NVIDIA is making an aggressive bid to own the entire physical AI stack today, launching both the new Cosmos 3 'omnimodel' and a line of power-efficient Jetson Thor modules. That full-stack approach lands just as the broader industry coalesces around competing control platforms, including Google's ongoing Gemini Robotics 2 rollout and a direct hardware challenge from an AMD-Analog Devices partnership.
Figure AI has reportedly formed a new partnership with Nike to deploy its humanoid robots in the apparel giant's manufacturing and warehouse operations. The collaboration, reported Monday, signals growing enterprise confidence in using humanoid robots as practical solutions for industrial automation.
Why it matters
This partnership with a global consumer brand like Nike is a significant milestone for the commercial adoption of humanoid robots. Following its deployments with BMW, this move suggests Figure AI is successfully moving its robots from pilot programs to mainstream industrial applications. It underscores the perceived value of physical AI in improving supply chain efficiency and addressing labor shortages in large-scale manufacturing and logistics, further cementing Figure's position as a leading contender in the race to commercialize humanoids.
The Enterprise World views this as a sign of growing enterprise confidence in humanoids. The Motley Fool points to this as another proof point for Figure's go-to-market strategy, which focuses on securing major industrial partners early on.
Chinese automotive giant BYD confirmed it will debut its first humanoid robot in early August. The initial deployment will be in the company's 'Di Space' showrooms, where the robots will greet customers and explain vehicle features. Following this, BYD plans to use the robots for factory automation tasks, with a mass rollout targeted within one to two years. The company is leveraging its existing expertise in batteries, motors, and control systems from its automotive manufacturing.
Why it matters
BYD's entry into the humanoid space is significant. By leveraging its vertically integrated manufacturing infrastructure, the company may have a considerable cost and scaling advantage over competitors. Starting with a customer-facing role in showrooms is an unconventional go-to-market strategy that could help normalize human-robot interaction in public spaces, while also serving as a real-world data collection and testing ground before wider industrial deployment. This move intensifies the competition among automakers like Tesla and Hyundai who are also investing heavily in robotics.
The Autowire emphasizes the strategic advantage of repurposing automotive manufacturing infrastructure. WinFuture frames this as a direct challenge to Tesla, highlighting BYD's focus on interactive demonstrations over purely technical presentations.
As Hyundai formalizes its acquisition of Boston Dynamics and the dedicated US robotics factory we've been tracking, the automaker has committed to deploying over 25,000 Atlas humanoid robots across its manufacturing plants by 2028. However, this massive rollout is facing immediate, organized opposition from the Korean Metal Workers’ Union, which is blocking deployment pending a new labor-management agreement.
Why it matters
This story encapsulates both the immense ambition and the significant real-world friction of the humanoid robot revolution. Hyundai's massive deployment number signals a firm belief in the technology's readiness for large-scale manufacturing. However, the immediate and organized labor opposition highlights that the biggest hurdles to widespread automation may be social and political, not just technical. How Hyundai navigates this dispute could set a major precedent for labor relations in the age of humanoid workers.
GCN highlights this as a defining moment for integrating humanoids into automotive manufacturing. The ongoing labor dispute has been a recurring theme, with unions demanding job security guarantees and a say in how the robots are implemented.
The consumer robotics market for outdoor maintenance is seeing a wave of new, sophisticated products. On Monday, Ecovacs launched a full suite of outdoor robots, including two robotic lawn mowers (GOAT A2000 and O600) with LiDAR and RTK navigation, a window cleaner, and a pool cleaner. Concurrently, Roborock launched its RockMow X120H mower, which relies on 360-degree LiDAR, while Yarbo is introducing a modular system that can swap between mowing, snow blowing, and leaf collection.
Why it matters
This product blitz signals a major expansion of consumer robotics beyond indoor vacuums into the more complex and unstructured environment of the yard. The use of advanced navigation technologies like LiDAR and RTK, previously found only in high-end research or industrial robots, is becoming standard. The modular approach from companies like Yarbo also points to a future of integrated home robotics platforms rather than single-purpose gadgets. For the consumer robotics industry, this is a significant step toward delivering on the long-held promise of a truly automated home.
A buyer's guide from Glass Almanac highlights that RTK, LiDAR, and automated mapping are now essential features for high-end robotic mowers. First Class Checkin praises Yarbo's modular system for offering long-term flexibility, while Marine Life Photo notes Roborock's bet on LiDAR-only navigation comes at a premium price point, testing consumer appetite for high-tech lawn care.
