China is moving to formalize its humanoid robot industry with the first regulatory approvals for physical AI models, coinciding with a massive $200 million angel round for Beijing's Anyverse Dynamics. Meanwhile, major tech players are bringing chip design in-house and acquiring software firms to reshape the hardware foundation for the entire sector.
Beijing-based Anyverse Dynamics, a general embodied intelligence robotics company founded in 2025, announced on Friday it has closed an angel funding round exceeding $200 million. The company, which is already mass-producing its K15 humanoid robot, stated a Pre-Series A round of nearly $200 million is also nearing completion. Anyverse has reportedly secured substantial commercial orders for its robots, which are powered by a proprietary 'Multimodal World Architecture Embodied General Brain' (MWA) technology.
Why it matters
This massive, rapid-fire funding for a one-year-old company underscores the extreme velocity of capital deployment into China's humanoid robotics sector. For an entrepreneur in the space, Anyverse's claimed ability to move from founding to mass production and significant commercial orders in under two years sets a new, aggressive benchmark for operational speed. If their MWA technology proves effective, it could represent a significant leap in embodied AI, but the speed of these claims also warrants skepticism until independently verified deployments and performance data are available.
The significant funding and rapid progress claimed by Anyverse Dynamics showcase the intense investment and innovation within China's embodied AI and robotics sector. The company’s unique MWA technology and swift transition to mass production, if validated, could establish new industry standards for efficiency and adaptability, impacting sectors from manufacturing to logistics.
Early-stage investment in South Korean startups has become heavily concentrated in AI and robotics, which accounted for 54.7% of funding through May of this year, according to a report on Saturday. This is the first time these sectors have captured over half of early-stage investment in the country. The trend is fueled by mega-seed rounds, such as the 40 billion won (approx. $28.8M) raised by Config Intelligence, as the nation strategically positions itself in the global robotics race.
Why it matters
This intense concentration of capital signals South Korea's all-in bet on becoming a global robotics leader, backing up the Goldman Sachs forecast that it will account for 30% of humanoid production by 2035. For an entrepreneur, this makes South Korea a critical market to watch for talent, components, and potential partnerships. The flood of capital will accelerate development, but it also raises the risk of a bubble and creates intense competition for resources and market share.
The surge in funding is seen as a strong indicator of investor confidence in the future growth of AI and robotics in South Korea. This capital influx highlights both the increasing demand for advanced robotic solutions and the fierce competition for funding in other, less-hyped industries within the country's startup ecosystem.
As we noted alongside its purchase of Fauna Robotics, Amazon has acquired Swiss firm RIVR (formerly Swiss-Mile) to bolster its last-mile delivery capabilities. The strategic focus on a hybrid wheel-leg design—capable of combining the efficiency of wheels with the obstacle-traversal capabilities of legs—aims to better navigate complex urban environments.
Why it matters
Amazon's acquisition of RIVR, following its recent purchase of Fauna Robotics, signals a clear strategy of buying, not just building, key robotics technology and talent. The focus on a hybrid wheel-leg design suggests that pure-wheeled or pure-legged solutions are not fully solving the challenges of last-mile urban delivery. This move could accelerate the deployment of more versatile and robust delivery robots, further consolidating Amazon's dominance in logistics automation.
This acquisition is expected to redefine the economics and operational strategies of last-mile delivery. By integrating RIVR's technology, Amazon is signaling a move to transition robotics from experimental prototypes to scalable, real-world applications, potentially lowering delivery costs and reshaping labor roles in logistics.
Following its initial demonstration on Friday, Chinese startup MindOn Robotics provided more details on Sunday about its Mind-0, a unified AI foundation model capable of controlling both a Unitree G1 humanoid robot and a traditional dual-arm robotic system. The model uses a modular, two-stage architecture that separates high-level reasoning from low-level motion control. It is trained on human-centric data and employs a 'Real-World Execution Compensation Model' to bridge the sim-to-real gap, allowing different robots to coordinate on complex tasks.
