Preface:

The market is increasingly paying attention to [Embodied Brain] technology. The underlying reason is that domestic development in this area is still limited, particularly lacking an [Embodied Brain] specifically tailored for the robotics industry and product level.

However, the absence of advanced brain capabilities severely hampers the autonomous decision-making intelligence of robots.

Author | Fang Wensan

Image Source | Network

2024 Chip Company Revenue TOP10: A New Era Begins

The 2024 global chip company revenue rankings released by market research firm Omdia have caused a significant stir in the semiconductor industry.

This year, the global chip market size reached an impressive $683 billion, marking a 25% year-on-year increase.

Behind this substantial growth, the surging demand for AI-related chips was undoubtedly the primary driver.

NVIDIA successfully topped the list of global chip company revenues, thanks to its outstanding performance in the AI chip field, emerging as the new champion.

This achievement not only showcases NVIDIA's robust capabilities in technological innovation and market strategy but also signifies the deep penetration and widespread application of AI technology within the semiconductor industry.

From obscurity to industry dominance, NVIDIA's rise is nothing short of legendary.

Capitalizing on the rise of AI technology, NVIDIA swiftly seized this historical opportunity, significantly increasing its R&D investment in the AI chip sector.

With its leading technology and superior product performance, NVIDIA's AI chips quickly gained widespread market recognition, becoming the preferred choice for many technology companies.

In stark contrast to NVIDIA's unparalleled success, some traditional chip giants such as Infineon and STMicroelectronics failed to keep pace with market changes and fell out of the top 10.

As a leading global provider of semiconductor solutions, Infineon boasts deep technical expertise and a substantial market share in automotive electronics and industrial electronics.

However, in the face of the rapid expansion of the AI chip market, Infineon's transformation pace was somewhat sluggish, failing to timely introduce competitive AI chip products, leading to its gradual decline in market competition.

Similarly, STMicroelectronics, a company with significant influence in the analog and power chip fields, failed to adjust its product strategy promptly in response to changes in market demand structure, resulting in its market share being gradually eroded by competitors.

Traditional chip vendors like STMicroelectronics and Infineon also face their respective challenges.

STMicroelectronics' declining market share in the analog and power chip market is primarily due to shifts in market demand and intensified competition.

With the ever-increasing demand for chip performance and functionality in sectors such as new energy vehicles and industrial automation, STMicroelectronics' products failed to keep up with the evolving market demand, leading to its market share being gradually taken over by competitors.

Infineon also encounters challenges in the analog and power chip market, with its automotive electronics business impacted by the global automotive market downturn, resulting in revenue decline.

Moreover, Infineon's transformation in the AI chip field is relatively slow, failing to seize market opportunities in a timely manner, leading to its disadvantage in market competition.

To counteract this, STMicroelectronics has increased its R&D investment in emerging application fields such as the Internet of Things and artificial intelligence, aiming to enhance its market competitiveness through technological innovation and product upgrades.

On the other hand, Infineon, through acquisitions and collaborations, has expanded its product line and market share while accelerating R&D and deployment in the AI chip field to adapt to market changes. However, the effectiveness of these measures remains to be seen over time.

These rises and falls profoundly reflect the monumental changes in the global chip market landscape.

Meituan Shines: Independent Variable Robot Secures Latest Round of Funding

Recently, the embodied intelligence company [Independent Variable Robot] successfully completed a Series A funding round of hundreds of millions of yuan, led by Meituan Strategic Investments with follow-on investment from Meituan Dragon Ball.

The funds raised from this round of financing will be utilized to continually propel the synchronous development of fully independently developed end-to-end general embodied intelligence large models and robot bodies, as well as to facilitate cooperation and implement intelligent solutions for multiple application scenarios in the future.

Since its establishment in early 2023, Independent Variable Robot has completed a total of seven funding rounds, raising over 1 billion yuan, attracting investments from top-tier institutions, industrial capital, and state-owned funds.

In February of this year, the company completed a Pre-A++ funding round led by Lightspeed China Partners and Legend Capital, with funds primarily allocated for the training and scenario application of the next-generation general embodied large model.

Shortly thereafter, it secured a Pre-A+++ funding round from Huaying Capital, Qiming Venture Partners, and GF Securities.

Apart from maintaining its strong financing momentum since the beginning of the year, the most significant highlight of this round of financing is the exclusive investment by Meituan.

Wang Qian, founder of Independent Variable Robot, explained to the media following the completion of this funding round that the company plans to implement and practice in domestic open service scenarios, dedicated to exploring the realization of service closed loops in diverse scenarios.

This investment in Independent Variable Robot constitutes a crucial aspect of Meituan's layout in the robotics field, aiding Meituan in building a full-stack technology system encompassing perception, decision-making, and execution, thereby achieving deep integration of robotics technology with its own business and enhancing service efficiency and user experience.

Technical Uniqueness: [End-to-End with Unified Cerebrum and Cerebellum] Approach

In the realm of embodied intelligence, the choice of technical approach is paramount.

Independent Variable Robot opted for the [End-to-End Large Model with Unified Cerebrum and Cerebellum] technical approach, which can be described as an innovative path.

While Unitree represents the core capabilities of humanoid robots and Galaxy General focuses on grasping and manipulation technology, Independent Variable Robot has emerged as an industry star with its [Unified, End-to-End Embodied Large Model].

