Physical AI explained: Why a bigger shakeup may be round the corner

The most visible disruption caused by artificial intelligence (AI) has been in generating content, images, videos, reports, infographics and so on with applications like making chatbots smarter. But the bigger play is in embedding AI in physical products, robots, autonomous vehicles, drones etc. For example, robots used in factories or warehouses are task specific — like running a paint shop in car manufacturing units or lifting goods in warehouses. But when AI is embedded in these robots and physical products, they become smarter, more productive and capable of multi-tasking.

 

That shift marks a deeper transition underway across industries. It upgrades how robots and machines operate. Vishesh Rajaram, founding partner, Speciale Invest, says: “We are moving from AI that generates content to AI that acts in the physical world.” Speciale invests in deep tech ventures and has also backed CynLr, a Bengaluru-based robotics venture. It is a shift from words and pixels to motion, touch and decision-making in real environments — on factory floors, farms, warehouses and even inside controlled biological systems.

 

At its core, physical AI refers to systems that combine intelligence with physical action. As Mrutyunjaya Nadiminti, co-founder, CBO & co-CTO, Perceptyne Robots, explains, “unlike traditional AI, which focuses on data analysis, physical AI enables machines—such as robots and autonomous systems—to perceive their surroundings, make decisions and execute actions in real time.” 

 

Unlike traditional software AI that analyses data in the cloud, physical AI integrates vision, sensors, touch and advanced models, allowing machines to operate in dynamic environments. So, a robot is no longer just executing pre-programmed commands; it is responding to what it sees, feels and learns.

 

This distinction is critical. For decades, industrial automation has been rigid — machines designed to repeat the same task thousands of times with precision. But they struggled with variability. A slightly misaligned component, a new product variant or an unstructured environment could disrupt operations. Physical AI changes that. Nadiminti adds that embodied

 

AI brings intelligence directly into physical systems, allowing them to perform tasks with precision and responsiveness.

 

When AI is embedded into physical systems, automation shifts from fixed tasks to adapting to workflows. Raghav Gupta, vice president Endiya Partners says: “A robot becomes a continuously improving operator, process engineer, quality engineer and safety manager.  For enterprises this means higher uptime and lower dependence on manual intervention. In daily life, it translates into safer factories and better products.”

 

According to Deloitte Smart Manufacturing survey, research by Polytechnic University of Madrid and Siemens Electronics, robots use results in an 18-20 per cent drop in error rates and with physical AI, it can drop to near zero defects in hi-tech manufacturing.

 

Why does the intelligence layer matter?

 

Deep tech investors like Endiya Partners and Speciale Invest, tracking the space, believe the real value of physical AI lies not in hardware, but in the intelligence powering it.

 

According to Rajaram, physical AI is the next major frontier. ‘’What excites us most is not the hardware — it is the intelligence layer. For example, when we backed CynLr, we saw a team rebuilding the foundational perception stack for robotics, inspired by how the human brain actually processes the physical world,” he points out. Rajaram cites breakthroughs such as robots learning to handle a new object in under 15 seconds without months of retraining. The implication is profound: factories could become software-defined, adapting quickly to new products and processes.

 

So companies can upload a new design, say a new SUV model or a phone, and physical AI systems will interpret these and adjust according to new specs, rather than being built again for the new task. For hi-tech firms in India, this opens up new opportunities. With its engineering talent and manufacturing demand, India could move from being a consumer of automation technologies to a provider— much like it has done in the IT services business. On the shop floor, the impact is already becoming visible.

 

“AI will serve as the backbone of modern Indian shop floors, driving a shift towards efficiency and precision,” says Vinod Sharma, managing director, Deki Electronics (which makes capacitors), and chairman of CII’s National Committee on Electronics Manufacturing. Predictive maintenance systems now forecast machine failures before they occur, while computer vision identifies micro-defects invisible to the human eye. But the next leap lies in combining these capabilities (like vision and computing power) with physical systems.

 

For instance, “physical AI will be applied to solve one of the biggest gaps in industry—automating dexterous, real world tasks that still rely on human hands,” says Nadiminti of Hyderabad-based Perceptyne Robots. In sectors like manufacturing, physical AI can enable precision-driven operations like assembly, inspection and machine tending — ensuring consistency in high-volume environments. By combining vision, force and tactile sensing, robots can now handle variability on the shop floor—adjusting to differences in components, orientation or environment.

 

The approach is pragmatic. Rather than aiming for fully autonomous factories, companies are deploying robots for specific high-value tasks, working alongside humans.

 

Beyond factories: farms that behave like labs

 

Physical AI is also expanding beyond traditional industrial settings into unexpected domains like agriculture. At agriculture technology venture Panama Hydro X, AI is embedded into controlled cultivation environments, where sensors, imaging systems and IoT infrastructure work together to monitor plant health, disease, nutrients and micro-climate conditions.

 

Vivek Raj, founder & CEO, Panama Hydro X, elaborates. “Unlike traditional automation, which follows a predefined set of instructions, physical AI systems are adaptive. They learn from live biological responses.”

 

In one instance, physical AI models analyse leaf-level data to detect early stage plant stress and disease, even before visible symptoms appear. Raj adds that these systems operate in a closed loop framework, where sensing, analysis and control continuously refine each growth cycle. The outcome is not just automation, but consistent, traceable and repeatable   biological production.

 

A growing market

 

Globally, physical AI is still in its early stages but scaling rapidly. According to US-based market research and consulting company Grand View, the AI-in-robotics market, was valued at about $12.7 billion in 2023 and is projected to grow nearly 10x to over $120 billion by 2030.

 

Large-scale deployments are already visible—with companies like Tesla and Amazon deploying physical AI systems in factories and warehouses. In India, adoption is nascent but accelerating. AI usage in manufacturing has risen from 8 per cent to 22 per cent in FY2024, according to TeamLease Digital. Industry estimates suggest the domestic AI-in-manufacturing market could grow from about $1.2 billion to $8 billion by 2030.

 

Yet, as Nadiminti points out, “Adoption of physical AI in India and globally is still at a very early stage, with a large share of such operations continuing to be manual.” However as enterprises aim to be globally competitive, more physical AI systems will be deployed. “Any new technology only helps amplify our capabilities. AI-driven robots are the way forward. Today Indian manufacturing in electronics has low local value addition and is mostly assembly of imported components. That’s partly because of lack of tech deployment compared to global best,” explains Deepak NG, managing director, Dassault Systèmes India. The French software company provides software to create virtual twins (virtual replicas) of products, processes and systems for industries like automotive, aerospace, defense and life sciences. The India division aids in design, simulation, and manufacturing innovation.

 

As advances in sensors, computing power and AI models converge to make physical AI viable, systems will become more reliable, adaptable and easier to integrate into existing workflows. Physical AI vision is clear: factories that can reconfigure themselves, farms that optimise output autonomously, and machines that operate with a level of awareness once reserved for humans. 

AI in physical world 

What is physical AI

  Fusion of AI with robotics and hardware systems to enable machines to sense, act, and adapt in the real world

  Opportunities…

  The opportunity set is vast and includes enhancing capabilities of manufacturing companies, logistics, healthcare, agriculture, defence and even consumer applications

  …and challenges

 

• High capital costs make deployment expensive, especially for small manufacturers, when human labour is much cheaper
• Talent gaps in AI-robotics integration persist
• Energy demands require advances in batteries and efficiency
• Safety and governance concerns demand robust regulatory frameworks 

Source: Industry

 

The writer is a New Delhi-based independent journalist