Why it matters
The ability to use a single AI model to control heterogeneous robot hardware is a significant step toward general-purpose robotics. This hardware-agnostic approach could drastically reduce development time and costs, enabling more flexible and scalable automation. For a robotics entrepreneur, MindOn's architecture offers a potential blueprint for overcoming the costly process of creating bespoke control software for every new robot platform, accelerating the path to deploying collaborative, multi-robot systems.
Mind-0's architecture is seen as a key innovation in embodied intelligence. By separating reasoning from motion, the system becomes more adaptable. This could enable more flexible AI systems that can control a variety of robotic platforms, representing a scalable alternative to traditional teleoperation and fostering more advanced human-robot collaboration. The approach is particularly promising for accelerating the deployment of humanoids and robotic arms in diverse applications.
The Beijing Humanoid Robot Innovation Center announced on Saturday that its dual foundational AI models, 'TianHu' and 'WoWu', have received regulatory approval in China. This marks the first time the country's regulators have issued generative AI licenses specifically for models designed to control autonomous physical hardware. The approval is a key part of China's strategy to standardize and scale its humanoid robot industry.
Why it matters
This is a significant milestone in the global race for physical AI dominance. By creating a formal regulatory pathway for robotics-specific AI models, China is building the institutional infrastructure needed for mass commercial deployment. While Western firms navigate a patchwork of general AI regulations, China is creating a standardized, state-sanctioned software layer for its industrial robotics push. This could accelerate their deployment timeline and create a new kind of competitive advantage built on regulatory efficiency.
This regulatory approval is seen as a critical step in China's strategy to gain an edge in the global physical AI rivalry. It establishes a structured pathway for commercial automation, potentially giving Chinese companies a head start in deploying and scaling humanoid robot solutions compared to their Western competitors.
Hugging Face on Sunday introduced VLX-Go, a new lightweight vision-language waypoint planner for embodied navigation. The model is designed to process visual history, current frames, and natural-language instructions to predict short-horizon local waypoints. This enables robots to perform more effective target following, local navigation, and dynamic obstacle avoidance.
Why it matters
VLX-Go offers a practical approach for integrating multimodal understanding with robot control, a core challenge in embodied AI. Its focus on short-horizon waypoint prediction and a closed-loop design allows for more robust and adaptable robot behavior in dynamic, real-world environments. As an open-source tool released on Hugging Face, it provides the robotics community with an accessible building block for developing more sophisticated navigation systems, potentially accelerating the transfer of research from simulation to physical robots.
Developers see VLX-Go as a significant step in making embodied AI more practical for real-world deployment. By providing a lightweight, open-source model that handles complex navigation tasks, Hugging Face is lowering the barrier for researchers and startups to experiment with and build upon advanced robotic navigation techniques.
Researchers at Cornell University have developed a method called Direct Electrode-to-Electrode Regeneration (DEER) that can restore worn-out lithium-ion batteries. The process involves submerging the battery in an electrochemical solution that dissolves the solid electrolyte interphase (SEI) layer, a primary cause of capacity loss. A report on Sunday states that initial tests show restored batteries regaining up to 95% of their original capacity, with the potential to reduce recycling costs by 56%.
Why it matters
Battery life, cost, and sustainability are critical limiting factors for the widespread deployment of mobile robots. This breakthrough offers a path to significantly extend the operational life of robotic systems and reduce their total cost of ownership. For a robotics entrepreneur, this could make battery-powered robots more economically viable and environmentally sustainable, opening up new applications where battery replacement cycles were previously a major barrier.
The DEER method is being hailed as a potentially transformative approach to managing the lifecycle of lithium-ion batteries. By reviving batteries instead of dismantling them for recycling, the process could significantly mitigate environmental impact and resource consumption. This is particularly crucial as the demand for batteries for EVs and robotic systems continues to soar.
