Physical AI: From Thought to Action

Without doubt, artificial intelligence has been the hottest tech topic of recent years. To date, generative AI and agentic AI have been center stage. But 2026 will see a marked shift – to Physical AI, as AI evolves out of the digital and into the physical realm.

This development marks artificial intelligence’s transition from an efficiency tool to a strategic enabler, with its first deployments in critical sectors like industry and defense. The growing transformative potential of Physical AI is also visible in the European M&A space – for example, in Softbank’s recent USD 5.375-billion purchase of ABB Robotics.

Such is the promise of Physical AI that market observers expect there to be almost 1 billion humanoid robots in use by 2050, with the vast majority performing industrial and commercial tasks.

Physical AI 101

But what exactly is this new type of AI and how does it differ from its predecessors? Physical AI refers to systems that perceive, reason, and act in the physical world by integrating various technologies, such as sensors, machine learning/AI models, and robotics or hardware systems.  

The following key characteristics distinguish Physical AI from other form of artificial intelligence:

• Adaptiveness: It learns from real-world data to improve performance over time.  

• Context-awareness: It understands and responds to its environment.  

• Action-orientation: It executes tasks autonomously without human intervention.  

  
  

Unlike conventional automation, which is rule-based and static, Physical AI is dynamic and capable of learning, adapting, and working autonomously.  

The Latest Step in AI Evolution

Physical AI can be seen as the latest step in the evolution of artificial intelligence. While generative AI concentrates on understanding and creating, and agentic AI centers on decision-making, Physical AI focuses squarely on taking action in the real world.  

Initial Limitations: The Situation in 2025

As recently as 2025, Physical AI was regarded as promising but not yet transformative. While there was considerable hype around the technology, deployments were confined to pilot and experimental projects in controlled environments. In fact, there were a number of significant limitations that put the brakes on the widespread adoption and transformative potential of technology.

One major challenge was limited generalization. Physical AI systems were task-specific and unable to adapt to different use cases. What’s more, the high cost of hardware and deployment made implementing the tech prohibitively expensive for many organizations.

Restricted Applicability/Excessive Reactivity

Another critical limitation was reliance on structured environments. To function effectively, Physical AI systems needed highly controlled settings, restricting their applicability in dynamic or unpredictable scenarios. In addition, most systems were reactive rather than predictive. In other words, they responded to inputs but couldn’t anticipate or forecast outcomes.  

These limitations showed that further advances in technology and deployment strategies were needed to unlock the full potential of Physical AI. But while its impact was not all it could have been, Physical AI still demonstrated its potential in specific, controlled scenarios.

2026: The Inflection Point  

By 2026, greater technological maturity, cost reductions, and enhanced AI capabilities combined to enable a shift toward widespread adoption of Physical AI. This inflection point was driven by four key developments:  

1. The Shift from Automation to Autonomous Operations  

Systems evolved from executing predefined tasks to acting independently and adapting in real time. Examples here include autonomous surveillance, real-time threat response, and self-optimizing production.  

2. Dual-Use Convergence (Defense ↔ Industry)  

The same Physical AI capabilities are being used in both industry (for efficiency) and defense (for security and operations). This can be seen, for instance, in drones used for inspection and reconnaissance, and robotics for logistics and defense support.  

3. Real-Time Decision Cycles (OODA Loop Compression)  

Physical AI enables continuous sensing, instant analysis, and immediate action. This compresses the OODA loop (Observe-Orient-Decide-Act) and shifts advantage to those operating at machine speed.  

4. The Move to Edge Intelligence 

AI is shifting from centralized cloud systems to edge-based AI, operating on devices, in vehicles, or within infrastructure. This delivers benefits by reducing latency, removing connectivity constraints, and safeguarding data sovereignty – all of which are critical factors for defense and industrial environments.  

Thanks to these developments Physical AI has now successfully transitioned from experimental deployments to transformative, real-world applications.

Industry Implications: Where Physical AI Matters Most

The sectors in which Physical AI has the greatest impact are those characterized by high labor intensity, physical operations, real-time requirements, and risk exposure. These attributes make the tech especially beneficial in industries like manufacturing, logistics and supply chain, defense, and energy and infrastructure.  

More Autonomy, Better Predictions, and Greater Speed

In manufacturing, Physical AI enables autonomous factories, AI-powered predictive maintenance, and advanced robotics. In the automotive sector, for example, autonomous factory robots streamline production processes while AI-powered predictive maintenance analyzes machine data to prevent failures before they occur.  

In logistics and supply chain operations, Physical AI is driving the development of autonomous warehouses and delivery systems. Next-generation fulfillment centers leverage AI-driven robots and automation systems, while smarter fleets of robots are now being built.

Enhanced Decision-Making, More Effective Maintenance, Plus Cost Savings

Physical AI is enhancing surveillance and reconnaissance, autonomous systems, and accelerated decision-making in the defense sector. AI-driven decision acceleration analyzes satellite and drone data to identify potential threats, significantly improving operational efficiency, for example. 

In the energy and infrastructure sector, Physical AI supports grid operations and inspection and maintenance activities. For instance, AI-based solutions for grid predictive maintenance have reduced transformer failures by 48%, resulting in an estimated USD 800,000 in annual savings in operations and maintenance costs.  

The Strategic Implications of Physical AI  

Physical AI has profound strategic implications across four key areas, reshaping how organizations and nations approach technology and operations.  

First, the distinction between technology strategy and operational strategy is becoming increasingly blurred. Physical AI is a core operational capability that impacts entire businesses – and not just their IT.  

Meeting the New Needs of IT/OT Convergence

Second, the convergence of Information Technology (IT) and Operational Technology (OT) calls for new architectural approaches and governance frameworks to ensure seamless and secure operations.  

Third, governance must expand to cover both cyber and physical risks. As Physical AI systems gain ground, new risks such as physical safety, system failures, and cyber-physical attacks will require governance models that go beyond traditional IT security.  

Finding the Right Approach to Deployment

Finally, Physical AI has significant implications for strategic positioning and geopolitics. It is emerging as both a competitive differentiator and a critical national capability. Leaders must address core questions, including whether to build or partner, how to balance control and dependency, and how to prioritize innovation versus sovereignty.  

The Transformative Power of Physical AI 

The evolution of AI from digital to agentic to physical represents more than just an incremental change; it is a structural transformation. Physical AI doesn’t simply extend existing capabilities; it marks a fundamental shift in how AI interacts with and impacts the real world.  

The potential of Physical AI was already clear in 2025. In 2026, the time has come to deploy the tech at scale. But one critical question remains: Now that AI is no longer just augmenting work but starting to actually perform it in the physical world, how prepared is your organization to design, deploy, and govern AI systems capable of acting in the real world?  

Your Thoughts?

Want to find out more Physical AI and what it could mean for your organization? Then feel free to reach out to me. How do you see this highly promising technology shaping business and beyond? Let me know by leaving a comment below.