Industry executives and experts share their predictions for 2026. Read them in this 18th annual VMblog.com series exclusive.
By Rob Hirschfeld, CEO, RackN
Let’s cut through the noise. AI dominates every conversation heading into 2026, and it should. But here’s what nobody’s saying out loud: the bill comes due in 2026.
The Coming Token Economics Reality Check
AI companies can’t keep subsidizing token costs forever. What looks like “successful pilots” today is going to turn into CFO sticker shock when real production workloads start burning through tokens at scale.
And here’s what makes this worse: optimization strategies (RAGs, MCP, etc) do not create predictable outcomes. Worse, the infrastructure for AI pipelines is still immature, and enterprises are about to discover that in production.
We predict that economics will force companies to reconsider owning their own bare metal instead of paying per token. Not because it’s trendy, but because the math stops working at scale. Plus, at real scale, security, control, and data sovereignty aren’t optional; they’re requirements.
A Familiar Platform in New Clothes: VMware’s Gift to Kubernetes
The exact same dynamics are playing out on the virtualization side, driven by VMware license increases. Companies that spent twenty years standardizing on vSphere are suddenly doing the math on alternatives.
At RackN, we’re seeing this firsthand. Enterprise clients are actively evaluating bare metal Kubernetes for both virtualization consolidation and AI workloads. The conversations we’re having with Fortune 1000 companies today would have been unthinkable three years ago. As a result, we’ve been helping companies learn bare metal Kubernetes and OpenShift skills through targeted, operations-focused pilots.
So what do AI workloads and enterprise VMs have in common? The potential for Kubernetes to become the common control platform.
This strategic alignment is heating up. Virtualization teams running VMs with KubeVirt. AI ops running GPU workloads with Kubeflow. Platform teams focused on helping AppDevs. Different problems, different teams, and different clusters, but leveraging the same platform underneath.
That’s not a radical vision. That’s what survival looks like in a post-VMware, AI-driven infrastructure world. And it’s the main ask we’re getting from customers to build.
The Expertise Gap Nobody Talks About
But there are serious storms on the horizon. A decade of cloud abstraction killed bare metal expertise in most enterprises. The people who knew how to operate physical infrastructure at scale have retired, moved to cloud providers, or retrained up stack.
So we believe that 2026 is when enterprises rediscover, painfully, that rebuilding bare metal capabilities from scratch is an unrewarding place to spend mental cycles.
What does that look like? After 15+ years in infrastructure automation, I’m watching this pattern repeat:
- Budget stress as pilot costs balloon into production realities
- Failed deployments where infrastructure never quite works reliably
- Million-dollar GPU clusters sitting idle because operational processes aren’t there
- Engineering teams drowning in firmware updates and network configs instead of shipping features
But we’ve learned that it doesn’t have to be this way from dozens of deployments.
From FUD to Confidence for Infrastructure Operators
Working with enterprise clients on OpenShift bare metal deployments, we’re seeing the operational patterns that separate successful rollouts from stalled projects. The difference isn’t technical sophistication; it’s operational discipline and knowing where to focus scarce expertise.
We’ve seen that the winning approach starts with zero-touch provisioning. Then builds repeatability into the foundation. Getting the operational foundations resilient and automated takes some time upfront, but pays big dividends right away in terms of faster learning cycles, improved scale and operational experience.
You can’t spend 18 months reinventing bare metal automation when your real goal is running AI models or consolidating virtualization costs.
The companies that win in 2026 will rebuild operational muscle strategically and partner where it makes sense to accelerate delivery. While bare metal is a competitive advantage, leaders don’t need to become experts in firmware and hardware life-cycle to have their infrastructure running reliably. This is exactly what Digital Rebar delivers: it gives you infrastructure that just works so you can focus on problems that actually differentiate your business.
Looking Ahead to 2026
I’m both excited and nervous about the IT challenges that AI and virtualization are adding into an already demanding profession. I’m hopeful because I’ve witnessed remarkable operational improvements among RackN customers who adopt repeatable, well-managed infrastructure pipelines. The potential is real, but so are the operational challenges and expertise gaps.
What are you seeing in your infrastructure planning for 2026? Are you grappling with AI costs, VMware alternatives, or the bare metal skills gap? Let’s compare notes in the comments.
##
ABOUT THE AUTHOR
Rob Hirschfeld, Founder & CEO
Rob has been in the cloud and infrastructure space for 20 years and has done everything from start-ups working with early ESX betas to serving four terms on the OpenStack Foundation Board and as an executive at Dell. As leader on the2030.cloud, he believes that the technology of running data centers and applications on cloud is just part of the bigger story. He trained as an Industrial Engineer and carries a passion for applying Lean and Agile processes to software delivery. Rob has received degrees from Duke University and Louisiana State University.