Scientists in China have developed a new ceramic lithium-ion micro battery that offers significantly enhanced safety, thermal stability, and energy efficiency. According to a report on Saturday, the battery uses a ceramic-based solid structure, which eliminates the risk of electrolyte leakage and greatly reduces fire hazards associated with traditional lithium-ion batteries. The new design can reportedly operate safely at temperatures up to 150°C while maintaining high energy density.
Why it matters
Battery safety is a major concern in robotics, especially for systems operating in close proximity to humans or in sensitive environments. This breakthrough in solid-state battery technology addresses the critical risks of overheating and fire. For a robotics entrepreneur, safer, more thermally stable batteries could unlock new applications and simplify the safety certification process for commercial robots, making them more reliable and robust.
This development is a significant step forward in battery technology, addressing long-standing safety issues. Its enhanced durability and stability in extreme environments make it a promising candidate for applications in miniaturized electronics, electric vehicles, and aerospace, where safety and reliability are paramount.
A report on Sunday indicates that nearly 20 Chinese motor manufacturers are in a competitive race to supply critical components for the burgeoning humanoid robot market. The focus is on two key areas: frameless torque motors for joint actuation and hollow cup motors for dexterous hands. Companies like Leadshine Technology and Step Electronics are reporting significant growth in sales for these specialized components as robot manufacturers ramp up production.
Why it matters
The intense competition at the component level is a strong leading indicator of an industry preparing for mass production. For a robotics entrepreneur, this is a double-edged sword: it signals a maturing and potentially commoditizing supply chain, which can lower costs, but it also highlights China's rapid consolidation of critical hardware dependencies. Securing a reliable supply of high-performance actuators remains a key challenge, and this boom in China could make it harder for Western firms to compete on both price and access.
This race to supply motors highlights a crucial bottleneck and opportunity in the humanoid robot supply chain. As the industry moves towards mass production, the ability to manufacture high-quality, specialized motors at scale will be essential. This is expected to drive further innovation and potential consolidation among component suppliers.
Figure released a new demonstration on Sunday showing its humanoid robots autonomously performing complex, coordinated household chores. Powered by the company's Helix AI model, the robots were shown making a bed together and hanging a coat. The demonstration highlighted the system's Vision-Language-Action (VLA) capabilities, which enable real-time adaptation and implicit communication between robots to handle deformable objects like blankets.
Why it matters
While many humanoid demos focus on single-agent tasks, this demonstration of multi-agent coordination on a non-rigid object problem is a significant step forward. It showcases an ability to handle the unpredictability of the real world, a key challenge for domestic robotics. For the consumer robotics market, this points toward a future of more capable and genuinely helpful home assistants, though the timeline for commercial viability remains a major question.
The demonstration is being viewed as a significant leap in domestic automation and human-robot collaboration, showcasing the robots' ability to handle complex and unpredictable tasks. Experts note that the seamless coordination and handling of deformable objects could lead to widespread deployment of humanoid robots in daily life, transforming household chores and potentially many other sectors.
DeepReinforce has released Ornith-1.0, an open-weight family of coding models with a novel capability: it can generate its own reinforcement learning (RL) scaffolds for training. According to the release notes from Thursday, this approach allows the model to dynamically create and adapt its own problem-solving strategy for coding tasks without needing human engineers to pre-define the RL training harness.
Why it matters
This is a fascinating development in 'self-improving' AI. If the self-authoring scaffold mechanism proves robust, it could significantly reduce the human expertise and effort required to design complex RL training pipelines. For the open-source robotics community, this could accelerate the development of more capable and adaptable AI agents by automating a tedious but critical part of the training process, allowing developers to focus on higher-level goals rather than intricate reward-shaping.
The release is being met with interest in the open-source AI community. The ability of a model to essentially design its own training regimen could represent a paradigm shift in reinforcement learning. However, experts caution that the robustness and generalizability of this self-authoring capability need to be thoroughly tested across a wide range of problems before its practical impact can be fully assessed.
