Integrated Computing and Control — The Next-Generation Standard Architecture for Industrial Control Systems
In the past, PLCs handled control, industrial PCs handled vision, and servers handled AI — but the coordination among these multiple devices has become increasingly complex. Today, integrated computing and control is emerging as the new direction. The real challenge is not just putting AI into a PLC, but ensuring that real-time control and AI inference can coexist stably over the long term. This i
If your future PLC still needs an accompanying industrial PC, your control architecture may already be outdated.
Why is that?
Over the past decade, a classic combination has taken shape in factories.
PLC handles control.
Industrial PC handles vision.
Servers handle AI.
Edge gateways handle data collection.
Each does its own job, and the division of labour looks clear.
But the problems are also becoming more apparent.
More equipment, more complex networks, and more difficult data synchronisation.
Many projects end up stuck – not because the AI model is inaccurate, but because control, vision, and diagnostics operate like separate combat units, with ever-increasing coordination costs.
01 A Shift Underway
In recent years, a shift has become increasingly evident.
More and more vendors are no longer promoting "AI + PLC", but are instead embracing a new architecture – Unified Control and Computing.
Many assume this is simply cramming an industrial PC into a PLC.
That is not at all the case.
What it truly aims to solve is the hardest problem in industrial control:
Can real-time control and AI inference coexist within the same system?
02 An Inherent Contradiction
These two tasks are fundamentally contradictory.
PLC demands determinism.
1 millisecond means exactly 1 millisecond.
Every control cycle must remain consistent.
Robots, machine tools, and motion control all rely on this stable operation.
AI, by contrast, demands computing power.
A single inference may take 10 ms or 50 ms, and compute resources fluctuate with the model.
If AI consumes control resources, control cycles will jitter.In mild cases, shutdowns; in severe cases, mechanical collisions.
That is why, in the past, people preferred to buy an extra industrial PC rather than put AI directly into the controller.
03 The Real Challenge
The real challenge has never been about having enough computing power.
It is about how to keep "determinism" and "uncertainty" stably co-existing within the same system over the long term.
This is also why, over the last two years, the development of domestic real-time operating systems, multi-core processors, and industrial control platforms has drawn increasing attention.
What they truly solve is not who runs faster.
Rather, it is who can isolate control tasks from AI tasks – giving control top priority, allowing AI to fully utilise the remaining computing capacity, and ensuring the two do not interfere with each other.
Only when this problem is solved can Unified Control and Computing deliver real value.
04 It Changes More Than Just a Controller
Why does this matter?
Because what it changes is not just a single controller, but the entire industrial software architecture.
In the past, a robot might require a PLC, a vision industrial PC, data acquisition devices, and an AI analytics server to work together.
In the future, these capabilities will continue to converge onto a single platform.
Control, vision, quality inspection, equipment health monitoring, and predictive maintenance are no longer four separate systems, but different capabilities within a single system.
05 A Shift in Competitive Dimensions
This means the competitive focus in the industrial control industry will also change.
The past was about who controls more precisely.
The future will be about who can ensure real-time control, run AI continuously, and maintain system stability and reliability for hundreds of thousands of hours.
These are two entirely different competitive dimensions.
So, I increasingly believe that Unified Control and Computing is not a new product, but rather the new standard for next-generation industrial control systems.
Just as PLC replaced relays – not because it had more functions, but because it represented a new control architecture – the same applies to Unified Control and Computing today.
It is not simply putting AI into a controller.
Rather, it gives the controller, for the first time, the ability to continuously understand, analyse, and optimise equipment.
The controller of the future will not just execute actions.
It will become the brain of the equipment.
And the biggest competition in the industrial control industry over the next decade will most likely take place right here.