Explore every episode of the podcast Physical AI Daily

Uber and Zipline want to scale autonomous drone delivery toward one million orders per day, while Diligent Robotics rolls out Moxi 2.0 with a World Model trained by experience from real hospital deployments. Serve expands its delivery network with Grubhub, San Jose, Washington, DC, and a new micro-depot model, while Seeing Machines brings human-awareness AI from millions of cars into robotics. Today: why the deployed robot fleet itself may become Physical AI's most valuable dataset.

What does “autonomous” actually mean when a robot reaches production? Today we propose a practical scorecard based on interventions, recovery, utilization, human attention, useful output and cost per task. Plus, the World Humanoid Robot Games move toward more realistic autonomous testing, RoboSynChallenge measures synthetic-to-real transfer, and Actuate 26 brings Physical AI builders from Physical Intelligence, Google DeepMind, Aurora, Wayve, NVIDIA, 1X, Zipline and others to San Francisco this week.

HII expands its physical-AI shipbuilding push with Path Robotics and GrayMatter Robotics, Palladyne AI demonstrates one operator controlling drones from four manufacturers, and new ARGUS research tackles one of robot learning’s hidden data problems: changing camera viewpoints. Today’s episode looks at why the next physical-AI platform may sit above the individual robot..

Avatar Robotics raises $6.5 million to deploy remotely operated warehouse robots that create customer value and training data before achieving full autonomy. The U.S. Air Force and Lockheed Martin also demonstrate an AI agent using live infrared data to direct an autonomous airborne intercept. Today’s episode examines how human control evolves into shared autonomy—and why useful work per human hour may be the metric that matters.

Waymo opens its fully autonomous Dallas service to everyone, moving beyond an invitation-only rollout into a real test of public demand and fleet operations. Blue Water Autonomy also gains eligibility to compete for U.S. Navy open-ocean survey missions. Today’s episode examines the five stages separating an autonomy demonstration from a scalable operating business.

Procore agrees to acquire DroneDeploy for approximately $845 million, combining drones, ground robots and jobsite imagery with construction drawings, schedules and operational workflows. Aptiv also says its robotics activity is advancing toward commercial programs. Today’s episode examines who will own the physical-AI loop from observation and understanding to action and verification.

Former DeepMind robotics leaders launch Reimagine Robotics with industrial robots workers can teach directly, Agile Robots projects a doubling of annual revenue, and XTEND demonstrates distributed production across seven robot platforms and five countries. Today’s episode examines how worker knowledge, integration expertise and common software platforms may become the real foundations of scalable physical AI.

No weak weekend filler today. This episode connects the week’s biggest developments—from Gemini Robotics and humanoid teleoperation to robot components, engineering agents, regulation, implementation and real customer metrics—to explain the complete stack required to turn physical AI into a scalable operating business.

Tau Robotics begins sending humanoids into San Francisco homes for paid cleaning sessions, Unitree’s IPO filing reveals its financial exposure to new U.S. restrictions, and major airlines tighten rules for transporting robots with large lithium batteries. Today’s episode examines the economics, geopolitics and logistics that begin when physical AI leaves the laboratory.

Physical AI Daily celebrates its first month with a recap of the strongest trend that emerged across July: robotics is becoming an infrastructure industry. From foundation models and edge compute to factory deployments, robotaxis, industrial investment, safety, and national strategy, this special episode examines why the competition is shifting from impressive demonstrations toward complete systems that can be deployed, improved, supported, and scaled.

Google DeepMind expands Gemini Robotics to coordinated whole-body humanoid control, Exyn reports major productivity and safety gains from an autonomous underground-mine deployment, and Tesla purchases a full-body locomotion and teleoperation platform for Optimus development. Today’s episode examines how general robot intelligence, human demonstration data and focused commercial deployments are beginning to connect.

Zoox wins the first U.S. approval for paid robotaxi service using a vehicle built without steering wheels or pedals, Stabilus and Synapticon prepare integrated humanoid actuators for industrial production, and Teradyne reports $100 million in quarterly robotics revenue. Today’s episode examines the regulatory, manufacturing and commercial foundations required to scale physical AI.

