Insights

Edge Computing: Process Where Data Is Born | SUMāTO

Written by Andrés Lozada | Jul 9, 2026 7:18:48 PM

A few days ago, walking a manufacturing plant alongside a client in northern Mexico, I stopped in front of a packaging line that generated more data in an hour than many companies produce in a month. Vibration sensors, inspection cameras, temperature controllers: everything measuring, everything talking. And then the operations lead asked me the question that defines this moment in the industry: "Does all of this really have to travel to the cloud for something to decide whether or not I stop the machine?" That question, so concrete, sums up why edge computing has stopped being an academic curiosity and become a business conversation. Let me share how I see it in early 2019.

In short: Edge computing means processing data close to where it is generated, instead of sending it all to a remote data center. It does not replace the cloud: it complements it. When latency, bandwidth or autonomy matter, deciding at the source changes the rules of the game.

What exactly is edge computing?

The term "edge" refers to the edge of the network, the point where physical devices meet the digital world: a machine on the plant floor, a point of sale in a store, a camera, a sensor. Edge computing proposes placing computing capacity right there, or very close by, to analyze and act on data in the place where it is born.

For years, the dominant model was simple to describe: devices captured information and sent it to the cloud, where it was stored, processed and a response was returned. It works very well for enormous volumes of historical analysis. But it begins to show strain when the response has to be immediate, when connectivity is intermittent, or when moving every byte to a data center is expensive or, quite simply, unnecessary.

Why process near the source?

There are three reasons that come up again and again in conversations with our clients. I'll sum them up this way:

  • Latency: the time it takes for a data point to travel, be processed and return. When a machine must stop within milliseconds in response to an anomaly, that round trip to the cloud is too long. Deciding locally eliminates that delay.
  • Bandwidth: sending every reading from every sensor saturates networks and drives up operating costs. Processing at the edge allows for filtering: only what matters is transmitted—an alert, a summary, an exception—instead of the full torrent.
  • Autonomy: an operation cannot come to a halt because the internet link went down. Edge computing lets a site keep running and making decisions even if it temporarily loses its connection to the cloud.

To these three we can add a fourth factor that carries more and more weight: control over where certain sensitive data resides. Processing it locally and sending only the aggregate helps govern that information better.

Does this mean the cloud loses relevance?

Quite the opposite, and I want to be very clear on this point because it causes confusion. The edge is not a rival to the cloud; it is its natural partner. The most useful way to understand it is to think of a division of labor:

  • The edge handles the immediate, the local, what cannot wait: detecting, filtering, reacting.
  • The cloud handles the large-scale and the strategic: training models, consolidating information from many sites, preserving the historical record, providing the big-picture view.

A detection model can be trained in the cloud with data from dozens of plants and then deployed to the edge to run in real time. The edge, in turn, returns to the cloud the results that enrich the next learning cycle. It's a virtuous circle. That is why, when we design cloud infrastructure architectures, we now think of them together with the edge, not as separate worlds.

Use cases where the edge already makes a difference

I'm not talking about a distant future. In early 2019 we already see concrete applications in the region:

  • Industrial IoT: predictive maintenance based on vibration, temperature and consumption. The edge detects the anomalous pattern and triggers the alert before the failure becomes costly.
  • Retail: foot-traffic analysis, shelf inventory management and in-store experiences that cannot depend on the connection being perfect at peak hours.
  • Manufacturing: visual quality inspection with cameras that evaluate every part on the line, discarding defects without taking the decision off the plant floor.
  • Logistics and energy: monitoring of distributed assets—fleets, substations, remote facilities—where connectivity is precisely the weakest link.

How does the edge relate to analytics and intelligence?

Here is, to my mind, the most interesting part. Processing at the edge is not only about filtering data: it is about bringing intelligence to the point of action. Analytics models that once lived exclusively on large servers are beginning to run on ever more capable and efficient devices.

That turns the edge into the first point where a data point becomes a decision. But for that decision to be a good one, the model behind it has to be well built, well fed and up to date. That is why I insist to our teams that an edge strategy without a solid data analytics strategy is a house with no foundation. The edge executes; analytics gives it judgment.

What challenges should you anticipate?

I don't want to paint a picture without nuance. Adopting edge computing involves new decisions:

  • Managing many devices: administering, updating and securing dozens or hundreds of distributed nodes demands discipline and the right tools.
  • An expanded attack surface: every computing point at the edge is also a point to protect. The surface to guard grows.
  • Architecture design: defining what is decided at the edge and what goes up to the cloud is not trivial; it is an engineering and a business decision at the same time.

None of these is insurmountable, but all of them carry weight. That is why it pays to start with a bounded scope and grow on the basis of real lessons learned.

Frequently asked questions

Will edge computing replace the cloud?
No. They solve different problems and reinforce each other: the edge provides immediacy and local autonomy; the cloud provides scale, historical memory and training capacity. The sensible approach is to combine them.

Do I need a large upfront investment to get started?
Not necessarily. The recommendation is to begin with a specific, measurable use case—one line, one store, one type of asset—to validate the value before scaling to the whole operation.

What kind of company benefits most?
Those with distributed physical operations and a need for fast response: manufacturing, retail, logistics, energy. Wherever latency, bandwidth or continuity matter, the edge has a great deal to contribute.

How do I know whether a process should be decided at the edge or in the cloud?
A good guide: if the decision cannot wait for the round trip to the cloud, or must keep working without a connection, it belongs at the edge. If it requires consolidating a lot of information or a historical view, it belongs in the cloud.

The first step

Edge computing is not just another tech fad: it is a different way of thinking about where intelligence happens in your operation. And as with almost everything in digital transformation, it isn't about adopting it wholesale in one leap, but about starting where it hurts most today: that decision that arrives late, that link that drops, that data that's costly to move.

At SUMāTO we help companies in Mexico, Colombia and the rest of LATAM identify exactly that point. We propose starting with a short assessment: mapping your data sources, understanding where latency or connectivity is costing you, and defining together what is best processed at the edge and what to leave in the cloud. From that comes a concrete roadmap, not an abstract promise.

If this conversation resonates with you, let's talk. Write to us through our contact page and let's arrange that first assessment. Processing where data is born could be the change your operation has been waiting for.