Deep Dive | Robot Carnival and Commercial Cold Reflection

07/10 2024 382

Image Source: Visual China

Blue Whale News, July 9 (Reporter Zhang Han) With the curtain falling on the 2024 World Artificial Intelligence Conference (WAIC), "embodied intelligence" undoubtedly became the core focus of this year's conference. Riding on the strong winds of artificial intelligence, this concept proposed by Turing over 70 years ago is gradually moving from theory to reality, and humanoid robots are the most intuitive expression of embodied intelligence.

At this exhibition, humanoid robots that can walk, dance, and perform splits attracted the most attention from visitors. Near Hall H1-2, the orderly arranged "Eighteen Arhats - Pioneering Array of Humanoid Robots" became a popular photo spot, with "star" products such as Tesla Optimus, Fourier GR-1, Kuafu 4th version, Unitree H1, and Kepler Pioneer K1 gathered, filling the air with a sense of science fiction becoming reality.

As the representative of humanoid robots, Tesla's second-generation humanoid robot Optimus, encased in a glass protective cover, could only be seen and not touched, yet visitors to its booth were still endless, requiring strategic timing to squeeze in for a close-up look at Optimus' true form, making it the most popular attraction in the entire exhibition hall.

During the exhibition, Blue Whale News reporters interviewed over a dozen humanoid robot companies exhibiting, as well as some investors, industry insiders, and visitors, attempting to analyze the real situation behind this embodied intelligence feast from more diverse perspectives.

"At this juncture, any company that can make humanoid robots will take them out to boost their image, but let's not be blindly optimistic; commercialization is still a distant prospect." A staff member from an exhibiting company told Blue Whale News. These robots that increasingly resemble humans are filled with technological charm, but when asked about mass production and commercialization, most company staff members would hem and haw. Just like the question repeatedly asked by most visitors, "It's fun to watch, but what would I buy it for?"

During the interviews, Blue Whale News reporters also noted that for the current status and development of humanoid robots, education, healthcare, industrial manufacturing, and service industries are the scenarios commonly being explored. However, regarding pricing, some believe that it is now time to seize the market with relatively low prices, while others argue that initiating a price war in the early stages would have little effect and could even be detrimental to the industry.

In the face of this trend, robot companies strive to tell good stories to the market and investors, but is the future of humanoid robots really bright enough? The answers in their hearts seem less than certain.

Consensus: Attempting to Take the Lead in Landing in TO B Scenarios

"The changes this year have been too rapid. Last year's conference focused on software aspects like large models, but this year, everyone came to see humanoid robots." A staff member from an exhibiting company told Blue Whale News.

According to official WAIC data, a total of 45 smart robot companies and 25 humanoid robot companies participated in this year's exhibition, compared to only around 10 humanoid robot companies a year ago. According to media reports, there were no bipedal humanoid robots at last year's AI conference. But this year, companies like Zhuoyide led their humanoid robots around the exhibition hall. At the Humanoid Robots and Embodied Intelligence Development Forum on July 4, Wang Xingxing, founder and CEO of Unitree Robotics, a leading domestic robot company, delivered a keynote report titled "Embodied Intelligence is the Most Effective Path to Achieve AGI."

Beyond the embodied intelligence concept carnival, Blue Whale News reporters also noticed that most visitors would ask exhibiting companies, "What can this robot do? Can it help me with housework?" The companies' answers were generally consistent: "Not yet, but there is unlimited imagination for the future." Liu Yufei, technical director of the National-Local Co-constructed Humanoid Robot Innovation Center, said at the July 4 "Humanoid Robots and Embodied Intelligence Development Forum" that the difficulties in applying humanoid robots lie in challenging scenarios and difficulty in obtaining high-quality data.

Difficult though it may be, companies are already moving forward through trial and error. Blue Whale News reporters noticed during on-site interviews that most companies would choose sectors such as education, scientific research, industrial manufacturing, and services as the first scenarios for landing. The full-size robot "Kuafu," which made its debut at Huawei Developer Conference 2024 and is the first in China to run on Huawei's HarmonyOS, was also present, with developer Leapmotor stating that the robot has begun mass production and is expected to be priced between 600,000 and 800,000 yuan, with primary use scenarios in education and scientific research rather than industrial manufacturing.

A responsible person from Leapmotor told Blue Whale News, "Leapmotor robots have been exploring and trying in the field of education since 2016, and product development requires channels for sales, so there is a certain accumulation. In addition, humanoid robots are still in a relatively early stage, and they have not yet reached a highly refined level in terms of systems, perception, decision-making, control, and algorithms, so it is necessary to cooperate with domestic universities to accumulate technological advantages."