Open Robotics on Sunday announced the release of a new package, `ros2_realtime_support`, that provides robust, real-time functionality for the core ROS 2 client libraries (rcl and rclcpp). The package is designed to provide a unified mechanism for real-time support without requiring modifications to the core ROS packages, a long-standing challenge for the community.
Why it matters
This is a critical development for the professional and industrial adoption of ROS 2. True real-time performance is essential for applications requiring high-speed, deterministic control, such as advanced manufacturing, autonomous navigation, and safety-critical systems. By providing an official, non-intrusive solution, Open Robotics is making it significantly easier for developers to build more reliable and responsive robotic systems, solidifying ROS 2's position as a production-grade framework.
The announcement on the official ROS Discourse forum was met with positive feedback from developers who have been seeking a standardized approach to real-time capabilities. A related field report from another developer details successful implementation of workspace-level virtual environments in ROS 2 using the `uv` package manager, indicating the ecosystem is rapidly adopting modern software engineering practices to improve dependency management and reproducibility.
Following last week's unveiling of Google DeepMind's Gemini Robotics 2, new demonstrations are highlighting the system's ability to execute fluid, multi-step tasks like replacing a light bulb or tying a garbage bag. By utilizing the suite's vision-language-action (VLA) and embodied reasoning (ER) models we tracked, robots like Apptronik's Apollo 2 can coordinate their entire bodies for complex chores rather than controlling limbs in isolation.
Why it matters
This release represents a significant move to create a universal, hardware-agnostic 'brain' for robots. By decoupling the intelligence layer from the physical body, Google could commoditize robot hardware, making the AI model the primary source of value. For robotics developers and entrepreneurs, this could drastically reduce the time and data needed to bring new robots to market, as they could license a powerful, pre-trained control system instead of building one from scratch. The focus on whole-body coordination, rather than controlling limbs in isolation, is a critical step toward more fluid, human-like motion and capability. This positions Google's software as a potential industry standard, especially as partners like Apptronik begin integrating it into their humanoids.
Analysts at Value Add VC note this is a direct challenge to vertically integrated players like Tesla, shifting the battleground from hardware to the foundation model layer. Logistics Viewpoints highlights that this could create a more portable intelligence layer, reducing the engineering effort needed to make different machines useful in supply chain environments. TechRound frames the announcement as an aggressive pivot by DeepMind toward 'physical AGI,' aiming for broad physical adaptability that could revolutionize industries but also raises questions about AI safety and containment.
A new AI memory compression technique from NVIDIA could enable Tesla to deploy its next-generation Full Self-Driving (FSD) V14 software on older vehicles equipped with HW3 hardware. According to a report on Monday, NVIDIA's KV Cache Transform Coding (KVTC), which claims to compress a large language model's working memory by up to 20x, would dramatically reduce the VRAM requirements for advanced FSD models. The analysis also points to a potential Tesla-SpaceX merger by 2027, driven by a joint chip manufacturing facility known as Terafab.
Why it matters
This breakthrough addresses a major bottleneck in deploying ever-larger AI models on resource-constrained edge devices, like a car's computer. If Tesla can use this to bring FSD V14 to its huge existing fleet of HW3 vehicles, it would be a massive win, avoiding the need for costly hardware retrofits and vastly increasing the data-gathering and revenue-generating potential of its FSD program. This highlights how software and AI architecture innovations can be as critical as new silicon. The rumored Terafab project further signals Tesla's ambition to control its entire AI stack, from chips to models, across both automotive and robotics.
Essiale suggests this compression technology is crucial for democratizing advanced autonomous features. UPA Photo's deep dive into FSD v14.3 notes it already features a rewritten AI compiler for 20% faster reaction times, indicating Tesla is actively optimizing its software stack for existing hardware. Both reports emphasize that Tesla's ambitions extend far beyond cars, with the Terafab project seen as foundational for its AI, Optimus robot, and energy businesses.
NVIDIA has launched Cosmos 3, a new foundational model for physical AI that it calls an 'open omnimodel.' Announced on Monday, the model is designed to unify reasoning, world generation, and action generation for robots, autonomous vehicles, and other smart devices. Built on a Mixture-of-Transformers (MoT) architecture, Cosmos 3 aims to significantly accelerate development cycles by providing a flexible and scalable platform with high physics accuracy and native multimodal understanding.
Why it matters
Cosmos 3 is NVIDIA's play to create the definitive operating system for physical AI. By offering a single, powerful model that can handle simulation, perception, and action, NVIDIA aims to streamline the incredibly complex process of training and deploying intelligent machines. For the robotics community, this provides a powerful foundation to build upon, potentially reducing redundant development efforts and accelerating innovation. The 'open' nature of the model is particularly key, inviting developers into NVIDIA's ecosystem to build the next generation of autonomous systems.
