AI and Robotics Daily Wrap: Humanoid Focus
SoftBank Acquires ABB Robotics for $5.4 Billion to Advance ‘Physical AI’
The global robotics landscape saw a major consolidation this week with investment giant SoftBank Group announcing its definitive agreement to acquire ABB Ltd.’s robotics business for approximately \$5.375 billion. The acquisition, which is expected to close in mid-to-late 2026 pending regulatory approvals, is a significant move in SoftBank’s strategic shift toward what its CEO, Masayoshi Son, has termed “Physical AI”.
The Swiss-Swedish firm ABB Robotics specializes in industrial robots, particularly large robotic arms used for precise movements in factory settings. By combining ABB Robotics’ leading technology and deep industry expertise with SoftBank’s state-of-the-art capabilities in AI, robotics, and next-generation computing, the company aims to accelerate innovation in AI robotics. SoftBank, already a major investor in AI firms like OpenAI, views this acquisition as essential to its broader mission of realizing Artificial Super Intelligence (ASI).
The acquisition is part of SoftBank’s expanding business focus on four essential areas: AI chips, AI robots, AI data centers, and energy. SoftBank plans to integrate the ABB business with its existing robotics-related investments, which include companies such as SoftBank Robotics Group, Berkshire Grey, and Agile Robots, to drive progress toward the realization of its ASI mission. The move underscores the growing strategic importance of integrating advanced AI with physical hardware, even in traditional industrial automation sectors, setting the stage for more intelligent and autonomous systems, including humanoids, to operate in real-world environments.
* **Key Points:**
* SoftBank Group is acquiring ABB Ltd.’s robotics business for \$5.375 billion.
* The deal is central to SoftBank’s strategy to advance “Physical AI” and fuse Artificial Super Intelligence (ASI) with robotics.
* The acquisition will merge ABB’s industrial robotics expertise with SoftBank’s AI and computing capabilities.
* The transaction is anticipated to close in mid-to-late 2026.
Figure Unveils ‘Figure 03’ Humanoid Robot with Advanced Vision and Dexterity
Figure AI, a California-based startup, has debuted its newest humanoid model, the Figure 03, designed to be a general-purpose android for both factory floors and domestic use. The new machine is built around the company’s proprietary AI model, **Helix**, which combines vision, language, and action capabilities to enable the robot to rapidly learn and master everyday chores. The Figure 03 boasts significant hardware upgrades, particularly in its visual and manipulative systems.
The robot’s head-based cameras feature a higher frame rate, lower latency, and a wider field of view, which the company claims is essential for intelligent navigation and precise manipulation in complex, cluttered environments like homes. A novel feature is the inclusion of new cameras in the robot’s palms, providing “visual awareness even when the main cameras are occluded” when reaching into confined spaces. This hand-eye coordination is further enhanced by new pressure sensors in the hands that are sensitive enough to register the weight of a paperclip, allowing for better grappling of fine-textured materials and detection of slippage.
Figure 03’s design moves closer to a consumer aesthetic, featuring a sleek, fabric-wrapped design with swappable knitwear covers, a deliberate choice to “soften the uncanny valley” as the company designs for living room environments. The robot is powered by a new high-density battery pack, offering up to five hours of runtime, with inductive charging on the heels for seamless power-ups. This reveal follows the successful deployment of the previous Figure 02 model, which has reportedly been assisting on the BMW X3 production line for five months, operating for up to 10 hours per day.
Google DeepMind Unveils AI Pair for Complex Robot Reasoning
Google DeepMind has introduced a pair of new artificial intelligence models designed to significantly boost the intelligence and reasoning capabilities of robots, enabling them to perform complex, multi-step tasks that were previously out of reach. The new system acts as a supervisor and worker, allowing general-purpose robots, including humanoids, to better understand and interact with the physical world.
The system comprises two models: **Google Robotics-ER 1.5** (the “brain”) and **Google Robotics 1.5** (the “hands and eyes”). Google Robotics-ER 1.5 is a vision-language model (VLM) that processes natural language commands, gathers information about a space, and uses advanced reasoning and tools to generate instructions. Google Robotics 1.5 is a vision-language-action (VLA) model that translates those instructions into physical actions and movements for the robot’s hardware.
In a demonstration, the robot was able to perform tasks such as sorting fruit onto plates based on shared characteristics, spatially locating objects, identifying colors, and then providing a natural language output to describe its reasoning. This level of sophisticated capability moves beyond simple automation and is crucial for general-purpose humanoids that need to operate in unstructured and unpredictable environments, such as homes and offices. The new AI models are an evolution of the company’s earlier Gemini Robotics model, specifically specialized for robotics, and are intended to power more sophisticated humanoids for complicated daily tasks.
China Accelerates Humanoid Robot Deployment with Focus on Industrial ‘Vocational Schools’
China’s robotics sector is rapidly commercializing humanoid technology, moving from laboratory demonstrations to practical, real-world applications in a strategy dubbed “China Speed”. The Chinese government has actively promoted this development through initiatives like the “Robot+ Application Action Plan,” fostering a tripartite alliance between the government, companies, and research institutions to build a vertically integrated industrial ecosystem.
Instead of focusing solely on viral acrobatics, Chinese robot makers are prioritizing the deployment of their machines in “vocational schools”—real-world industrial settings like automotive plants, warehouses, and eldercare facilities—to bridge the data shortage and train them for practical skills. This pragmatic approach has resulted in significant success rates in tasks such as returning empty crates, sorting materials, and weighing, packing, and boxing wares.
For example, Leju Robotics’ Kuavo robots have already achieved success rates over 95 percent in practical tasks in these test environments. Additionally, the country is seeing rapid commercialization milestones, with models from Unitree and Agibot recording cumulative shipments of 1,000 units, and UBTECH’s humanoid robot expected to sell 500 units this year. The country’s strategy leverages its strengths in manufacturing and its established supply chain, which includes specialized component makers like Manifold Tech, a startup that transformed to become a dedicated “robot eye” manufacturer to meet the industry’s growing needs. This aggressive push is aimed at accessing a projected trillion-dollar market, with one think tank forecasting over 100 million humanoid robots deployed across Chinese industries by 2045.