articleFeb 7, 2026
Mobility & Physical AI
Enterprise & AI

Why Physical AI Will Transform the Next Decade of Industry

From global incumbents to venture-backed startups, physical AI is moving from experimentation to deployment, reshaping how physical systems are built, operated, and scaled.

At the 2025 VivaTech conference in Paris, Nvidia CEO Jensen Huang made a statement that resonated far beyond the AI community: “This is going to be the decade of autonomous vehicles, robotics, and autonomous machines.”

To some, it sounded like another bold claim feeding the AI hype cycle. To others, it reflected a deeper shift already underway. Viewed through the lens of industrial reality, Huang’s words appear less speculative and more diagnostic of where the global economy is heading.

The structural pressures reshaping industry

Modern manufacturers are operating at a critical juncture. The convergence of labor shortages, intensifying cost pressures, and fragile global supply chains - compounded by geopolitical and market volatility - is actively undermining productivity, profitability, and long-term resilience. Simultaneously, a shift in consumer behavior toward rapid delivery, bespoke customization, and environmental sustainability is forcing a fundamental step-change in operational flexibility.

Incumbents are already moving

The actions of global incumbents suggest this shift is well underway:

  • Walmart has partnered with Symbotic to automate its regional fulfillment centers using AI-driven robotic systems. 
  • Mercedes-Benz has partnered with Apptronik to test and co-develop the humanoid robot Apollo for repetitive intralogistics tasks in production environments. 
  • Amazon now operates more than 750,000 robots across its logistics network, coordinating picking, sorting, and transportation at scale. 
  • Tesla has long relied on industrial robotics and is now deploying Optimus, its humanoid robot, directly within factory environments. 
  • In agriculture, Deere & Co. is commercializing autonomous tractors capable of operating without human drivers. 

These are not pilot projects. They are large-scale deployments that signal a structural transformation in how physical work is performed.

From robotics to physical AI

This transformation is driven by the rise of physical AI. While traditional robotics excelled in structured, predictable environments, physical AI enables machines to perceive, decide, and act autonomously in dynamic, real-world conditions. Earlier AI paradigms evolved largely in the digital domain, with perception AI interpreting the world through sensors, generative and agentic AI producing content and executing workflows. Their interaction with the physical world, however, remained indirect. Physical AI closes this gap by embedding intelligence directly into machines, operating through a continuous feedback loop: sensing, processing, deciding, acting, and learning in real time without explicit reprogramming.

This shift is particularly evident in the emergence of humanoid robots. Declining hardware costs, advances in AI and machine learning, edge computing, and improved sensing technologies are making embodied intelligence economically viable. Humanoid systems combine perception, cognition, and motor control within a single embodied platform, enabling interaction with human environments in fundamentally new ways. Humanoid robotics represents a qualitative leap in automation, bringing machines closer to human-like adaptability and dexterity.

Startups as the signal

As with every technological inflection point, startups provide the clearest signal of where innovation is heading. Across Europe, AI and robotics startups are transforming manufacturing, healthcare, home, aerospace and defense, agriculture, and home. In manufacturing, startups are automating complex processes to reduce costs and improve efficiency. Healthcare ventures are deploying AI-driven robotics for surgery, rehabilitation, and patient care. Agricultural robotics is advancing sustainability and productivity, while enabling technologies in AI, IoT, and machine learning are accelerating progress across the ecosystem. Over the past decade, Europe has seen some of the largest funding rounds in AI and robotics, underscoring the continent’s growing influence in shaping the global physical AI landscape.

According to PitchBook data, since 2020, approximately $23 billion has been invested in physical AI-related startups across nearly 2,000 deals. Over the past five years, aerospace and defense have emerged as the most heavily funded verticals, driven largely by companies such as Anduril, which has raised $6.3 billion to date, and Shield AI, with $1.3 billion raised. This is followed closely by humanoid robotics, led by companies such as Figure, which has raised $1.5 billion overall.

Humanoid Robotics | 2025 Report Teaser
Read Now

What are the enablers of physical AI?

The rise of physical AI is enabled by a convergence of technological, economic, and methodological factors:

  • Hardware commoditization: Falling costs have made high-performance sensors, actuators, and robotic platforms widely accessible. GPUs, batteries, and LiDAR sensors are now affordable even for startups, allowing more organizations to experiment, iterate, and scale intelligent machines. Data from the International Federation of Robotics suggests that the average cost of a robot has declined significantly over time, falling 80% just between 1995 and 2017.
     
  • Simulation-to-reality: Synthetic data generated in virtual environments enables safe and efficient training, bridging the gap between digital models and real-world conditions. This accelerates development, reduces risk, and improves reliability.

    Startup example: 
    Saturn Dynamics, a company accelerated through the Takeoff Accelerator Program, is developing a generative world model and virtual simulator powered by an AI-first, physics-aware engine. Its mission is to deliver accurate and controllable simulations for autonomous systems.
     
  • Edge computing and AI hardware: Processing data directly on devices allows real-time learning, low-latency responses, and enhanced privacy. Edge computing empowers robots, drones, and autonomous systems to perceive, decide, and act efficiently in dynamic environments, turning AI intelligence into practical impact.

    Startup example: 
    Gemesys, a Plug and Play portfolio company, is developing an analog chip inspired by the way the human brain works.
    The chip is designed to enable both training and inference directly on the device, completely avoiding dependence on the cloud or large data centers.

What are the challenges of physical AI?

Physical AI faces three fundamental challenges that distinguish it sharply from purely digital AI systems.

  • Robustness and safety: Operating in unpredictable real-world conditions, physical AI decisions have direct consequences. Errors can cause physical damage, safety risks, or environmental harm, making rigorous testing, certification, and monitoring essential.

    Startup example:
    Axion Robotics provides a universal trust layer for humanoid and autonomous robots, combining safety validation, live compliance monitoring, and continuous assurance scoring for regulators and enterprise buyers.
     
  • Scalability of physical learning: Learning in the real world is slower and costlier than digital training, as trials consume energy, wear materials, and involve safety constraints. Simulation, synthetic data, and transfer learning are critical to making physical systems as adaptable as digital ones.

    Startup example:
    Cyberwave builds the operational bridge between artificial intelligence and physical machines by abstracting hardware into programmable digital twins. This enables developers to simulate, control, and orchestrate machines with just a few lines of code.
     
  • Data orchestration: Physical AI relies on continuous, heterogeneous data from sensors, simulations, and control systems. Fragmentation and asynchrony make integrating these streams challenging, limiting scalability and robustness.

    Startup example:
    Rerun has developed a visual introspection platform for building time-aware visualizations of multimodal data, helping teams monitor, debug, and optimize complex physical AI systems.

Looking ahead

Physical AI is moving rapidly from promise to deployment. Its adoption is being driven by structural industrial pressures, enabled by technological advances, and validated by both incumbents and investors. At the same time, its success will depend on how effectively companies address safety, scalability, and orchestration at scale.

As intelligence becomes embodied and autonomy becomes operational, physical AI is poised to redefine how value is created in the physical economy over the next decade.

GET IN TOUCH

Ready to start your innovation journey?
PLUG AND PLAY AROUND THE WORLD

Discover our global locations

We have a community of experts all over the world. We work as a bridge between Silicon Valley and the rest of the world, making innovation open to anyone, anywhere.

Sign up for our global newsletter