Neros raises $250 million to scale autonomous and interceptor drones, with a long-term ambition of manufacturing one million drones per year. X Square Robot demonstrates embodied AI sorting 1,816 random parcels in an hour, while Galaxea's G0.5 challenges conventional robot foundation-model architecture by putting reasoning and action into one autoregressive stream. Today: what happens when Physical AI has to multiply across hardware, tasks and robot bodies.

The United States restricts authorization of new Chinese humanoid and quadruped robots, LG Innotek and TDK partner to develop integrated vision and tactile sensing, and Nauticus completes an early electric manipulator prototype for autonomous subsea work. Today’s episode examines supply-chain trust, sensory trust and physical execution in robotics.

Northrop Grumman launches a robotic spacecraft designed to service and extend the life of satellites, Wandercraft secures Ecuadorian clearance for its self-balancing rehabilitation exoskeleton, and XTEND integrates thousands of reported Atlas systems into its European robotics operation. Today’s episode examines how robots are evolving from standalone products into long-term operational infrastructure.

NVIDIA is giving AI engineering agents access to physics models, simulation and specialized solvers, while Wetour Robotics works to assemble an implementation structure for a proposed warehouse rollout. Today’s episode examines both developments and asks a critical industry question: what should actually count as a robotics deployment?

Johnson & Johnson receives FDA authorization for its Ottava surgical robot and prepares its first U.S. hospital deployments. Foundation partners with AMD while claiming its humanoids have already contributed to automotive production, and AgiBot begins the process of entering the public markets. Today’s episode explores physical AI’s new accountability phase—where hospitals, factories, and investors demand measurable performance rather than impressive demonstrations.

Hyundai lays out a physical-AI roadmap connecting robots, factories, vehicles, operational data, and a planned 30,000-unit U.S. production facility. Exyn prepares to bring autonomous inspection into U.S. Air Force aircraft maintenance, while Applied Intuition introduces agentic engineering infrastructure for building and deploying intelligent machines. Today’s episode explores why the physical-AI race is shifting from individual robots to complete operating systems.

Atoms raises $1.7 billion to bring robotics and AI into mining, construction, transport, and food production, while Ropedia raises $22 million to capture real-world human experience for training machines. Today’s episode connects these developments to the recent TerraFirma, Gritt, and Wetour stories and explores the two economies forming around physical AI: industrial deployment and the data needed to make robots better.

Japan reveals the enormous scale behind its national physical-AI strategy, with more than $2.3 billion in initial government backing and a goal of 10 million AI-enabled robots. Wetour takes a different approach by collecting human physical expertise directly inside real warehouse operations, while DEEP Robotics says its humanoid is entering hazardous industrial deployments. Today’s episode explores one of physical AI’s defining battles: who can build the fastest loop from real-world experience to better robots?

Samsung creates a dedicated robotics division, Serve Robotics prepares to test whether its autonomous delivery operation can scale internationally, and NEURA expands its network of physical-AI training facilities. Today’s episode connects these developments to the recent TerraFirma and Gritt stories and asks what comes next after proving robots can work: building the infrastructure that makes every new deployment easier than the last.

Gritt launches with $32.4 million and robots already working on real construction sites, the U.S. Marines select an autonomous counter-drone system for integration, and Knightscope reports rapid growth as it combines robots, software, and people into a managed security service. Today’s episode explores a major shift in physical AI: customers may increasingly buy outcomes, while the robots disappear into the operation behind them.

TerraFirma raises $115 million to scale robotics in real construction operations, Sunday Robotics prepares to put its Memo robot into actual homes, and XPeng targets automotive-scale humanoid production. Today’s episode explores the emerging physical-AI bottleneck: turning impressive robot capabilities into scalable, reliable operations.

Pony.ai and Uber plan more than 2,000 robotaxis across five European cities, LG and NVIDIA combine humanoids, factory robots and a Physical AI data factory, and new U.S. drone tariffs could reshape autonomous-system supply chains. Today: what changes when robotics moves from individual machines toward infrastructure built for thousands.

AGIBOT launches four new robot platforms while ACE Robotics pushes world models closer to real-world action. Today’s Deep Dive examines the emerging physical-AI stack—from robot bodies and predictive world models to the deployment infrastructure that could turn every machine in the field into part of a continuous learning system.

