$Meta Platforms (META.US)$ To become the Android developer of the robotics industry.

According to the latest report by The Verge, Andrew Bosworth, Chief Technology Officer of Meta Platforms, revealed in an interview at the company’s headquarters that Meta’s investment in humanoid robots is comparable in scale to its previous focus on augmented reality—planning to invest tens of billions of dollars in related research and development. This initiative is directly driven by CEO Mark Zuckerberg, with the internal team code-named “Metabot.”

Bosworth disclosed that earlier this year, under the guidance of Zuckerberg, he formed a robotics “research team.” Although the existence of this team has been previously reported, Bosworth’s latest interview elaborated on Meta’s robotics strategy—focusing on software.

Bosworth frankly stated, “I don’t think hardware is the difficulty… I’m not saying hardware isn’t challenging, but it is not the bottleneck for robots; the bottleneck lies in software.” He noted that dexterous and precise operations are among the major challenges in robotics. “Many robots can stand, run, and flip because the ground is very stable.” In contrast, picking up a water cup is much more difficult for robots.

Bosworth envisions that Meta will license its software platform to other robot manufacturers, stating, “I don’t care if we become a hardware manufacturer.” Instead, Meta plans to adopt a model similar to Google’s—licensing its software to mobile device manufacturers. Bosworth emphasized that the core idea is: Meta will develop an open software platform allowing third-party hardware manufacturers to use, aiming to make the platform an industry-wide standard.

He also mentioned that Meta’s newly established “Superintelligence AI Lab” is collaborating with the robotics team to build a “world model” capable of simulating real-world physical laws, enhancing robots’ understanding and operational capabilities in their environment, and helping robots “conduct software simulations to achieve dexterous arm movements.” This could potentially expand to more complex actions and tasks in the future. The project is led by former Cruise CEO Marc Whitten, with several top experts, including MIT professor Sangbae Kim, also joining the team.

Ahead of Meta, Google has actively pursued creating “the Android system for the robotics industry” in recent years. Unlike its early efforts in developing robot bodies, Google now aims to provide powerful AI models for different robot manufacturers to purchase, akin to the Android system’s role in the mobile phone industry.

Software is the “brain” and “soul” of humanoid robots, while hardware determines the limits of a robot’s “physical capability,” and software dictates how much of that capability can be utilized in real-world applications. Currently, there is a consensus in the humanoid robotics industry: hardware development has relatively advanced, while the lag in software severely restricts the versatility, practicality, and commercialization progress of robots.

This is precisely why tech giants such as Meta Platforms and Google, along with numerous ventures, are focusing their efforts on software development—particularly in large AI models, reinforcement learning, and world models.

Huatai Securities recently released a research report stating that the robotics sector initially experienced thematic market dynamics. Subsequently, technological breakthroughs were gradually achieved in both software and hardware domains, leading to the emergence of relatively mature hardware solutions. These advancements have enabled applications in simple industrial scenarios and specialized use cases, transitioning the sector toward trend-driven market behavior. Core investment targets have expanded from the T-chain to domestic supply chains. The market’s ‘evaluation criteria’ for robotics has evolved from an early focus on ‘motion control capabilities demonstrated in video edits.’ The institution believes that the ultimate outcome of intense competition in the robotics industry may involve major players engaging in strategic rivalry. Leading technology giants may be the first to achieve breakthroughs in general-purpose embodied intelligence models, while automotive manufacturers and large-scale manufacturing enterprises could pioneer the large-scale deployment of B2B humanoid robots. A small number of ventures pursuing differentiated strategies may also emerge. Progress of companies with the potential to implement general-purpose humanoid robots should be closely monitored going forward.

Guotai Haitong Securities noted that the defining characteristic of embodied intelligence is the ability to understand and interact with the physical world. Compared to traditional robot control systems, the core transformation of embodied intelligence lies in enabling autonomous decision-making, evolving robots from machines executing specific tasks based on instructions into intelligent agents continuously interacting with the world. Therefore, large models and data represent the most critical components of humanoid robots.

Editor/KOKO