Kubernetes at the Edge • Charles Humble & Hannah Foxwell • GOTO 2026

In a conversation with Hannah Foxwell, author and consultant Charles Humble dives deep into the concepts from his e-book, “Kubernetes at the Edge,” providing a nuanced view of edge computing, its real-world applications, and the broader responsibilities of the tech industry.

Defining “Edge” Computing

Humble clarifies that in the context of his book, “edge” refers to a physical location of computing and storage, rather than a network topology. The primary focus is on two main types:

• Far Edge: Infrastructure located in immediate proximity to end-users and devices, such as at the base of a cell tower or in a point-of-sale system.
• Device Edge: Specialist sensors and controllers, like industrial sensors or medical imaging scanners.

This is distinct from “near edge” (small, localized data centers) and the network security definition of edge (e.g., accessing a corporate network via VPN). The core challenges at the edge often revolve around constraints like size, weight, and power (SWaP), as well as intermittent network connectivity—a factor many software systems are not designed to handle.

Real-World Use Cases and Architectures

The discussion highlights several inspiring and practical applications of edge computing.

Precision Agriculture and Healthcare

• Agriculture: Tractors and combine harvesters are now essentially “huge computers on wheels.” Edge computing enables precision agriculture, where technologies like drones and machine learning can create detailed field maps to detect stressed crops or weeds. This allows for targeted application of water, nutrients, or treatments, increasing crop yields while reducing reliance on fossil fuels and pesticides, which is critical in the face of climate change and growing global populations.
• Healthcare: The company RapidAI provides a powerful example. By using ML-based image processing on MRI scans directly at the edge (in the hospital), they have successfully extended the critical treatment window for stroke victims from approximately 6 hours to 24 hours. This has a profound, life-saving impact, especially for patients in rural areas far from specialized hospitals.

Retail and Day-Two Operations

Retail provides a common but complex edge environment. In-store systems, from point-of-sale to inventory management, must function reliably even with poor internet connectivity, as supermarkets can act like Faraday cages.

A key challenge that many organizations underestimate is Day-Two Operations. Humble stresses the need to move beyond manual “truck roll” deployments. Success at the edge requires extending DevOps and infrastructure-as-code practices to these remote environments. This involves:

• Automating deployments through a common platform.
• Using version control for remote infrastructure configuration.
• Implementing centralized, common monitoring to manage edge sites like “cattle, not pets.”

Vendor Selection and Strategy

When choosing a Kubernetes distribution for the edge, the answer often “depends” on the organization’s specific needs and existing infrastructure. Humble offers a breakdown of the landscape:

• Specialists: For greenfield projects, specialist vendors like Sidero (with its opinionated, secure Talos Linux) and Spectro Cloud offer powerful, focused solutions. While Sidero has a steep learning curve due to its unconventional design (no bash, no SSH), it provides a highly declarative and secure platform.
• Established Players: Companies already invested in an enterprise Linux ecosystem may find it easier to adopt solutions from Red Hat (Device Edge) or SUSE. While these are strong products, they are part of a broader portfolio and may lack the specific focus of the specialist vendors.

The decision should be guided by factors like the required level of GPU support for ML workloads, existing team skills, and purchasing relationships.

Sustainability and Green Software

Humble, a passionate advocate for sustainability, explains the significant environmental benefits of edge computing.

Operations First, Code Second

A crucial argument is that for sustainability, operations efficiency trumps code efficiency. This is due to a concept called “energy disproportionality,” where servers consume a significant amount of power even when idle.

“In most data centers that I go into, more than half of the hardware in that data center is not doing anything at all. You can literally turn it off and no one would notice.”

Optimizing operations by turning off “zombie servers” and improving hardware utilization (aiming for 50-80%) yields far greater energy savings than rewriting an application in a more performant language like C++ or Rust. This principle is equally, if not more, important at the edge, where a lack of a robust asset inventory can lead to forgotten, insecure, and wasteful hardware running for years.

Reducing Network Traffic

With network transmissions accounting for over half of the IT sector’s energy consumption, edge computing offers a direct solution. By processing data locally, it drastically reduces the amount of information sent to the cloud. This is particularly relevant for AI, where federated learning allows models to be trained on-device, with only the small deltas being sent back to a central server, enhancing both privacy and energy efficiency.

AI, Responsibility, and the Future of Tech

The conversation concludes with a frank discussion about the tech industry’s role and responsibility, particularly with the rise of generative AI. The recent hype around Large Language Models (LLMs) has unfortunately pushed sustainability down the priority list for many companies.

Humble argues for a more responsible approach:

• Use the Right Tool: Instead of defaulting to a massive, general-purpose LLM, organizations should consider smaller, specialized models that are more efficient, reliable, and cost-effective for specific tasks.
• Acknowledge the Ethical Debt: The industry needs to confront the fact that many foundational models were trained by scraping the internet and using the intellectual property of creators without consent or compensation. Humble warns, “There will be a reckoning.”
• Grow Up as an Industry: Technologists must take responsibility for what they build. To avoid a future where heavy-handed regulation stifles innovation, the industry must proactively address these ethical issues and have a broader conversation about the societal impact of its creations.

Ultimately, Humble encourages everyone in tech to get involved and think critically about how technology can be used for good, ensuring that this pivotal moment in the industry leads to positive outcomes for humanity.

• Tags:
• Feature
• Kubernetes
• AI Ethics
• Green Software
• Federated Learning
• Sustainability
• K8s
• Edge Computing