Where AI meets robotics: humanoid robots and labs racing to ship them, embodied foundation models, warehouse and industrial deployments, and the investor and policy backdrop.
Topic brief
Brief updated Jul 11, 2026
Robotics is the engineering of machines that sense, decide, and act in the physical world, and its current wave is defined by the fusion of robotics with modern AI, often called physical AI or embodied AI. Where classical robots were programmed for narrow, repetitive tasks, today's systems increasingly learn general skills from data, using foundation models, world models, and vision-language-action architectures to perceive their surroundings and adapt. Humanoid robots, autonomous vehicles, drones, and industrial arms are the visible edge of this shift.
For AI, ML, data-science, and engineering practitioners, robotics is where models leave the screen and take on real-world consequences. It couples perception, control, and learning with hard constraints such as safety, latency, energy, and reliability that pure software systems rarely face. The tooling is maturing into open stacks for data collection, simulation, teleoperation, and evaluation, which lowers the barrier to building embodied systems and makes robotics a fast-moving applied-ML frontier.
The stakes are also economic and geopolitical. Automakers, logistics firms, and defense agencies are among the earliest adopters, national governments are pledging enormous sums to build domestic robot industries, and safety standards for autonomous machines are being written in real time. Because physical AI touches manufacturing, transport, defense, and labor, decisions about which robots to deploy, how to train them, and how to govern them carry weight well beyond the lab.
The newest signals show robotics scaling outward from humanoids into adjacent physical-AI niches. DEWALT commercially launched DALE, a fleet-capable downward-drilling robot for data center construction developed with August Robotics, reporting 10x faster drilling, 99.97 percent accuracy across more than 230,000 holes, and 190 weeks saved across 26 construction phases. MIT CSAIL researchers separately described FloatForm, modular robotic boats about 21 centimeters square that use onboard sensing, thrusters, and magnetic latches to self-assemble into larger floating structures. Maritime and underwater autonomy moved forward too: Hanwha Systems put a 30-ton unmanned surface vessel through sea trials near Busan, and ZenaTech began U.S. field testing of its IQ Aqua underwater mine-detection drone in Florida. On the transport side, Hanjin launched South Korea's first paid autonomous freight service, a 25-ton truck running a 118-kilometer route from Gunsan Port to the Daejeon Mega Hub three times a week with a safety officer aboard, a production-data milestone rather than a short demo.
Institutional commitment is deepening in parallel. Israel's IDF is finalizing a new General Staff structure dedicated to AI, robotics, and unmanned systems, formalizing military ownership of the category, while Seattle's Overland AI became the first ground-autonomy company to serve as prime contractor on a U.S. military production deal, a $19.7 million Marine Corps contract to help resupply the Marine Air Defense Integrated System. On the corporate side, Xpeng lost its robotics business head as it pushes the Iron humanoid toward an end-2026 mass-production target, the second leadership departure from that program in recent months, a reminder that humanoid programs are now judged on manufacturing execution as much as research novelty.
Several threads remain unresolved and worth tracking. Xpeng is pushing its Iron humanoid toward an end-2026 mass-production target even after its robotics head departed, a test of whether the program can hold its timeline through leadership churn. Overland AI's Marine Corps contract has initial deliveries due in early 2027 and full delivery by October 2027, an early proving ground for autonomous ground vehicles moving from demonstration to production. Israel's IDF has not yet finalized its new AI, robotics, and unmanned-systems general staff structure, and New Jersey's lidar-for-robotaxis bill is still advancing through the legislature. On the capital side, Japan's target of 10 million robots and the 312 trillion-won Korean conglomerate investment, which includes Samsung's committed spend on humanoid-robot mass production, will play out over years rather than months, and NVIDIA's Halos safety stack is only beginning to roll out, with Agility as an early adopter.
| maker | robot | status |
|---|---|---|
| Figure | Figure 03 | Deploying in a logistics sequencing use case at BMW's Plant Spartanburg. |
| Apptronik | Apollo 2 | Being trained at a newly opened robot park. |
| Genesis AI | Eno | Unveiled as a general-purpose robot. |
| UBTech | UWORLD U1 | Debuted in Shenzhen with launch-day orders reported above 13,361 units. |
| UMA | Northstar | Unveiled by ex-Tesla scientist Remi Cadene with a Real-Time Learning architecture. |
It is the combination of robotics with modern AI, where machines learn general skills from data rather than being programmed for narrow tasks. Recent examples include foundation and world models for robots, Mistral's single-camera Robostral Navigate model, and NVIDIA's Isaac GR00T stack, all aimed at giving robots more general perception and control.
Yes, increasingly, though execution risk is real. BMW is deploying Figure 03 humanoids in a logistics sequencing use case at Plant Spartanburg, Apptronik opened a robot park to train its Apollo 2, and Genesis AI unveiled the Eno general-purpose robot. At the same time, Xpeng's robotics head departed as the company pushes its Iron humanoid toward an end-2026 mass-production target, a reminder that manufacturing execution, not just demos, determines whether these programs ship on schedule.
A great deal, from both companies and governments. Four Korean conglomerates, Hanwha, Hyundai Motor, Samsung, and SK Group, pledged a combined 312 trillion won (about $201.7 billion) for a southeastern industrial region that includes humanoid-robot mass production, Hyundai separately committed $27.3 billion to mobility and physical AI, Japan set a target of 10 million robots, and defense-oriented autonomy drew funding through Quantum Systems' $1.2 billion raise and Overland AI's $19.7 million Marine Corps contract.
Open stacks are maturing quickly. Hugging Face released LeRobot 0.6 with a fuller evaluate-correct-train loop, NVIDIA brought Isaac GR00T 1.7 and Teleop into LeRobot, and Mistral released Robostral Navigate, an 8B single-camera navigation model that substitutes for LiDAR.
It is becoming a first-class layer. NVIDIA announced Halos, a full-stack robotics safety system with Agility as an early adopter, while regulators act too: New Jersey advanced a bill requiring robotaxis to use lidar plus multiple sensing modalities and complete 50,000 miles of testing, and Israel's IDF is finalizing a new general-staff structure specifically for AI, robotics, and unmanned systems.
In logistics, maritime, defense, and construction. Hanjin launched South Korea's first paid autonomous freight service, DEWALT commercialized a downward-drilling robot for data center construction, Hanwha Systems and ZenaTech are running sea and underwater trials of unmanned vessels and mine-detection drones, and Overland AI became the first ground-autonomy company to serve as a U.S. military prime contractor.
No recent Robotics articles
We'll have fresh Robotics coverage here soon. In the meantime, explore the full news feed.
Go to the news feed