Ship2PL reports that Cosmos 3 is intended to blur the lines between the digital and physical realms. CWSX highlights its ability to reduce training and evaluation cycles from months to days. This launch complements NVIDIA's hardware push with the new Jetson Thor modules, creating a comprehensive, full-stack platform for robotics development.
A project in Los Angeles is exploring a new model for robotic autonomy where a major telecommunications company provides cognitive support to food delivery robots, effectively acting as a distributed, external 'brain.' According to a report on Sunday, the concept involves offloading complex reasoning tasks—like navigating unexpected obstacles or interpreting novel situations—to the telco's edge network, which already supports thousands of autonomous vehicles in the region. This would allow the on-board robot AI to focus on immediate navigation and control.
Why it matters
This hybrid autonomy model could solve a critical scaling problem for robotics. Instead of equipping every robot with expensive, power-hungry compute capable of handling every possible edge case, the system outsources the 'heavy thinking' to a powerful, shared network. This could dramatically lower the cost and complexity of individual robots, making large-scale deployments more economically viable. For telcos, it opens up a significant new revenue stream, positioning them as essential infrastructure providers for the machine economy. This is a novel approach to the architecture of robot intelligence that is highly relevant to anyone building or deploying autonomous systems.
The project envisions a tiered system where robots handle most situations autonomously but can 'call for help' from the network AI when encountering a problem beyond their local capabilities. This leverages the telco's massive existing investment in edge computing and high-bandwidth connectivity.
Researchers at Seoul National University have developed a new artificial muscle for robots that integrates both sensing and actuation into a single component. According to a report on Monday, the system mimics biological muscle-tendon structures, allowing a robot to simultaneously feel what it is touching and know its own position, providing a form of human-like proprioception.
Why it matters
This is a fundamental breakthrough in component technology. Current robots typically rely on a complex and distinct set of motors, actuators, and external sensors to move and perceive their environment. By integrating these functions into a single 'smart muscle,' this research points toward a future of more elegant, lightweight, and biologically-inspired robots. This could dramatically simplify robot design and improve 'physical intelligence,' where the robot's body inherently understands its state without relying solely on a central computer and external sensors.
Another related innovation in actuation comes in the form of 'Electrofluidic Fiber Muscles,' a new design that uses electrofluidic pressure to mimic biological muscle strands, aiming for more efficient and human-like movement. Together, these developments signal a clear trend towards creating robotic components that are more analogous to their biological counterparts.
LG Group is significantly expanding its collaboration with NVIDIA to accelerate its push into physical AI and robotics. On Sunday, NVIDIA highlighted LG Electronics as a key partner for its new Cosmos 3 foundation model. LG is leveraging the platform for its 'LG Cloi' AI home robot and is also expanding its robotics components business, positioning AI and robotics as core future growth pillars for the company.
Why it matters
This strategic alliance signals a major commitment from a global consumer electronics giant to become a key player in the physical AI landscape. By leveraging NVIDIA's advanced platforms, LG aims to compete in the commercial humanoid robot market and strengthen its position as a supplier of critical robotic components. For the robotics ecosystem, this means another major, well-capitalized company is entering the fray, likely accelerating development and competition, particularly in the consumer and smart factory sectors as LG pursues its 'zero-labor home' vision.
Europe Says reports that LG Chairman Koo Kwang-mo is personally driving this initiative, indicating its high strategic importance. The partnership is not just about end-products; it's also about LG supplying AI server components and other hardware, creating a vertically-integrated strategy.
Robbyant, the embodied AI unit of China's Ant Group, has successfully raised 15 billion yuan (approximately $2.07 billion) in its first funding round. This massive capital injection follows the unit's release of the LingBot-VLA 2.0 open-source model we tracked last month, and will be used to aggressively accelerate its robotics and artificial intelligence development.
Why it matters
This is one of the largest single funding rounds for a robotics-focused AI company to date, signaling immense investor confidence in Ant Group's potential to become a major player. Combined with its prior open-source releases, this level of investment will likely intensify competition and significantly accelerate the pace of physical AI innovation in China.
The Standard reports this as the unit's first external funding round. The funding builds on Robbyant's 'embodied-native' approach, as seen with LingBot-VLA 2.0, which was designed from the ground up for physical tasks rather than being adapted from digital models.
Reimagine Robotics, a startup founded by former Google DeepMind leaders, officially came out of stealth on Monday. The company has developed AI-powered robots that can be taught new tasks, corrected, and adapted directly by factory workers without needing specialized programming. The company is already deploying its robots in advanced manufacturing and electronics disassembly, allowing clients to rapidly prototype new automated behaviors on-site.
