By Rajesh Subramaniam, CEO & Founder, Embedded Systems
Until recently, the global discussion about AI has been focused on creating larger models, training on bigger data sets, and investing in more data centres.
However, today, as AI moves beyond digital interactions and starts being used in critical spaces such as factories, hospitals, transportation systems, and telecom networks, the conversation has changed. And that shift highlights the challenges that come with this new direction.
Data should no longer be processed at centralised cloud infrastructures. Instead, they must be computed where they are created, allowing for real-time decisions. This need is giving rise to a new phase of innovation called Physical AI, where intelligence is directly embedded into real-world systems and infrastructure.
Physical AI – Its Definition and Demands
In simpler terms, Physical AI is the type of AI that can be embedded in everyday devices. When integrated with AI, these devices can understand their surroundings, make decisions, and interact with the physical world.
Traditional AI systems that are currently used in businesses depend on cloud-based processing, which can result in some delays. And this might be tolerable in such time-insensitive operations.
However, the same can’t be said for mission-critical environments where even a few minutes can make a big difference. For instance, a manufacturing system monitoring production lines, a robotic arm performing surgery, or a telecom network handling traffic cannot afford to wait for data to be sent back and forth to a distant data centre before making decisions.
This shows the need for Physical AI, but it also throws light on a key challenge.
With the influx of more connected devices that are necessary for daily operations, and their need to constantly transmit, issues like latency, high bandwidth costs, energy usage, and reliance on network availability are unavoidable.
Therefore, rather than relying only on centralized infrastructure, AI must be brought closer to where data is generated and decisions are made – the network edge. And this need paves the way for distributed intelligence.
The Enablers of Distributed Intelligence
Physical AI needs more than just powerful models; it requires systems that can make decisions and act in real time. This is where AI agents and Edge AI come into play.
AI agents are independent entities that allow systems to interpret information, reason with multiple inputs, and learn with minimal human help. As AI becomes a crucial part of operational environments, these abilities are increasingly important for handling complex and dynamic situations. This trend is already evident with around 40% of organizations having implemented agentic AI, while another 50% plan to do so by 2026; with adoption particularly strong across manufacturing, retail, healthcare, and life sciences.
At the same time, Edge AI allows intelligence to function closer to the source of data. By processing information directly on devices and edge networks, organizations can cut down on delays, avoid depending on the cloud, and improve their overall resilience. According to IDC, world spending on edge computing is expected to surpass USD 378 billion by 2028, driven by the need for real-time analysis, automation, and better customer experiences. As organizations focus more on quick decision-making and local processing, edge infrastructure is becoming a vital base for the next generation of AI deployments.
While Edge AI helps with local data processing, AI agents strengthen the reasoning part, helping intelligent systems to sense, decide, and respond in real time across various physical environments.
Why India Is Well Positioned for the Physical AI Era
According to BCG’s AI at Work 2026 report, India is leading the world in workplace AI adoption. This, along with nationwide 5G deployment and rising investments in manufacturing, healthcare improvements, and smart mobility, shows that the country is creating an environment where intelligent systems can thrive.
In manufacturing, Physical AI can help with predictive maintenance, quality assurance, and automated process improvements. In healthcare, intelligent medical devices and remote monitoring systems can aid in better access to medical support. Telecommunications industry can benefit from AI-driven automation to enhance performance, manage data traffic, and boost operational efficiency. As intelligence becomes part of such key domains, the ability to effectively use AI in real-world settings may be as important as developing the models themselves.
For India, this presents an opportunity to emerge as a global leader in deploying intelligent systems.
Building the Foundation for Physical AI
Although Physical AI has huge potential, scaling it comes with its own challenges.
Running AI agents on devices often needs more processing power, extra memory, and highly optimized software that can balance performance with efficiency. The systems will also have to meet strict latency requirements while working within limited power supplies. As organizations use AI across a wide range of devices and infrastructure, challenges around model optimization, hardware compatibility, and large-scale deployment will become even more critical.
Overcoming these obstacles will be essential for unlocking the full potential of distributed intelligence and catalysing the adoption of Physical AI in different industries.
The Way Forward
The next stage of AI innovation will not only be about algorithm improvements, but also where intelligence is located.
The advent of Physical AI marks a shift from centralized intelligence to distributed intelligence, where AI agents, edge computing, and connected infrastructure work together to make decisions closer to the action. For India, this transition offers an opportunity to move beyond being just a leader in AI adoption and becoming a pioneer deploying intelligent systems that power the economy. The future of AI may not be entirely built in data centres, but at the edge, where the physical and digital worlds collide.