Mikael Holmberg at BalticNOG: Untangling the Myths Around AI in Networks
In an industry that moves fast and markets even faster, few voices bring the depth of experience and clarity of thought that Mikael Holmberg does. With over 30 years in networking and telecommunications and the title of Distinguished Engineer at Extreme Networks, Mikael has seen technology trends rise, fall, and evolve—often distorted by hype before their true potential is understood.
At BalticNOG, Mikael will deliver a presentation that challenges one of the most pervasive narratives in today’s tech role of artificial intelligence in modern networking. His “Myth Around AI Networks” session explores the promises to cut through inflated expectations and spotlight the technical realities often lost in vendor messaging and marketing campaigns.
Beyond the Buzzwords
Artificial Intelligence has become a catch-all term, applied liberally across industries to evoke innovation, efficiency, and automation. In networking, the term has been used to describe everything from anomaly detection to traffic prediction, fault recovery, and beyond. But as Mikael points out, AI is rarely the game-changing, autonomous intelligence people envision in this context.
The presentation opens by dissecting what AI in networking entails. It’s not artificial general intelligence or decision-making in the human sense. Instead, it relies on structured models, data pipelines, and rule-based systems—tools that are incredibly powerful when applied correctly, but far from autonomous.
One of the central ideas Mikael explores is that AI doesn’t replace solid engineering—it depends on it. Every AI-assisted insight, predictive suggestion, and intelligent alert must be grounded in data from a robust, observable, and scalable network fabric. Without that foundation, the AI layer is little more than a marketing veneer.
The Infrastructure Beneath the Intelligence
Contrary to some misconceptions, today’s high-performance Ethernet networks are not bottlenecks to AI development—they’re enablers. Mikael walks the audience through how modern switching fabrics—capable of 51.2 Tbps and beyond—form the backbone of AI infrastructure, even in the most demanding environments such as GPU clusters and hyperscale data centers.
He challenges long-standing beliefs that specialized interconnects are essential for AI workloads. In reality, the evolution of Ethernet has closed many of those gaps, offering standardization, scalability, and cost-efficiency that dedicated alternatives struggle to match.
This isn’t a rejection of AI or automation. Instead, it’s a call to reevaluate where real value lies. AI can optimize, predict, and assist—but it does so effectively only when the network is designed to support that intelligence. That includes having the right telemetry, observability tools, and well-structured data architectures. AI is only as good as the visibility and control the network affords.
Engineering Wisdom in the Age of Hype
Throughout his talk, Mikael deconstructs several familiar narratives: that AI can operate without human input, design or repair networks on its own, or obsoletes the need for architectural expertise. These ideas, he argues, are not only misleading but dangerous. They lead to overengineered solutions, wasted investment, and operational complexity.
Instead, Mikael advocates for a more disciplined and realistic approach where AI is a tool, not a solution. One where the focus remains on engineering fundamentals—efficiency, scalability, reliability—and only layered AI is thoughtfully on top.
He brings examples from the field that illustrate how practical automation, combined with clear visibility and strong architectural decisions, delivers far more impact than any black-box “AI” system. His framing invites engineers, architects, and decision-makers to remain skeptical of grand promises and instead pursue solutions that are explainable, maintainable, and rooted in operational needs.
A Talk for Builders, Not Buzzwords
Mikael’s session isn’t about riding the AI wave—it’s about shaping the conversation for those who have to live with the consequences of technical decisions. It’s for the builders: the engineers working in ISPs, cloud environments, data centers, and enterprise networks who know that reliability is earned, not marketed.
If you’ve ever questioned the practical limits of AI in networking or felt the tension between automation potential and operational complexity, this talk will resonate deeply. Mikael doesn’t promise silver bullets, but he does offer clarity, perspective, and decades of experience that can help guide decisions in an increasingly noisy landscape.
Join us at BalticNOG 2025 and hear directly from one of the most grounded voices in the networking community. Building the future of networking requires more than bold claims—it requires engineering wisdom.