At 3:30 p.m. on July 5, the booth of Fourier Intelligence, a robot company, was packed with people as they demonstrated their upgraded humanoid robot GR-1. At last year's AI conference, Fourier Intelligence released the full-size humanoid GR-1 and took the lead in achieving mass production. The relevant person in charge of Fourier told Blue Whale News, "GR-1 will primarily be deployed in scenarios such as rehabilitation and companionship, which are areas where we excel."

"Overall, current humanoid robots are relatively similar, and it's not clear which is stronger. The key is to find suitable landing scenarios." An industry insider admitted, "Whether it's industrial manufacturing, services, or education, the technical difficulties are not significantly different. However, healthcare and medical care have relatively higher requirements because robots need more data training to achieve precision and ensure safety when interacting with humans."

During the interviews, Blue Whale News reporters noted two reasons why companies prefer to initially try landing in sectors such as scientific research and education: first, due to the influence of high prices, only B-end enterprises have the ability to pay; second, it is related to their own industry accumulation, such as healthcare for Fourier and education for Leapmotor. Another exhibitor who wished to remain anonymous said they chose to invest in the industrial sector due to previous related experience, but it's unclear which sector can better leverage the capabilities of humanoid robots.

Divergence: Continue Focusing on Product Development or Join the Price War

Robots are not a particularly advanced concept. From industrial robots to collaborative robots and now humanoid robots, robots have always existed in the industrial chain and have been the goal of generations of entrepreneurs and investors. In recent years, humanoid robots have quickly gained attention, partly due to the development of AI technology and the popularity of embodied intelligence, but more importantly, due to the driving force of benchmark companies like Tesla.

A few years ago, although the American robot company Boston Dynamics also generated buzz with its cool demonstration videos, the industry did not perceive much commercialization potential for humanoid robots until Tesla changed that. Humanoid robots are no longer just a "castle in the air."

Tesla stated that the second-generation humanoid robot Optimus will subsequently be sold to individuals, with a price expected to be around US$20,000 (equivalent to RMB 145,000), and limited production will begin next year. According to Tesla CEO Elon Musk, by 2025, over 1,000 Optimus humanoid robots will be put into work at Tesla factories. He predicts that in the future, humanoid robots will become the main force in industry, potentially outnumbering humans, reaching 10 to 20 billion. Tesla aims to produce 1 billion units annually, capturing over 10% of the market and generating a market value of US$25 to 30 trillion for Tesla.

Along with Tesla, Unitree Technology also lowered prices when it launched the G1 humanoid robot in May this year, with a starting price of RMB 99,000, though it did not attend the exhibition. Unitree Technology told Blue Whale News that G1 did not attend because it has not yet started mass production, which is expected to begin in Q3 of this year.

Compared to the H1, which was released in August 2023 with a price of RMB 650,000, the price of G1 dropped rapidly. According to the relevant person in charge at Unitree Technology, "The cost reduction is mainly due to Unitree's self-research of all core components, such as drivers, controllers, motors, code, and production. However, humanoid robots are still in the early stages of development, and both H1 and G1 are products for early market exploration."

According to the Economic Daily, the most expensive part of a humanoid robot is the joint, which is mainly composed of motors, drivers, reducers, encoders, and power sensors, accounting for about two-thirds of the overall cost. For example, servo motors used to ensure the flexibility of robots are priced between RMB 2,000 and 3,000 each abroad, and an average humanoid robot requires 20 to 30 of them, making the cost of motors alone reach RMB 100,000. Reducing prices is indeed a challenge, as not every humanoid robot company has achieved self-research of relevant core components.

At the exhibition, Blue Whale News reporters noticed significant differences in market pricing for humanoid robots among different companies. Tesla Optimus and Unitree G1 fall into the RMB 100,000 range; CloudMinds XR4 and Kepler's Pioneer K1, which plans to start mass production in the second half of this year, are around RMB 300,000; Unitree H1 and Leapmotor "Kuafu" are in the RMB 500,000 to 600,000 range; while Fourier, StarEra, and other companies have not announced official prices for their products.

Among them, staff at Kepler and CloudMinds booths believed that although humanoid robots are still in their infancy, competition is fierce, and it is time to seize the market through pricing. However, staff at Taihu and Fourier argued that price plays a negligible role in the current market, and the focus should be on refining products first, as initiating a price war at this stage would be detrimental to the industry.

