For 14 years, our community has been at the forefront of industry shifts. We pioneered the SD-WAN marketplace, now a multi-billion dollar industry. We championed multi-cloud and hybrid-cloud strategies when single-cloud was the prevailing thought. Our journey continued through orchestrating and automating virtual and leased infrastructure, and tightly integrating networking and security. Today, ONUG stands on the precipice of our fifth wave: AI networking, evolving into the enterprise AI community. Our focus for 2026 and 2027 is clear: building an AI-native enterprise network fabric. 

The Double Tsunami: AI and Data

We are currently experiencing two monumental “tsunamis” that will either lead to destruction or unprecedented advantage, depending on how we ride them.

The first is the AI tsunami. AI capacity is growing exponentially. Consider the scale of Colossal, Elon Musk’s AI data center for XAI: from 100,000 GPUs consuming 150 megawatts in May, it has already surged to 250,000 GPUs using 450 megawatts. By next year, it’s projected to reach one million GPUs, demanding 1 to 1.2 gigawatts—enough to power the entire city of Denver. By 2030, we’ll see AI data centers with 5-10 million GPUs, consuming tens to hundreds of gigawatts. This phenomenal growth underpins a burgeoning “token economy,” with Microsoft, for instance, processing 500 trillion tokens in just three months.

The second is the data tsunami. The sheer volume of data flowing across networks is staggering. eBay processes 2-5 petabytes daily, while FedEx handles multiple petabytes every 30 minutes. This massive scale strains existing network protocols and device capabilities. Furthermore, AI workloads, both in the cloud and on-premise, introduce unpredictable and dynamic traffic patterns, making traditional network management increasingly complex. The operational data – system data, trace data, logs, alerts – is also growing geometrically, exemplified by Intuit pumping 100 petabytes of system data into its data lake last year. This necessitates a complete reimagining of how we build and operate these systems, a shift far greater and faster than the transition from SNA to TCP/IP in the 1990s.

The Pace of Autonomy and the Need for Readiness

How quickly are these waves approaching? When we consider “model autonomy”—how smart AI models are becoming—the pace is astounding. Today, an AI can complete tasks that would take a human an hour and a half. By 2028, this will extend to a full week’s worth of human effort, shifting incident response towards AI agents as humans take on supervisory roles requiring deep domain experience. By 2030, a month’s worth of human work could be done by AI, leading to “agentic operations centers” where humans focus on policy, governance, and exceptions. Some projections even suggest models processing 20,000 years of human work by 2030. This systemic shift impacts every facet of business, from HR to engineering. As John Chambers warned, 70% of today’s Fortune 500 could disappear by 2030 if they fail to embrace AI.

Our community’s readiness, however, presents a mixed picture. A recent poll revealed that only half of us believe our current network architecture needs significant changes for AI workloads, with three-quarters earmarking less than 25% of their infrastructure budget for AI modernization. This suggests a potential underestimation of the coming changes. Yet, there’s widespread agreement: 80% view agentic AI as a transformative force for enterprise infrastructure and business value. Our collective mission statement, distilled from your feedback, affirms: “AI is a transformative shift for enterprise infrastructure, automating workflows for efficiency and cost savings, enabling real-time decision support with proactive risk management, and doing so with a secure zero-trust governance framework that ensures business innovation scales safely.”

While many anticipate having AI-ready infrastructure in one to three years, the urgency is palpable. The biggest hurdle remains the lack of skilled staff and workforce, a challenge echoed by industry leaders like Gene Sun (FedEx), who states, “AI will not take your jobs, but people with AI skills will,” and James Beason (Pfizer), whose mantra is “either change the people or change the people.” Furthermore, nearly half of our respondents are already seeing reductions in skilled staff, with AI expected to bridge the productivity gap—a trend exemplified by Walmart’s projected 4% annual growth without adding headcount, driven by AI tools and new processes.

Engineering Miracles: Collaborative Projects

As S. Fial, CTO of Memorial Sloan Kettering Cancer Research, wisely put it: “Boards are pressuring CEOs to adopt AI, but without the right infrastructure and governance, you’ll only get false starts.” This summit is designed to equip you with that crucial infrastructure.

We’re already making strides through collaborative projects within the ONUG Community. One is the “Agentic AI Overlay Reference Architecture.” In an agent-driven world, where ephemeral agents communicate dynamically with databases, tools, and models across various trust domains, traditional static role-based authentication and authorization mechanisms are obsolete. Our working group is defining a zero-trust, interoperable framework to address dynamic security control, rogue agent isolation, governance, data exfiltration prevention, privilege escalation, and audit trails. Parantap Lahiri from eBay has been closely involved with this project and presented the project at the Summit. 

Another critical project, “ONUG Connect,” tackles the archaic and manual process of procuring networking circuits, which often represents a top 10% budget item. Eric Powers from Citigroup and Tony Farinacci (ex-JPMC) demonstrated an MVP of an API-driven reference architecture at the Summit. This MVP automates circuit discovery and selection, incorporating route diversity and AI-powered analysis to drastically reduce procurement time from months to seconds.

These are just two examples of our community’s work in 2026. Our framework centers on AI-native platforms, driving business value through two networking pillars: data center and AI WAN Edge, supported by agentic AI overlay, orchestration, observability, and pervasive security.

Enterprise IT are the infrastructure professionals that will enable the enterprise AI era. Together, we are building an AI-native enterprise network fabric that will engineer miracles for our companies and, in doing so, transform the world economy, one company at a time.