Why it matters
This 'learn-on-the-job' approach directly tackles a major barrier to automation adoption: the high cost and long lead times associated with expert programming. By empowering non-technical workers to train robots, Reimagine could make flexible automation accessible to a much wider range of industries, especially in settings with high-mix, low-volume production. This represents a significant step towards making robots more like collaborative tools than fixed infrastructure.
VentureBeat notes the company was founded in April 2025 and has received pre-seed funding, with plans for a new fundraising round. According to a company press release, the system allows workers to 'show, tell, and text' instructions to the robot, which then generalizes the task. This is a fundamentally different interaction model than traditional industrial robotics.
Construction management software giant Procore Technologies has agreed to acquire DroneDeploy for approximately $845 million in cash. The deal, announced Monday, aims to integrate DroneDeploy's reality capture platform, which supports drones and ground robots, into Procore's software. The goal is to provide a unified platform for jobsite monitoring, using visual data and AI to enhance decision-making.
Why it matters
This acquisition represents a major consolidation in the AEC (Architecture, Engineering, and Construction) tech space, signifying that robotic data capture is becoming a core feature of modern construction management, not just a niche add-on. For the robotics industry, it validates the market for automated reality capture and demonstrates a clear exit path for startups focused on industrial verticals. Integrating robotic data directly into project management workflows is a critical step for automating construction processes at scale.
The AI Insider reports this move will combine Procore's vast project data with DroneDeploy's visual intelligence tools. This follows a broader trend of major software platforms acquiring specialized robotics and AI companies to build end-to-end vertical solutions.
Indonesia's Health Ministry is launching a major initiative to integrate robotic surgery into its national healthcare system. A new Ministry Robotics Committee will guide the procurement of 40 surgical robot units for government hospitals. The government is also working to include robotic surgery in its national healthcare financing system and is mandating technology transfer to foster a domestic medical device industry.
Why it matters
This is a significant example of a government-led, national-scale adoption of advanced medical robotics. Unlike market-driven adoption in other countries, Indonesia's top-down approach aims to ensure widespread access and build local industrial capacity simultaneously. For the surgical robotics market, this represents a large new opportunity, but also one that requires partnership and technology transfer. The initiative could become a model for other developing nations looking to modernize their healthcare infrastructure.
Observerid.com reports that a key focus is training a new generation of surgeons to use the technology. This contrasts with a recent critique from an Indian surgeon, reported by News Now Nation, who called current systems 'slave robots' without intelligence and argued for more AI integration to improve accessibility and lower costs.
AMD, in partnership with Analog Devices (ADI), has launched the Kria AI Robotics Developer Platform. The new offering, announced Monday, is an integrated system-on-module (SOM) that combines AMD's Kria K26 SOM processing capabilities with ADI's physical sensor interfaces. The platform is designed to accelerate the deployment of industrial and mobile robots by providing a pre-validated, centralized architecture for perception, navigation, and motion control, with support for open-source environments like ROS 2.
Why it matters
This is AMD's most direct challenge yet to NVIDIA's dominance in the robotics hardware market. By creating an integrated, pre-validated platform with a major sensor company, AMD is addressing a key pain point for developers: hardware and software integration. This could significantly reduce time-to-market for new robotic systems. For the robotics ecosystem, it provides a powerful and much-needed competitive alternative to NVIDIA's Jetson platform, fostering more innovation and potentially driving down costs in the AI hardware space.
This launch builds on AMD's recent introduction of its Ryzen AI Embedded X100 series, signaling a comprehensive, full-stack strategy for the robotics and industrial automation markets. The collaboration with Analog Devices is critical, as it tightly couples the compute with the all-important sensor data pipeline.
A study by researchers at the Open University of Catalonia (UOC), analyzing nearly 5,000 Spanish manufacturing firms from 1990 to 2016, found that adopting industrial robots significantly increases productivity in small and medium-sized enterprises (SMEs). However, the research concluded that robotization alone does not directly lead to an increase in exports. The ability to export was more strongly linked to the presence of 'localization economies'—regional clusters of suppliers, skilled labor, and infrastructure.
Why it matters
This research provides a nuanced, data-backed view on the impact of automation. It suggests that for SMEs, robotics is a powerful tool for internal efficiency but not a silver bullet for global market access. For an entrepreneur considering an investment in automation, this is a crucial insight: the ROI of robotics may be primarily in cost reduction and output gains, while cracking international markets requires a broader strategy that includes ecosystem and supply chain considerations, not just factory-floor technology.