An industry insider told Blue Whale News, "As long as a suitable landing scenario is found and mass production is achieved through market demand, prices can come down quickly; it's not a particularly difficult problem to solve."

At the exhibition, Blue Whale News reporters also encountered some company executives conducting market research, who agreed that the current selling price of humanoid robots is not their primary concern. As long as large-scale procurement and mass production are achieved, price is not an issue. The focus of early investigation is still on performance and scenarios, such as the robot's battery life, load-bearing capacity, and obstacle avoidance abilities, which are viewed favorably in the long run and are still under consideration.

Future: Some Are Firmly Optimistic, While Others Remain Observant

This year's WAIC, with bipedal robots walking around and accompanied by explanations from company staff, truly ignited the enthusiasm of visitors and the market for this industry. Just like the temperature in Shanghai during the exhibition, despite the air conditioning, the exhibition hall felt far from cool.

During Blue Whale News' interviews, the industry was found to have two factions: one that is firmly optimistic, with resolute and confident eyes. Whenever visitors asked when the robot could serve them, booth staff would reply, "Soon, very soon. People used to think it was incredible to drive electric cars, but now they're all new energy vehicles, right? Technology develops rapidly, and in a few years, we may not even need blind massages anymore."

Global tech leaders like Tesla's Elon Musk and NVIDIA's Jen-Hsun Huang, as well as domestic robot company founders like Wang Xingxing of Unitree Technology and Tan Min of UBTECH, are representatives of the optimistic camp.

There is also relevant data to support this optimism: The "Humanoid Robot Industry Research Report" released at the China Humanoid Robot Industry Conference and Embodied Intelligence Summit predicts that the Chinese humanoid robot market will reach RMB 2.76 billion in 2024 and RMB 75 billion by 2029, accounting for 32.7% of the world's total and ranking first globally. By 2035, it is expected to reach a scale of RMB 300 billion.

Wang Xingxing told Blue Whale News, "The field of humanoid robots has experienced decades of ups and downs, with peaks and troughs. Especially after the sensation caused by AI large models last year, new technologies and AI advancements have provided impetus for the practical application value of robots. The government is also actively promoting development in this field, which can be seen as a new starting point. Of course, it will take another three years to truly achieve the popularization of humanoid robots."

UBTECH's Chief Brand Officer Tan Min has also publicly stated that the development of humanoid robots is currently embracing its own "iPhone moment," with AI technology rapidly advancing towards the critical juncture of breakthroughs in general AI. Humanoid robots are also developing at an unprecedented pace, with widespread applications expected within 10 to 20 years.

However, not everyone shares this optimism. "We've been making robots for many years, but we can't really imagine a scenario where they can be mass-produced and commercialized. It's quite troubling. Moreover, as far as I know, no company currently has a clear commercialization model. Their way of earning money is still to rent out their robots at exhibitions and collect commissions, which isn't very meaningful." A staff member from an exhibiting company told Blue Whale News.

At the exhibition, Blue Whale News reporters saw more static demonstrations of companies' humanoid robots or simple walking around their own venues, with only a few sporadically navigating the exhibition hall.

"Actually, a week before the conference, everyone planned to have their robots walk around the venue, but after testing, we found too many uncontrollable factors, so it eventually became mainly static displays. Right now, the way these humanoid robots walk is more like human babies, with a lot of room for growth." An industry insider told Blue Whale News candidly.

A staff member from an exhibiting company also told Blue Whale News, "Don't be too optimistic about humanoid robots. As it stands, they still exist more in imagination. It's not just because of issues with scenario landing; another reason is that to make humanoid robots, financing is inevitable. When capital comes in, it's easy for ideas to diverge, and it's not easy for the team to stay united."

According to incomplete media statistics, there were over 22 financing events in the global humanoid robot sector in the first half of 2024, with a total financing amount exceeding RMB 7 billion. Domestically, companies like StarEra, Unitree Technology, Zhiyuan Robotics, and Kepler Robotics have all received funding.

Within the WAIC exhibition halls, like large models, embodied intelligence was the hottest concept keyword of the year; but outside the WAIC exhibition halls, in actual production and life, robot companies are still amidst a complex tapestry of thoughts, hesitations, trials, and uncertainties.

Solemnly declare: the copyright of this article belongs to the original author. The reprinted article is only for the purpose of spreading more information. If the author's information is marked incorrectly, please contact us immediately to modify or delete it. Thank you.