A pop-up store named KOID Shop has opened in Lower Manhattan, New York City, featuring a range of general-purpose robots, with a notable presence of Chinese manufacturers. The store, a joint venture by KraneShares and OpenMind, showcases interactive humanoid robots to the public, demonstrating the current state of robotics technology and the growing influence of open-source AI platforms.
Why it matters
Bringing advanced, interactive robots directly to the public in a retail setting is a significant step for familiarizing consumers and normalizing human-robot interaction. For the open-source robotics community, this kind of public showcase provides invaluable feedback and demonstrates the maturity of open platforms. It also visibly highlights the increasing competitiveness of Chinese hardware in the global market, putting it directly in front of Western consumers and developers.
The pop-up is seen as both a commercial and educational initiative. It aims to gauge public interest and reaction to general-purpose robots while also showcasing the rapid advancements in embodied AI. The strong presence of Chinese brands and open-source systems underscores the global nature of the robotics race and the importance of accessible platforms in driving innovation.
Starship Technologies is strategically shifting the focus of its autonomous delivery robots away from college campuses and toward retail grocery chains and hot food delivery. The company has begun pulling its robots from universities like Bowling Green State University, citing a move to tap into the larger and more lucrative grocery market, which has seen a surge in online ordering. The company notes that robot-led grocery deliveries are already seeing significant traction in markets like Finland.
Why it matters
This pivot illustrates the challenging unit economics of campus delivery and the search for a more sustainable business model in the last-mile robotics space. For entrepreneurs in this sector, Starship's move is a crucial case study in market adaptation. It suggests that while controlled environments like campuses are good for initial testing, the real commercial opportunity lies in integrating with the high-volume, high-frequency logistics of grocery and food delivery.
Analysts view Starship's strategic shift as a natural evolution in the autonomous delivery market. College campuses served as effective incubators for the technology, but the larger, more consistent demand from grocery and retail partners offers a clearer path to profitability and scale. This move highlights the need for robotics companies to be agile and responsive to market demands to find their optimal niche.
Qualcomm formally confirmed its $3.92 billion all-stock acquisition of Modular, detailing the strategic bet we flagged earlier this week: breaking NVIDIA's CUDA software ecosystem. Modular, co-founded by LLVM creator Chris Lattner, has been developing a unified compiler and runtime that allows AI models to run efficiently across various hardware, directly targeting the vendor lock-in that has fortified NVIDIA's dominance.
Why it matters
This is one of the most significant strategic moves in the AI hardware wars, representing a direct assault on NVIDIA's software moat. For a robotics entrepreneur, the success of this play could be transformative. A hardware-agnostic AI development platform would drastically increase flexibility in choosing processors for robots, particularly at the edge, potentially lowering costs and fostering a more competitive and innovative market for specialized robotics chips. This shifts the competitive landscape from a battle of silicon to a battle of software portability.
Analysts see the acquisition as a bold attempt to democratize AI hardware and create a viable alternative to NVIDIA's closed ecosystem. By providing a hardware-agnostic platform, Qualcomm and Modular aim to empower developers to run AI models on a diverse range of hardware without costly and time-consuming rewrites. This could foster greater innovation by lowering the barrier to entry for new and specialized silicon.
Confirming reports from last Wednesday, OpenAI officially unveiled 'Jalapeño,' its first custom-built AI inference processor, developed in partnership with Broadcom. The chip is specifically designed for large language model inference and is part of OpenAI's strategy to reduce its reliance on NVIDIA GPUs. The company aims to optimize performance-per-watt for its gigawatt-scale data center deployments and plans to begin deploying the new chips later this year.
Why it matters
The 'Jalapeño' chip marks a major step in the trend of vertical integration among AI leaders. By designing custom silicon, OpenAI gains greater control over its cost structure, performance, and technology roadmap, reducing its dependence on a single supplier. For the robotics industry, this trend among hyperscalers could accelerate the development of more efficient and specialized chips, with innovations eventually trickling down to edge devices and robotic platforms.