The UOC press release states that the productivity gains were most pronounced in smaller and less innovative firms, suggesting robotics can be a powerful leveling-up tool. This aligns with other analyses showing that cobots are increasingly viable for small-scale manufacturing due to their flexibility and lower cost.
Researchers have developed a new magnetic coil system that can move microrobots across an entire workspace using a uniform magnetic field, eliminating the need for constant visual tracking. According to a report on Monday, the triaxial Helmholtz coil arrangement exerts a predictable pull on magnetic microrobots without needing to know their exact location, shifting control from a feedback-based system to one based on field properties.
Why it matters
This is a fundamental shift in how microrobots are controlled. The reliance on high-magnification imaging and complex feedback loops has been a major barrier to their practical use, especially in opaque environments like the human body or inside machinery. By creating a 'set and forget' field that pulls robots in a desired direction, this system could enable simpler, more robust, and more autonomous microrobotic applications for tasks like targeted drug delivery or internal inspection.
This innovation represents a philosophical move towards 'autonomy-forward' design at the micro-scale. It complements other recent breakthroughs in microrobotics, such as DNA droplets that can be controlled by light and liquid metal robots that can split and merge, all aimed at creating more versatile and controllable microscopic systems.
Following up on the NHTSA exemption Zoox secured for its wheel-free robotaxis over the weekend, the broader AV industry is facing a brewing federal-versus-local regulatory clash. While Zoox now has federal clearance to launch its paid service in Las Vegas, operators like Waymo are encountering increasing resistance and scrutiny from local officials in cities like San Francisco over incidents involving emergency responders.
Why it matters
This is a classic federal vs. local regulatory clash that will define the next phase of AV deployment. The federal green light for Zoox's custom, no-controls vehicle is a landmark moment, officially sanctioning the move away from retrofitted cars. However, the simultaneous pushback at the city level shows that operational reality is messy. The industry's ability to scale will depend entirely on its ability to solve these local-level safety and public trust challenges, regardless of federal approvals.
TechCrunch highlights the contrast between federal acceleration and local caution. KHON2 reports that federal regulators are also considering a proposal to permit robotaxis without brake pedals. Meanwhile, an analysis from Runtime Wire suggests Waymo is also scaling up its fleet rapidly, having imported at least 3,200 Zeekr-built vans despite high tariffs.
The 'Brain' vs. 'Body' Platform War Heats Up Google's Gemini Robotics 2 aims to be a universal AI layer controlling any robot's full body, suggesting a future where hardware is commoditized. Simultaneously, NVIDIA and AMD are rolling out powerful, robotics-specific hardware platforms (Jetson Thor, Kria), betting that the foundational layer will be custom silicon, not just software.
Venture Capital Flows to Robotics Startups Across the Stack A surge of funding is hitting the robotics sector. Ant Group's Robbyant raised 15 billion yuan, dexterous hand maker LinkerBot is valued at $6B, and firms like Reimagine Robotics and Astra Robotics secured significant rounds. This highlights investor confidence in both full-stack humanoids and specialized component makers.
Automakers Accelerate Humanoid Deployments and Development Automakers are moving aggressively into humanoid robotics. Hyundai is committing to deploy 25,000 Atlas robots, while BYD is set to debut its own model this month for showroom applications, leveraging its automotive manufacturing expertise to enter the race against Tesla and others.
The Consumer Robotics Market Expands Beyond Vacuums Companies like Ecovacs and Yarbo are launching a new generation of sophisticated, modular outdoor robots for tasks like lawn mowing, snow blowing, and pool cleaning. These devices, often featuring advanced LiDAR navigation, signal a broadening of the consumer robotics market into complex home and yard maintenance.
Open-Source Robotics Ecosystem Continues to Mature New tools and platforms are making robotics development more accessible. A new real-time support package for ROS 2 enhances performance for industrial applications, while projects like a low-cost AUV and 3D-printable humanoids highlight the growing power of open-source hardware and software to drive innovation.
What to Expect
August 2026—BYD is expected to debut its first humanoid robot at a showroom in Zhengzhou, China.
Mid-2027—Atlas Energy Solutions plans to expand its driverless truck fleet to 100 vehicles in the Permian Basin.
2027—Dubai's Autonomous Logistics Future Lab, SHIFFT, aims for its first commercial deployment.
2028—Hyundai Motor Group plans to have deployed over 25,000 Atlas humanoid robots in its manufacturing plants.
How We Built This Briefing
Every story, researched.
Every story verified across multiple sources before publication.
🔍
Scanned
Across multiple search engines and news databases
459
📖
Read in full
Every article opened, read, and evaluated
239
⭐
Published today
Ranked by importance and verified across sources
19
— 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