Pudu Robotics wants one shared AI stack to power an entire family of robot bodies, a medical exoskeleton shows how humans and machines can learn to move together, and SS Innovations reports more than 12,000 robotic surgeries and 178 telesurgeries. Today’s episode explores a fundamental physical-AI advantage: the machines deployed today may generate the data that makes tomorrow’s robots smarter.

NVIDIA brings Japan’s industrial giants into a major physical-AI push, Hyundai moves to take full ownership of Boston Dynamics, and Agility Robotics opens a new AI development hub in Silicon Valley. Plus, a self-dressing soft robot offers an important reminder: sometimes better mechanics can solve problems that others are trying to solve with more AI.

A new industrial dexterous hand takes a full-stack approach to robot learning, autonomous food robots enter military operations, eVTOL developers push toward uncrewed flight, and embodied AI prepares for space. Today’s Physical AI Daily asks a critical question: Can customers operate advanced robots without the engineers who built them?

Walden Robotics emerges from stealth with $300 million and robots already working at Toyota, Archer builds a foundation model specifically for aviation, Surgerii expands single-port robotic surgery into Europe, and Machina pushes robotic manufacturing into defense production. Today’s episode explores why physical AI is becoming increasingly specialized around real operating environments.

Autonomous vehicles head toward a new market in Türkiye, Quadric raises fresh funding for programmable edge-AI compute, and RoboCup puts 22 humanoid robots on the same soccer field. Today’s episode explores why physical AI is increasingly a systems challenge—not just a model or hardware challenge.

Today's Physical AI Daily examines Helsing's massive funding round through the lens of physical AI and explores why specialized robots continue to outperform general-purpose humanoids in commercial deployments. With a quieter news cycle, today's episode includes a Deep Dive into the economics and strategy behind successful robotics businesses.

Today’s Physical AI Daily covers Hippo Harvest’s $30 million robotic-farming round and Dryad Networks’ autonomous wildfire-response milestone. Because only two strong new stories qualified, the episode also includes a clearly labeled Deep Dive into why agricultural robotics is shifting from standalone machines to integrated production, workflow, and fleet systems.

Today’s Physical AI Daily covers 1X 25dof hand demo, AI² Robotics’ $735 million humanoid round, Asylon’s autonomous aircraft-inspection contract, Zeta Surgical’s FDA clearance, NHTSA’s robotaxi warning, and an early test of humanoid robots in surgery. The theme: physical AI is moving from demonstrating capability to earning institutional trust.

Today’s Physical AI Daily covers Emesent’s $17 million autonomy raise, RoboSense’s 510% robotics-LiDAR growth, ABB’s visual-SLAM forklift, Echodyne’s new radar factory, and Atlas operating at the World Cup. The theme: physical AI is moving beyond models and demos into sensing volume, manufacturing capacity, integration, and deployment engineering.

NEURA Robotics acquires Bosch Rexroth's field-proven ACTIVE Shuttle platform, Ondas reports $83.8 million in quarterly autonomous-systems revenue and a $757 million pro forma backlog, and Nauticus starts commercializing autonomy software for existing underwater robots. Plus, new market data shows North American robot demand shifting beyond automotive. Today: why Physical AI may scale by upgrading and connecting the robots already in the field.

Today’s Physical AI Daily looks at Mistral’s first robotics model, CarbonSix’s $40M manufacturing automation round, Robust.AI’s ShipLab deployment, Macrodata Labs’ robot data tooling, and Luxonis’ perception-layer funding. The theme: physical AI is becoming a full stack, not a single breakthrough robot.

Today’s Physical AI Daily looks at Forterra’s autonomous ground vehicles in Ukraine, Hive’s industrial-machine autonomy raise, Nomagic’s live VLA warehouse deployment, New Jersey’s robotaxi sensor bill, and VicOne’s cyber-safety stress test for robots. The theme: the real world is becoming the test bench for physical AI.

Today’s Physical AI Daily looks at Mowito’s demonstration-based factory robotics raise, Lockheed’s Ultra Maritime acquisition, Dominion’s defense autonomy funding, investor skepticism around humanoids, and maritime autonomy as a defense M&A hotspot. The theme: physical AI is moving toward specific jobs, urgent customers, and task-fit robot bodies — not one-size-fits-all robotics.