The introduction of custom AI silicon like 'Jalapeño' reflects a growing industry trend among hyperscalers to optimize their AI infrastructure and gain greater control over their technology stack. This could reshape the AI chip market, creating new opportunities for specialized hardware providers and potentially leading to more efficient and scalable AI solutions. Others see this as a defensive move to control runaway inference costs and supply chain risks associated with relying on a single dominant vendor.
Building on the 5,000 units we noted it shipped in 2025, embodied AI firm AGIBOT announced its 15,000th robot has rolled off the production line. The milestone unit, an industrial-grade AGIBOT G2, follows a recent six-day livestream demonstrating the wheeled humanoids' high success rate in electronics factory inspection tasks.
Why it matters
While much of the recent focus has been on bipedal humanoids, AGIBOT's milestone demonstrates the significant scale already being achieved with wheeled and 'semi-humanoid' platforms in industrial settings. This highlights a pragmatic, near-term path to commercialization for embodied AI that doesn't require solving the complexities of bipedal locomotion. For an entrepreneur, this serves as a reminder that market-specific form factors can often scale faster than general-purpose ones.
AGIBOT's achievement is seen as a key indicator of the rapid industrialization and commercial viability of embodied AI robotics. The company's ability to move from product validation to mass production signifies a maturing supply chain and the software-hardware integration necessary for large-scale, real-world deployment, shifting the industry's focus toward scalability.
Researchers have developed a novel soft pneumatic gripper featuring integrated 'teeth' designed to securely grasp irregularly shaped objects, particularly in food handling applications. A paper published on Saturday in Nature Scientific Reports details the design, manufacturing, and validation of the gripper. The device, powered by air pressure, was evaluated for its grasp success rate, pick-and-place performance, and load capacity with objects like produce.
Why it matters
Grasping and manipulating delicate, irregular objects remains a significant challenge in robotics, especially in the food and agriculture industries. This innovation offers a promising solution that combines the compliance of soft robotics with the enhanced grip of a toothed structure. For automation in these sectors, such a gripper could significantly reduce product damage and improve efficiency, opening up new possibilities for robotic harvesting and packaging.
This development is a notable advancement in the field of soft robotics, addressing a practical need in industries where traditional rigid grippers are often unsuitable. The design offers a new approach to automated handling that could lead to safer and more efficient systems for a variety of applications beyond food processing, where gentle but firm grasping is required.
Expanding on the MIT gel we covered last week, new details highlight the material's broader implications for soft robotics. By incorporating photo-ion generators (PIGs) into polyurethane rubber—allowing the gel to increase its electrical conductivity by a factor of 400 when exposed to light—the team is laying the groundwork for a new subfield dubbed 'soft photo-ionotronics'.
Why it matters
This is a foundational breakthrough for soft robotics and human-machine interfaces. The ability to dynamically and locally control the ionic conductivity of a soft material using light opens the door to creating adaptive systems that can respond to environmental stimuli in real-time. For robotics, this could lead to more sophisticated soft sensors, actuators, and even self-healing materials that can be controlled with precision using a non-invasive light source.
The technology is seen as having vast potential for revolutionizing wearables, bio-integrated electronics, and soft robotics. By enabling dynamic control over local ion populations, it paves the way for self-adaptive systems, essentially creating a new subfield of 'soft photo-ionotronics' with broad industrial implications.
The first-ever global technical regulation for automated driving systems (ADS GTR) has been officially approved, according to a report from Friday. The regulation was co-led by China, the EU, UK, US, Canada, and Japan. It establishes a common set of technical requirements for automated driving products, covering safety management, testing, validation, and post-use vehicle safety.
