Cloud-Edge Synergy — Who Should Do the Heavy Lifting on Computing?

When building industrial internet, should the edge compute more, or should the cloud compute more?
This question appears to be about technical architecture.
But many projects have asked the wrong question from the very beginning.
Because what should really be asked is not:
"Should data go to the cloud or the edge?"
Rather, it is:
"Where should this data create value?"
These two questions are completely different.

Take an electric motor, for example.
Its vibration sensor generates massive amounts of data every second.
If you send all the raw data to the cloud for FFT, filtering, and anomaly detection,
it sounds very advanced.
But once the project scales, you will find that bandwidth, storage, and latency all become issues.
More importantly, most normal data does not actually need to be uploaded.
So the edge should do one thing first:
turn raw data into valuable data.
For example: filtering, noise reduction, FFT, feature extraction, and anomaly detection.
Even telling you directly:
this equipment is experiencing abnormal vibration right now.
This is what the edge does best.
Because it is closest to the equipment.
It knows what is happening on-site, and it does not need to wait for the network.

The core of edge computing is not about being "closer to the cloud," but about being "closer to the site."

But then another question arises.
Should everything be placed on the edge?
Not necessarily.
Because the edge has an inherent limitation:
it can only see itself.
It can detect abnormal vibration in a single machine.
But if you ask:
"Why has the failure rate for this type of equipment across the entire factory been rising over the past month?"
The edge cannot answer that.
Because this requires correlating data from dozens or hundreds of machines,
along with maintenance records, production processes, ambient temperature, runtime, and even data from different factories.
This is where the cloud's value becomes clear.
So what the cloud should really do is not do the edge's job.
Rather, it should do what the edge cannot:
see the big picture, analyse historical patterns, make correlations, train models, and accumulate knowledge.

This creates a clear division of labour:

The edge handles on-site closed-loop; the cloud handles global closed-loop.The edge answers: "What is happening right now?"The cloud answers: "Why is it happening? What should we do next?"

01True Cloud-Edge Collaboration: Execute, Learn, Upgrade

But there is another crucial step.
Many projects stop here.
The cloud finishes its analysis and produces a report.
But this is not yet true intelligence.
True cloud-edge collaboration should be:
the edge collects and executes, the cloud learns and optimises, and then the new models, rules, and knowledge are deployed back to the edge.
In other words:
edge executes, cloud learns, edge upgrades.
Once this loop is running, the system becomes smarter and smarter.

Therefore, debating whether cloud computing or edge computing is more important is like asking a factory:
Is the production workshop more important, or is the management department?
That is not how it works.
The workshop is responsible for getting things done, while the management system oversees the big picture and makes decisions.
The real question is:
are the two sides forming a closed loop?

02The Value of Future Industrial Software: Connecting the Chain

So the future competition in industrial software will not be simply about "going to the cloud."
Nor will it be about stuffing everything into an edge computing box.
What truly matters is whether we can connect this chain:

Equipment perception → Edge intelligence → Cloud cognition → Model learning → Capability deployment → On-site execution.

At this stage, cloud-edge collaboration is no longer just an IT architecture issue.
It is actually redefining:
how industrial software should understand a machine, a production line, and ultimately the entire factory.

In the final analysis,
the edge turns data into on-site capability, and the cloud turns on-site capability into enterprise wisdom.
This may be what cloud-edge collaboration is truly worth discussing.