news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

A rare financing storm has emerged.

The investment community learned that Skild AI recently announced the completion of a $300 million Series A financing round, with investors including SoftBank Group, Bezos, Sequoia Capital, Lightspeed Venture Partners, Coatue, etc. So far, the company, which was established only one year ago, has a valuation of $1.5 billion (about 10.9 billion RMB).

Coincidentally, China's Zhuji Power also completed its Series A financing on the same day, with China Merchants Venture Capital, Shangqi Capital, Alibaba, Frees Capital, Oasis Capital and BrightSun Capital among the investors.

Recalling last month, Galaxy General won the largest angel round of financing this year, pushing the financing craze of embodied intelligence to a climax.

Half seawater, half fire. Controversy ensues: Is this a great innovation that will change the future, or another valuation bubble? Some people go left, some go right.

Two professors start a business

One-year valuation exceeds 10 billion

Who is Skild AI?

This is a startup company in the field of embodied intelligence based in Pittsburgh. It was founded in May 2023 by two former Carnegie Mellon University professors, Abhinav Gupta and Deepak Pathak. Most of the team members come from companies such as Meta, Tesla, and Nvidia, as well as prestigious universities such as Carnegie Mellon University and Stanford University.

Officially, Skild AI's long-term goal is to "develop general artificial intelligence (AGI) rooted in the physical world to create a general, flexible, and intelligent humanoid robot product."

Image source: Skild AI official website

In more general terms, Skild AI is building aUniversal BrainSkild Brain, this basic model can be installed on various robots, enabling them to complete basic actions such as climbing stairs, bypassing obstacles, picking up objects, etc. Through more than a thousand times the amount of training data, the robots can perform tasks they have never seen before.

The founder has been working on this for a long time. When Pathak was studying for his doctorate at the University of California, Berkeley, he developed a method for robots to "learn with curiosity", which encourages AI to browse more scenes and collect more data, and finally turn it into action. After the research was published in 2017, it has been cited more than 4,000 times.

VCs rushed in. In July this year, Skild AI announced that it had completed a $300 million Series A financing, with a post-investment valuation of up to $1.5 billion. The investment lineup was luxurious.

Led by Lightspeed Ventures, Softbank, Coatue and Amazon founder Jeff Bezos, the round also included CRV, Felicis Ventures, Menlo Ventures, Sequoia Capital, General Catalyst, SV Angel and Carnegie Mellon University. Earlier, in July 2023, Lightspeed Ventures also led the seed round of Skild AI.

It is surprising that the company quickly became a unicorn within just one year of its establishment.

According to the opinions of investors in this round, the two founders are called "catalysts" of robotics. Some of the progress made in the field of robotics in the past few years can almost be traced back to their academic achievements. At the same time, Skild AI has "revolutionized robotics technology, from pre-programmed robots to dynamic adaptive robots, which has the potential to subvert the entire real economy" in terms of building basic models for robot operation and movement.

It should be pointed out that although Skild AI emphasizes that its product model can be used in both "quadruped robots" and "humanoid robots", the company has not yet announced specific partners.

In fact, all embodied intelligence companies are facing competition from large manufacturers. For example, Google is iterating its large robot model, Nvidia has also launched a large model and development platform for the robotics field, and the robot body manufacturer Tesla Optimus Prime is also evolving rapidly. OpenAI has also recently restored its robot team... You can imagine how fierce the competition is.

Here, queue up to announce financing

Looking around, embodied intelligence is experiencing a wave of hot financing.

In detail, this is the hottest investment track after AI this year, with frequent large-scale financing of more than 100 million yuan. On July 15, general robot startupsZhuji PowerCompleted Series A financing, the investment community learnedThe amount is several hundred million yuanInvestors include China Merchants Venture Capital, Shangqi Capital, Alibaba, Frees Capital, Oasis Capital and Mingshi Capital.

Zhuji Power was founded by Zhang Wei, a tenured professor at the Southern University of Science and Technology. He is an international scholar in the field of robotics and automation and was a professor at Ohio State University in the U.S. At present, its products include humanoid robots, four-wheeled and legged robots, etc., and are seeking better landing scenarios.

Also recently announced funding:Titanium Tiger Robot, the tens of millions of yuan Pre-A round of financing was led by Matrix Partners, followed by Fangfu Ventures and Meihua Ventures. This is a hardware manufacturer focusing on robot joints and is about to release the T170A full-size humanoid robot model.

In the field of embodied intelligence, there is a saying of “Four Little Dragons, New and Old”

The “old four little dragons” include UBTECH Robotics, Cloud Robotics, Feixi Technology, and Yushu Technology; the “new four little dragons” are Zhiyuan Robotics, Galaxy General, Zhuji Power, and Youlu Robotics.

Among them, the 90s former Huawei "genius boy" ZhihuijunZhiyuan Robot, which is considered one of the hottest projects in the VC circle. In February 2023, Zhihuijun partnered with Yan Weixin, a professor at Shanghai Jiaotong University, and Zhiyuan Robotics was officially established in Shanghai. So far, it has launched the first generation of robots, Zhiyuan Yuanzheng A1.