Why it matters
While the UN adopted a framework last week, this GTR approval represents a more concrete and harmonized set of technical standards that manufacturers can build to. This is crucial for the safe and consistent deployment of autonomous vehicle technology worldwide. For companies in the AV space, it provides a clearer, more unified regulatory target, which can reduce development costs and accelerate the path to market by ensuring a baseline level of safety and interoperability across major international markets.
This approval is seen as a landmark achievement for the autonomous vehicle industry. By harmonizing technical standards, it provides a stable and predictable regulatory environment that is expected to foster innovation and build public trust. It addresses a key hurdle for global automakers who previously faced a fragmented landscape of national and regional rules.
Following up on the light-controlled DNA droplets we tracked last week, researchers at the Institute of Science Tokyo provided new mechanical details. By incorporating light-responsive molecules, the team can make the droplets change shape and mimic the swimming motion of a jellyfish, allowing for precise microscale maneuvering and cargo transport.
Why it matters
This innovation represents a new paradigm for controlling matter at the microscopic level without mechanical parts. The ability to use light to direct movement and transport cargo opens up significant possibilities for targeted drug delivery, mixing in microfluidic systems, and developing intelligent, soft microrobots, pushing the boundaries of what's possible in medicine and biotechnology.
The breakthrough transforms DNA from a passive carrier of genetic information into a programmable material for building dynamic micro-machines. This offers a simple yet powerful method for controlling complex tasks at the microscopic level, with potential applications ranging from advanced medical treatments to the fabrication of novel materials.
China's Humanoid Ecosystem Matures with Funding and Regulation The Chinese robotics sector is seeing immense activity, with Anyverse Dynamics raising over $200M for its mass-produced humanoids, the Beijing Humanoid Robot Innovation Center securing the country's first generative AI licenses for robotics, and nearly 20 motor makers competing to supply critical components for the rapidly scaling market.
Big Tech Redefines AI Investment Strategy Major technology firms are evolving their approach to the AI boom. Instead of outright acquisitions, they are opting for minority investments and complex partnerships. Concurrently, players like OpenAI, Google, and Amazon are bringing chip design in-house to reduce costs and reliance on third parties, a trend highlighted by the recent unveiling of OpenAI's 'Jalapeño' chip.
South Korean Robotics Attracts Heavy Investment South Korea is solidifying its position as a robotics powerhouse, with AI and robotics startups attracting over half of all early-stage investment in the country through May. This funding surge is fueling the nation's ambition to capture 30% of the global humanoid production market by 2035, leveraging its strong automotive and component manufacturing base.
The Hardware-Software Stack for Robotics Evolves The focus in robotics is increasingly on the full stack. On the hardware side, breakthroughs in battery technology promise safer and more sustainable power sources, while new sensors are enhancing robot perception. On the software side, unified AI models like MindOn's Mind-0 aim to create hardware-agnostic control systems, and open-source frameworks continue to democratize access to advanced capabilities.
Healthcare Robotics Advances on Multiple Fronts The medical robotics field is seeing significant progress, from Vinmec's launch of Vietnam's first integrated robotic surgery network to the development of novel systems in China and at Johns Hopkins. Innovations also extend to the micro-scale, with new algae-based microbots being developed for targeted cancer drug delivery.
What to Expect
2026-07-01—Paper published on design and prototyping method for pre-shaped soft growing robots using material patching.
2026-07-05—'Dielectric Elastomers Based Soft Machines' book to be published by Springer Verlag.
2026-09-27—IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2026 begins in Pittsburgh, PA.
How We Built This Briefing
Every story, researched.
Every story verified across multiple sources before publication.
🔍
Scanned
Across multiple search engines and news databases
461
📖
Read in full
Every article opened, read, and evaluated
208
⭐
Published today
Ranked by importance and verified across sources
20
— The Robot Beat
🎙 Listen as a podcast
Subscribe in your favorite podcast app to get each new briefing delivered automatically as audio.
Apple Podcasts
Library tab → ••• menu → Follow a Show by URL → paste