Today’s episode looks at physical AI as strategic infrastructure: maritime drones, dexterous robotic hands, manufacturing-tech funding, edge AI chips, and defense procurement for uncrewed systems. The theme is that robotics scale depends on the stack underneath the robot: components, compute, manufacturing, data, and customers with urgent real-world problems.

Today’s episode looks at the less flashy but critical deployment stack behind physical AI: dexterity, tactile sensing, robot programming, autonomous science testbeds, and middleware safety boundaries. The theme is that robots scale only when the systems underneath them become reliable.

Today’s episode looks at the deployment stack behind physical AI: humanoid safety, sensing and perception infrastructure, retail cleaning robots, public-space service robots, and AI-native industrial automation. The theme is simple: robots will scale when they are safe, reliable, easy to deploy, and economically useful.

Today’s episode looks at physical AI moving from demos into accountability: Unitree’s IPO path, Tesla’s Miami robotaxi launch, Katalyst’s orbital rescue robot, Weave’s home robot, and Davos’ public Robot City testbed. The theme is simple: robots are now being judged in markets, cities, homes, orbit, and real public spaces.

Today’s episode looks at where physical AI is actually finding market gravity: defense drones, mass-produced humanoids, delivery robot redeployment, shipped-unit reporting, and education pilots. The theme is deployment reality — not just who has the best demo, but where robots are getting funded, ordered, shipped, tested, and scaled.

Physical AI is becoming infrastructure across healthcare, construction, energy, autonomous vehicles, robotaxi operations, and multi-robot coordination.
Top stories:
1. Queue emerged from stealth with $18.6M total funding for autonomous robotic pharmacy fulfillment.
2. Built Robotics autonomous heavy equipment is pile-driving solar infrastructure tied to Meta’s Hyperion AI data center power buildout.
3. Momenta launched a Hong Kong IPO targeting up to $751.1M for autonomous-driving R&D and robotaxi expansion.
4. Aseon Labs raised $10M to build automated robotaxi pit stops for cleaning, charging, inspection, and lost-item retrieval.
5. Six Robotics raised €12M to scale swarm autonomy software for unmanned systems.

Autonomous trucks, construction robots, commercial cleaning and robot learning all face the same challenge: getting beyond the pilot. DAF and Einride move toward production-platform autonomy, RIC Robotics puts two autonomous construction systems to work across a 65+ home deployment, Pudu targets smaller commercial sites, and Surgical WAM shows how ordinary video could reduce expensive robot-training data. Today: what it actually takes for physical AI to scale.

Dyna Robotics says DYNA-2 learned physical intelligence from more than one million hours of human video, potentially attacking one of robotics’ biggest bottlenecks: expensive robot training data. Robo Inc. launches to build the U.S. integration layer around commercial robots, while QYSEA outlines a future of autonomous, connected underwater systems. Today’s episode asks whether physical AI is finally developing a scalable data-to-deployment flywheel.

Faraday Future launches a beyond-line-of-sight teleoperation and multi-robot control platform, while new AutoIntervene research teaches robot policies when to hand control to a human. A second study shows why even a safe autonomous-vehicle stop can leave a difficult recovery problem. Today’s episode examines the human exception layer—and the risk that teleoperation becomes the business instead of the bridge to autonomy.

Brain Corp reports more than 5.3 million autonomous operating hours in six months across a global fleet that has passed 50,000 deployed robots, while Oshkosh invests in Nextera Robotics to connect autonomous jobsite intelligence with heavy-equipment data. Today’s episode asks which metrics actually prove a robotics business is scaling: deployed units, productive machine hours, human attention, or economic output.

Unitree prices its Shanghai IPO at a roughly $9 billion valuation as DeepSeek takes a strategic stake, while X Square Robot reportedly files for a Hong Kong listing. New omega-zero research also tackles simultaneous walking and manipulation in humanoids. Today’s episode asks what changes when physical AI moves from demonstrations and private valuations into public-market accountability.

Unitree's first day as a public company ends at a roughly $50 billion valuation, creating a major new benchmark for humanoid robotics. But as global humanoid shipments surge, estimates that 50%-70% of this year's robots may primarily collect training data raise a more important commercialization question: how many productive robot-hours are actually being created? Plus, τ0-VLA explores giving robots more reasoning time before difficult physical decisions, and Pudu expands into autonomous forklifts.