During this period, Zhiyuan Robotics has been working non-stop to obtain several rounds of financing, not only gathering a number of leading institutions such as Hillhouse, Matrix Partners, CDH Investments, Gaorong, BlueRun, and Sequoia China, but also industrial capital such as BYD, SAIC Ventures, and Baidu Ventures. It is reported that its latest round of financing has been launched in June this year.

Galaxy UniversalIt has just received the largest angel round of financing this year, totaling 700 million yuan, including strategic and industrial investors such as Meituan Strategic Investment, BAIC Capital, SenseTime Guoxiang Fund, and iFlytek Fund; leading financial institutions such as Qiming Venture Partners, BlueRun Ventures, Matrix Partners, Source Code Capital, and IDG Capital; Guangyuan Capital also participated in the early investment.

Founded in May 2023, Galaxy General is led by Wang He, who graduated from the Department of Electronics at Tsinghua University and holds a PhD in Computer Science from Stanford University. He is currently an assistant professor and doctoral supervisor at the Frontier Computing Research Center of Peking University. Now the company's first-generation product, Galbot, has officially debuted.

In addition, Yushu Technology, Youlu Robotics, Kuawei Intelligence, and Steele have all received financing this year. Earlier, Pasini Perception Technology, Star Era, Dahua Robotics, Fourier Intelligence, and Leson Robotics all had investors behind them.

The AI ​​wave is coming, and embodied intelligence has become the consensus of investors in 2024. However, there are also hidden concerns behind the enthusiasm. When will humanoid robots, which are still difficult to walk independently, really enter factories?

“If humanoid robots cannot be put into practice, they are just a good story, not a good business,” a VC partner said.

Controversy behind the scenes

Kai-Fu Lee: Not investing in any company

The hotter the track is, the more you need to calm down.

Wang Tianmiao, honorary director of the Robotics Institute of Beihang University and director of Zhongguancun Zhiyou Research Institute, has previously pointed out that current robots are still relatively "retarded" in many environments, unable to understand scene requirements and unable to eliminate interference. "We have made progress in language and vision large models, but have not found similar models in the field of robotics. Only breakthroughs in robot large models can lay the foundation for the research and development of general artificial intelligence robots."

Wang Tianmiao believes that from an evolutionary perspective, how to coordinate the three computing spaces of language cognitive intelligence, behavioral intelligence, and visual intelligence, and efficiently utilize real, high-quality, and massive data, also requires breakthroughs. This is undoubtedly the top priority. The Google Robotics Research Team spent 17 months and obtained 130,000 pieces of data on more than 700 tasks based on 13 robots. This shows how difficult it is to collect a large amount of practical data.

The uncertain commercial prospects have also poured cold water on embodied intelligence.

UBTECH, once known as the "first stock of humanoid robots", has seen its share price go on a roller coaster ride recently. Compared to its historical high of HK$328 four months ago, it has now fallen by more than 60%, with a market value of more than HK$40 billion evaporating in an instant. "The technology is not mature enough in commercialization, and there may still be a gap in practicality," said an investor.

A VC from a US dollar fund shared that he recently discussed with some friends in the circle and everyone had a surprisingly consistent view that startups involved in embodied technology would not be able to truly land in 3 to 5 years and would need at least 10 years to get started; in addition, embodied intelligent companies have performed well in the financing process, but this has also caused everyone to worry about valuation bubbles, and the massive injection of funds will make some startups lose their rationality in market competition.

“Large companies have absolute advantages. Even if they lose money on R&D, they can make it back through stock prices. My VC friends have made it clear that they will no longer invest in current startups, but will wait together for future opportunities.”

Previously, Wang Xinyu, partner of Meituan Longzhu, told the investment community that humanoid robots will definitely enter thousands of industries and households, but the difficulty of achieving this is also proportional, and the right time may not have come yet.

Kai-Fu Lee's attitude is more direct. In his view, embodied intelligence is a combination of the physical world and AI. Once a large model is connected to the physical world, it will face various problems including safety, machine, mechanical, and failure, which will be many times more difficult.

At the same time, he also questioned the necessity of making robots in human form. "The vast majority of application scenarios do not require humanoid robots. A cooking robot should look like a pot, and a vacuum cleaner does not look like a human. Will cool, jumping robots like Boston Dynamics really have many application scenarios?"

At the end, Kai-Fu Lee bluntly said that many scientists and entrepreneurs have loved technology since childhood and hope to replicate a person, which is understandable. But if you ask VCs about their thoughts, he said that Sinovation Ventures has invested in 6 driverless companies, and 0 embodied intelligence and humanoid robot companies, "I believe in robots, but for embodied intelligence, we as investors certainly cannot invest in something that will only happen 10 years later."

As science fiction writer William Gibson said in Neuromancer, "The future is already here, it's just unevenly distributed." The birth of any epoch-making technology will be accompanied by countless controversies and challenges. The same is true for embodied intelligence. Time will eventually give the answer.