company that has only been established for 8 months is worth a closer look.

Genes

The most core logic for the early - stage leading institutions to collectively bet on Simplexity Robotics is definitely the people.

As mentioned earlier, the core founding team of Simplexity Robotics comes from Li Auto. Information shows that the CEO of Simplexity Robotics is Jia Peng, the former head of intelligent driving technology R & D at Li Auto; the chairman is Wang Kai, the former CTO of Li Auto; and the COO is Wang Jiajia, the former person in charge of intelligent driving mass production at Li Auto. So, when talking about why VCs collectively bet on Simplexity Robotics, two key factors are inevitable: the competitive technical background and the proven large - scale implementation experience.

This team has experienced survival crises and darkest moments. In the most competitive and brutally iterative market, they have witnessed the whole process from life - and - death struggles to breaking through against the odds, and completed the full verification from 0 to 1 and then from 1 to large - scale implementation. This means that in China's most competitive market, they have a set of combat capabilities that have been repeatedly polished by the market and are extremely systematic.

Technically, compared with previous industrial robots, the current core of embodied intelligence is generalization ability, which requires robots to be able to understand the physical world, make autonomous decisions, complete precise actions, and achieve self - iteration. The Simplexity team has not only realized this closed - loop system in the field of intelligent driving but also applied the same paradigm to the automated production and quality inspection production lines of new energy vehicles.

Embodied intelligence has also reached a critical point now. The discussions on "mass production, entering factories, and doing work" are very popular. This also requires a set of systematic capabilities including strategy, technology, brand, product, organization, and business, which is exactly what Jia Peng, Wang Kai, and Wang Jiajia have fully experienced, implemented, and successfully verified at Li Auto.

According to Jia Peng, the CEO of Simplexity Robotics, "Systematization is the differentiation of Simplexity Robotics."

Jia Peng is an expert in the field of VLA models and world models in China. He has had multiple core R & D management experiences at IBM, NVIDIA, and Li Auto. He is the initiator of the world's first VLM + end - to - end fast - slow dual - system and the first VLA model R & D and delivery. Wang Kai has worked at Nokia and Visteon. In 2021, as the CTO of Li Auto, he led the team to set a record in the industry by achieving the mass production of the intelligent driving system on vehicles in 7 months. Later, he served as a Venture Partner at Vision Plus Capital. Wang Jiajia was once one of the youngest R & D directors in the history of Bosch China. Later, at Li Auto, she led the full - process mass production and delivery of the intelligent driving system.

Therefore, it's actually not easy for VCs to find a core founding team that has fought together, is familiar with each other, has a clear division of labor, and has systematic combat capabilities. Otherwise, there wouldn't be so many "assembled" projects. Jia Peng, Wang Kai, and Wang Jiajia basically cover all the core aspects of current entrepreneurship in embodied intelligence, including strategy formulation, fundraising, and talent recruitment, full - stack technology and model R & D, and engineering implementation and large - scale mass production delivery and execution. Therefore, it's not surprising that leading market - oriented VCs and top Internet companies collectively bet on Simplexity Robotics.

Philosophy

Additionally, I think the "hunger" of VCs also comes from the expectation of the technological end - game.

From humanoid robots to embodied intelligence, as long as the goal is to enter the market and do work, the routes are gradually converging. There's no need to argue about bipedal or wheeled robots. The end - to - end VLA (Visual Language Action) model, the embodied brain, has become the standard, which is the basis for the explosion of financing in embodied intelligence.

However, there are still many problems and differences in the industry. For example, even though embodied intelligence companies are emphasizing generality and generalization, the performance of current embodied robots in unstructured and open environments is still not satisfactory. This is directly related to whether robots can achieve stable commercial use in scenarios such as factories. Stable commercial use can not only help users reduce costs but also, more importantly, continuously collect real - world data to feed back the model evolution, achieving a double - closed loop of commercialization and data.

Therefore, to enter the market and do work faster, traditional robot companies adopt a modular development approach, with one function per module, resulting in poor generalizability. Some embodied companies choose a semi - end - to - end approach, training mainstream functions separately after layering them and then "piecing them together" with protocols, which has a certain degree of generalizability but can't handle delicate tasks.

