Robots are advancing faster than ever. The “Coffee Test” measures a machine’s ability to brew a cup without error across three distinct, unstructured spaces that match the complexity of an ordinary home. According to “Wave 7: Robotics and Physical AI” from the Forecasting Research Institute, robots will be able to pass this test reliably from 2034 onwards.

Between that milestone and 2040, specialists surveyed for the report predict that robotic surgery will also arrive, enabling a machine to perform a successful appendectomy without human intervention. By then, we will need to have resolved major hurdles, including regulatory barriers, medical malpractice liability, and pushback from surgeons themselves.

Overcoming these misgivings about the impact of robotics on jobs will require genuine social acceptance, sparking a compelling debate. Professor Ainhoa Urtasun of the Public University of Navarre co-authored what the International Federation of Robotics named the world’s best research paper published in 2025. The study shows that US manufacturing plants adopting robotics increase their job listings by roughly 150% and expand their workforces by 15% compared to non-automated competitors.

At the latest CES in Las Vegas, consulting firm McKinsey presented a flood of data pointing to a clear reality: there is no time to lose. By 2033, US industry will face an unfilled gap of 1.9 million jobs; currently, only 6% of factories deploy robots at scale; factory robot installations in China outpace the US by ten to one; 13% of working hours are already suitable for robotic automation; and venture investments in robotics have surged 25-fold over three years.

The generative AI revolution is providing robots with genuine autonomy—the core paradigm shift behind so-called physical AI. Early deployments across logistics, manufacturing, mobility, healthcare, and defence project a market potential of approximately 430 billion euros by 2030, with Europe expected to generate between 80 billion and 110 billion euros of that total.

As AI expands its reach into robotics, it is taking on a fascinating array of new labels. At the Automate 2026 trade show in Chicago, Intrinsic—an Alphabet company and subsidiary of Google—unveiled a prototype powered by “adaptive AI”. Running on its bespoke operating system, IntrinsicOS, the system executes pre-packaged robotic actions that can be added to workflows with a few clicks using an intuitive drag-and-drop interface.

A joint initiative between FORT Robotics and NVIDIA is looking beyond traditional machine perception, which relies entirely on onboard sensors. Instead, they promote “outside-in safety”, using off-board sensors and cameras positioned around the workspace to guide the machine. Meanwhile, “acoustic AI” is emerging through platforms that analyse manufacturing sounds in under a second to spot production flaws in real time.

The next generation of robots will rely less on any single dominant technology and more on a fusion of AI, bio-inspired engineering, and autonomous decision-making in complex environments. Barbara Mazzolai of the Italian Institute of Technology argues that energy efficiency will be critical as robots enter daily life at scale. Her work looks to plants and octopuses, which perform complex tasks under severe energy constraints without a centralised control system.

Vision-language-action (VLA) models are also advancing rapidly as a promising route to give robots generalisable spatial reasoning. The π₀ and π0.6 models from Physical Intelligence already execute tasks like folding laundry across various robotic hardware setups without needing job-specific retraining.

Achieving human-level dexterity remains the ultimate frontier. Robotic hands require exceptional freedom of movement, often using more than 20 distinct joints. Advancements in 3D spatial intelligence—yet another descriptor for AI—will be essential. Balancing safety with operational flexibility remains a core engineering challenge. Interestingly, one of the most immediate benefits is occurring in workplace safety: at Amazon, facilities integrated with robotics have seen injury rates fall by 15%.

Commercial vehicle manufacturers are building dedicated operating systems for smart fleets. Autonomous trucks with safety drivers are already operating on public roads in Texas, US, where fully driverless cab operations are scheduled for approval starting in 2027.

Caterpillar, a leading manufacturer of heavy construction machinery, describes the shift as moving from intelligence assisting the operator to intelligence acting as the operator. By early 2026, its autonomous mining fleet had logged over 385 million kilometres without a human behind the wheel.

NVIDIA envisions a robotic future driven by physical AI through a “three-computer” architecture: one system trains the physical AI model, a second simulates the physical environment and operational workflows, and a third runs onboard execution. Making this system work requires more than solid hardware and software—it demands contextual data understanding, prompting major tech firms to forge strategic alliances with industrial, construction, and network management operators.

As hardware costs fall and physical AI improves, the business case for robotics is growing stronger. Converging technological breakthroughs have brought basic humanoids, like Unitree’s G1, down to prices below 15,000 dollars, while UBTECH’s K model has entered mass production after driving average unit costs down from three million to 100,000 dollars over a decade.

Although humanoids still suffer from key limitations that prevent deployment in unsupervised environments, they have rapidly captured global attention as the most visible application of physical AI. Several automotive manufacturers are already running trials with humanoid robots on assembly lines.

The year 2025 marked the start of mass production for humanoid robots in China, where over 140 domestic manufacturers have launched more than 330 models. OpenAI restarted its dedicated robotics division in early 2025, Tesla and XPeng are racing to roll out proprietary units, Goldman Sachs projects the total market will expand from 6 billion dollars today to 38 billion by 2035, and Morgan Stanley analysts estimate it will hit five trillion dollars by 2050.

Looking beyond humanoids, engineers are exploring frontier concepts such as bio-integrated machinery and quantum robotics. While practical quantum robots remain decades away, they promise unprecedented computing speed and operational capability. To fully capitalise on physical AI, businesses will need to rethink operator skill sets and workforce training requirements.

Source: 

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5114250
• https://www.automateshow.com/
• https://2026.ieee-icra.org/attend/
• https://www.youtube.com/watch?v=vI6k_mJeYU8