Guest: Rob Hirschfeld (LinkedIn)
Company: RackN
Show: 2026 Predictions
Topic: AI Infrastructure

As enterprises rush to build AI infrastructure in 2026, they’re heading into a perfect storm of challenges: insufficient power, unclear designs, and pressure to deploy before foundations are solid. Rob Hirschfeld, CEO and Co-Founder of RackN,  warns that the year ahead will separate companies with sustainable AI strategies from those destined for expensive false starts—and the difference comes down to one thing: process over speed.

The AI Infrastructure Challenge: Gold Rush Mentality Meets Reality

RackN provides full bare metal automation platforms for enterprises running their own infrastructure, managing everything from initial deployment through the complete lifecycle until decommissioning. As companies face mounting pressure to deliver AI capabilities in 2026, Hirschfeld sees a troubling pattern emerging.

“2026 is clearly the year when we start seeing the demands of AI translate into real production environments. But that also means real production challenges,” Hirschfeld explains. “Companies are going to have to start figuring out what to buy, and they’re going to have to start figuring out how to run those platforms.”

The gap between ambition and execution is widening. “Running an AI workload on your own is much, much more complex than VMware,” he notes. “But outsourcing it to a SaaS provider—where you pay by the token—means those costs are really going to start adding up.” As a result, enterprises find themselves caught between expensive token-based models and the complexity of building their own AI infrastructure.

The market itself is shifting dramatically. “There are new entrants emerging around NVIDIA power requirements and the new cycles of technology people are building, alongside broader power restrictions in data centers,” Hirschfeld observes. “We’re entering a phase where organizations are compelled to build AI infrastructure, yet the infrastructure they need to build is something they don’t have the power for, the designs for, or even a clear understanding of.”

VMware Exit: The 18-Month Reality Check

Despite the widespread desire to leave VMware following pricing changes, Hirschfeld predicts that most enterprises will renew their contracts in 2026. “The reality we’ve seen over and over again is that, as much as companies are getting pinched by VMware bills, the alternatives—and the ease and time required to implement them—are not practical for most organizations.”

However, this isn’t surrender; it’s strategic recalibration. “I do think companies are going to realize that they need real plans to get off VMware,” Hirschfeld says. “They’re going to have to start a serious 18-month to two-year cycle to build something better than VMware on their own, because that requires new hardware investments and a true exit strategy.”

This marks a shift from panic to pragmatism: “It’s going to be a much more practical VMware exit plan—still with high urgency, but a lot of resignation on how it will actually be executed.”

Agentic AI: Separating Promise from Hype

On agentic operations, Hirschfeld takes a measured stance that acknowledges both potential and limitations. “We do see agentic operations and using agents as actual, true material benefits for the coming year. And they are so incredibly hyped, it’s really difficult to be optimistic about agentic capabilities in the middle of all the hype.”

The key is understanding what agents actually are: “I see them as adjacencies to people’s infrastructure,” not standalone commercial products. “We expect our customers to actually want to run their own agents as agents are going to have to be trusted. They’re going to have to be paid for. They’re going to functionally be AI employees.”

Rather than replacing workers, agents should augment capabilities. “I think agents are going to make the employees you have and the situational analysis that they can do even more important,” Hirschfeld emphasizes. “Finding that balance is going to be the job for 2026.”

The Foundation-First Philosophy

Hirschfeld’s core message cuts through the noise with stark simplicity: “AI layered on top of bad processes will accelerate technical debt, and process debt will accelerate mistakes.”

This isn’t about slowing down—it’s about building sustainability. “If you want to move faster with AI, or with any type of infrastructure, you need to make sure the underlying foundational processes are solid,” he explains.

The danger lies in a gold-rush mentality. “The people who are successful aren’t necessarily the smartest in the room; a lot of times, they just got lucky.” Sustainable success requires a different mindset. “The ultimate winners are the ones who build and maintain solid foundations—who invest in strong processes and use them to accelerate.”

New Technologies Emerging

Despite the challenges, Hirschfeld sees genuine excitement in emerging technologies. “We’re seeing Ethernet technology begin to displace InfiniBand, a more diverse chip mix beyond just AMD and NVIDIA—including ARM architectures—and new network topologies enabling truly interesting platform designs.”

This shift represents opportunity on a generational scale. “We haven’t seen an opening like this in almost two decades. VMware’s long-standing lock on data center technology limited how we thought about infrastructure,” he says. “Now we can seriously consider true multi-vendor environments and genuinely innovative, high-performance systems in ways the IT industry hasn’t explored for an entire generation.”

Practical Guidance for Enterprise Leaders

For organizations navigating the challenges of 2026, Hirschfeld’s advice centers on process excellence. “It’s not just about the tools—it’s about the underlying processes, capabilities, and the quality and consistency you’ve built within your organization,” he stresses.

When it comes to AI tools themselves, the focus should be strategic. “We need to focus on storytelling and make sure we’re telling compelling stories with a strong human element.” AI, he argues, should free up time for higher-value work rather than simply generate more content. “When these tools help us improve our storytelling and human connection—and give us more time to do that—I get very excited.”

The warning is clear: enterprises that pursue AI infrastructure projects without solid foundations should expect “a lot of AI false starts” and “a lot of very expensive mistakes.” The path forward requires patience, process discipline, and realistic timelines—uncomfortable advice in a gold rush, but essential for sustainable success.