On one hand, through direct input of multimodal data such as vision, hearing, and touch, the model can output execution parameters like speed, position, orientation, and torque, thereby controlling the entire action process, which is independently completed by a single model.

On the other hand, tasks such as reasoning and decision-making are no longer decomposed into independent training processes but are integrated into the same model, enabling reuse in multiple different scenarios including industry, commerce, and services.

A notable advantage of this unified strategy is that the model can spontaneously form a [common structure] during the training process, where skills learned in one task can be transferred to other tasks, significantly reducing the amount of data required for training.

Furthermore, this technical path is gradually gaining global acceptance.

Leading international companies such as Skild AI, which recently completed a $500 million funding round, and Physical Intelligence (PI), a benchmark enterprise in the field of embodied intelligence, are also exploring similar end-to-end large model solutions.

The [WALL-A] model in the [GreatWall] series of operational large models independently developed by Independent Variable Robot possesses powerful multimodal information fusion capabilities, integrating visual, language, and motion control signals to achieve a complete closed loop from perception to execution.

In terms of versatility and generalization, the WALL-A model performs exceptionally well, capable of generalizing and transferring various physical environmental variables and action patterns with minimal samples.

It also boasts an absolute advantage in long-sequence complex operations, achieving a task success rate of over 90% for intricate flexible operations such as folding and hanging clothes within a few minutes.

Moreover, the WALL-A model has demonstrated zero-shot generalization capabilities in multiple unseen physical scenarios, which is a crucial threshold for general robot capabilities, particularly in long-sequence, complex motion control, where this advantage is fully realized.

After continuous iteration, the capabilities of the WALL-A model are now on par with those of SkildAI and Physical Intelligence, with some capabilities even surpassing foreign competitors.

Meituan's Strategic Layout: Behind Exclusive Investment is a Wide Net

Meituan's company strategy of [Retail + Technology] has driven its active investment in embodied intelligence, a cutting-edge technology field.

For years, Meituan has been laying out its investment landscape in the robotics field, encompassing a range of robots from cleaning robots, hotel robots to collaborative robots. By integrating its own home delivery and store-based business scenarios, Meituan has quietly rolled out its strategic deployment.

Meituan's robot investment portfolio brings together robot enterprises with diverse technical approaches.

Given that the robot technology path has not yet converged, it is prudent for investors who do not want to miss the next star enterprise like Unitree Technology to diversify their investment risks.

From a technical perspective, the zero-shot generalization capabilities demonstrated by the large model WALL-A developed by Independent Variable may be the key factor in Meituan's choice of exclusive investment.

It is noteworthy that at the end of 2023, Meituan invested in Galaxy General Robotics for the second consecutive year.

Prior to this, Meituan had made multiple investments in the robotics field, with Unitree Technology being a notable investment target.

In 2024, Meituan and its subsidiary Dragon Ball Capital invested in Unitree Technology twice, albeit not as exclusive investors.

According to Tianyancha data, after multiple rounds of financing, Meituan's Hanhai Information Technology became the second-largest shareholder of Unitree, holding an 8% stake, second only to founder Wang Xingxing's 27%.

Earlier, in 2021, Meituan participated in the B+ round of funding for cleaning robot enterprise Gaoxian Robotics alongside investors such as Tencent and China Securities Co., Ltd.

In 2022, Meituan Dragon Ball Capital invested in the A round of collaborative robot manufacturer FAORobot.

And behind CloudMinds Technology, which plans to go public in Hong Kong this year, there is also the figure of Meituan.

Currently, Meituan is emphasizing the synergistic potential between robots and the company's business.

The application of robots in actual scenarios is still in the concept verification stage, and precise calculations for labor substitution and efficiency improvement cannot yet be made, so Meituan will continue to seek more opportunities in various scenarios.

In the field of embodied intelligence, Meituan's previous post-investment collaborations also reflect a strategy of [scenario empowerment].

After obtaining investment from Meituan, in September 2024, Meituan and Galaxy General Robotics signed a strategic cooperation agreement, and the two parties will jointly create the world's first humanoid robot intelligent pharmacy solution.

In March of this year, Galaxy General established two 24-hour [unmanned] pharmacies staffed by humanoid robots in Beijing and plans to open 100 more nationwide by the end of the year.

CloudMinds Technology also collaborated with Meituan Delivery last year, connecting the Meituan Delivery system with hotels equipped with robot services, utilizing robots to solve the problem of the [last 100 meters] of food delivery.

Given that the technical field has not yet reached a broad consensus, Meituan prefers to employ investment strategies to strengthen its ecosystem.

Conclusion:

Models require continuous optimization to enhance their perception, decision-making, and execution capabilities in complex environments, improving their adaptability to diverse tasks and scenarios.

Simultaneously, with the continuous advancement of technology, it is imperative to stay abreast of cutting-edge technological trends and innovate continuously to ensure technological leadership.

Market competition is another challenge that cannot be overlooked. The field of embodied intelligence has attracted numerous enterprises and capital, making competition increasingly fierce.

Scenario implementation also poses difficulties. There is a significant gap between laboratory research and development and actual commercial application. Different application scenarios have varied requirements for robot performance, reliability, safety, and other aspects.

Currently, there is a consensus in the industry: the efficiency of data collection is crucial for the survival of embodied intelligence models.

As an important direction for the development of robotics technology, embodied intelligence will continue to break through technological bottlenecks and realize the transition from specialized robots to general robots.