Simplexity Robotics' solution is to create an integrated model of the world model and VLA, realizing the joint modeling, understanding, generation, and prediction of language logic, visual semantics, 3D spatial structure, and robot status through a unified transformer. While achieving a model architecture with a higher upper limit, it reduces manual design and has better scaling effects.

In other words, Simplexity Robotics has provided a solution closer to the "end - game", that is, eliminating the above - mentioned functional layering and integrating all the capabilities required by robots, such as language, vision, and physics, into a general large - scale model. The advantage of this approach is the strongest generalizability, but it's definitely not easy to train, and it requires more data.

Therefore, achieving a data - closed loop is crucial for Simplexity Robotics, even a matter of life and death. Simplexity Robotics has chosen to develop only one general - purpose body to cover as many scenarios as possible, improving data generalizability and reusability.

To solve the problems of model training efficiency and privacy, Simplexity Robotics has chosen to deploy and pre - embed additional computing power on the device side, enabling the robot to complete data collection, training, testing, and verification on the body, better binding to real - use scenarios. That is, "achieving device - side training and model testing and verification in user scenarios through the shadow mode, creating an extremely efficient closed loop of data collection, training, testing, verification, and deployment."

Of course, there's a price - high single - machine cost. However, these technological choices represent the survival philosophy of Simplexity Robotics. Like its name, Simplexity Robotics hopes to solve the complex problems in the field of embodied intelligence with a simple and scalable methodology, defining the body through the model and the hardware through software. The above - mentioned one model and one body are also based on this.

Specifically, Simplexity Robotics has currently built a technical architecture centered around "Four Os" (one model, on device, one hour, one body). It has launched the LaST₀ base model, the ManualVLA ultra - long - range task model, and the TwinRL real - machine reinforcement learning framework, corresponding to the three challenges of improving the robot's efficient reasoning ability for the dynamics of the physical world, enabling the robot to truly understand complex long - range tasks, and making the robot stronger in the real world.

LaST₀ base model: It fuses the world model's understanding and prediction of the physical world with the fast - slow thinking of VLA for the first time, significantly improving the efficient reasoning ability for the dynamics of the physical world and solving the problem of how the robot can "think while moving quickly".

ManualVLA ultra - long - range task model: Based on the powerful foundation of LaST₀, ManualVLA solves the problem of how to enable the robot to understand complex long - range tasks. The model can automatically generate multi - modal "operation manuals" similar to those used by humans from the target state, perfectly answering the question of how the robot can "think clearly before taking action" (this paper has been accepted by CVPR 2026).

TwinRL real - machine reinforcement learning framework: When the model has the ability to reason and execute, the final key lies in how to make it continuously evolve in the real world and achieve real - world implementation. TwinRL expands the exploration space of real - machine reinforcement learning with the help of digital twins. In multiple tasks, the robot can achieve a 100% success rate in the desktop area in less than 20 minutes, solving the challenge of "how to make the robot stronger in the real world".

As for the body, Simplexity Robotics adheres to full - stack self - research in design and manufacturing. According to China Venture Capital News, it only took less than 45 days from the arrival of the first employee to the birth of the first - generation self - developed body. Currently, Simplexity Robotics has completed the R & D of two generations of bodies for B - end and C - end users, achieved small - batch production of the bodies, and fully launched PoC verification.

As for specific implementation scenarios, Simplexity Robotics follows a progressive iterative path from closed to semi - open and then to fully open environments. In terms of scenarios, factory workshops, shopping malls, logistics, etc. are the first targets.

Currently, the company has completed strategic layouts in Beijing, Shanghai, and Suzhou. Next, Simplexity Robotics will fully invest in areas such as training the base model, body R & D and iteration, data collection, and core algorithm R & D, complete scenario expansion from factories to the manufacturing and service industries, and from domestic to overseas markets, continue to focus on technological innovation and industrial cooperation, empower the large - scale implementation of the embodied intelligence industry with technology, and accelerate the large - scale application of embodied intelligence technology in multiple scenarios.

This article is from the WeChat official account "China Venture Capital News", author: Zhang Nan. Republished by 36Kr with permission.