The ‘cloud-only’ approach is reaching its limits
For a decade, centralization in the cloud has been the dominant reflex. Yet a flood of on-the-ground data — from sensors, machines, cameras, and connected devices — is testing this model. According to Gartner, about 75% of data generated by companies is now produced and processed outside a centralized data center.
The cause is as physical as it is economic. The number of connected objects surpassed 15 billion units worldwide in 2024 and could approach 30 billion by 2030. Pushing all of this data to a central cloud inflates bandwidth costs, introduces latency, and creates a dependency on network connectivity that many use cases cannot tolerate.
Three limits become plainly evident. The latency: certain applications (industrial robotics, autonomous vehicles, quality control via vision) require responses in a few milliseconds, incompatible with round-trips to a distant data center. The costs: transferring and storing vast volumes far away is expensive. The resilience: a critical operation cannot be halted every time the network cuts out.
Adding to these constraints is sovereignty. Some sensitive data — health, defense, industrial secrets — cannot or should not leave a given site or territory, for regulatory or strategic reasons. Processing it locally, at the edge, rather than exporting it to a central cloud that could be subject to extraterritorial laws, becomes another compelling argument for a distributed architecture. Latency, cost, resilience, and sovereignty thus converge to challenge the reflex of “everything to the cloud.”
Edge computing, a nearby answer
The edge computing concept involves processing data as close as possible to its source, instead of sending everything to a central cloud. A local server, an IoT gateway, or hardened equipment analyzes data on-site, only forwarding the essential — results, aggregates, alerts — back to the cloud.
The benefits are immediate. The latency drops, enabling real-time responses; bandwidth needs shrink, since only useful data travels upward; and activity continues uninterrupted even during network outages, a hallmark of resilience. In manufacturing, edge gateways can cut latency by more than half and greatly reduce reliance on the cloud.
Edge is not a fad but a undercurrent, driven by Industry 4.0, the IoT, and AI. Global spending on edge computing grows by around 13% per year and could reach $317 billion by 2026. Use cases such as predictive maintenance, local monitoring, video analytics, or autonomous systems illustrate its hallmark benefits.
The factory serves as the most vivid illustration of this shift. A modern production line generates a flood of data — sensors on machines, quality-control cameras, collaborative robots — that would be imprudent and risky to move entirely to a distant cloud. Processing these streams on-site enables real-time defect detection, stopping a machine before damage occurs, or guiding a robot, without relying on a connection. It is precisely in these physical and mission-critical environments that the edge demonstrates its value most clearly.
Hybrid: the end of a binary choice
Edge does not replace the cloud: the two complement each other. The cloud excels at large-scale storage, advanced analytics, and training AI models; the edge excels at fast, local processing and resilience. The right architecture is not “all cloud” nor “all edge,” but a continuum where each data point is processed at the most appropriate place.
That is the very definition of a hybrid environment: an infrastructure that combines public cloud, private or on-premises resources, and edge elements, orchestrated together. For both industry and services, the challenge for 2026 is no longer to pick a side but to design this distributed architecture and orchestrate it intelligently.
This decision goes beyond technology. It touches on costs (edge requires initial hardware investment but reduces ongoing expenses), on resilience (business continuity), and on sovereignty (local processing of sensitive data rather than exporting it). It becomes a driver of competitiveness and a strategic concern, not just an IT issue.
Where to start
Approaching hybrid environments and the edge does not require an immediate, radical transformation of everything. The proven path is a gradual one:
- Start from use cases: identify applications that are genuinely constrained by latency, bandwidth, or resilience — these are the first edge candidates.
- Think about data placement: decide, for each data stream, where it is most appropriate to process it (cloud, on-site, edge).
- Begin with a pilot: test at a site or on a production line before scaling up.
- Plan for orchestration and security: managing a distributed infrastructure at scale requires dedicated tools and practices.
The challenge for a leader is not to pit cloud against edge, but to orchestrate the continuum between the two. As data begin to originate in the periphery and AI makes its way into the shop floor, the ability to process each stream where it makes the most sense becomes decisive for performance, resilience, and sovereignty. Hybrid environments are not an extra layer of complexity to endure; when well designed, they are the architectural answer to a world where data no longer lives solely in the cloud.
A final point deserves attention: the edge is also a enabler of on-site AI. Running an AI model directly at the edge — to analyze a video stream, detect an anomaly on a machine, or guide a robot — requires local compute power that only the edge can provide. As AI spreads through physical operations, mastery of the edge determines how much value can be extracted. Cloud and edge are therefore not rivals but two faces of a single modern infrastructure, where intelligence is deployed from the center to the periphery.
This content is published by Mentioned
