Enterprise AI is moving in data centers and the ongoing hardware crisis has thrown every CIO a monkey wrench.
Memory prices have surged over the past year, server lead times now stretch for months in some cases, and IT teams across industries are being forced to do more with less.
Global DRAM and NAND flash prices rose 200% to 400% in 2026 as memory manufacturers redirected supply toward high-margin AI infrastructure, a crisis that Kearney PERLab stated in its The Great Memory Reallocation report could persist until 2030. That convergence of accelerating AI adoption and a deepening component shortage represents a genuinely new kind of pressure for enterprise IT, according to Steve McDowell, chief analyst and founder of NAND Research, a boutique analyst firm focused on enterprise infrastructure.
"I don't know that in my history in the industry, we've had a moment quite like this," McDowell said during an interview at Nutanix .NEXT 2026 in Chicago.
"You have both AI as a disruptor and you have supply constraints. This is going to last through the end of 2027."
A Supply Crunch Unlike Any Before
The numbers underscore that warning. Industry reports show that the global shortage of DRAM and NAND flash memory, driven largely by a structural reallocation of manufacturing capacity toward high-margin AI infrastructure, could persist until at least 2030. In his May 2026 report Memory & NAND Flash Crisis: May 2026 Update, McDowell saw DRAM contract prices increase 90-95% between the fourth quarter of 2025 and the first quarter of 2026. New catch phrases like "RAMmageddon" or “Memflation” are bandied about tech circles to describe the financial impact on CIOs and IT decision makers.
The effects reach well past the server rack, according to McDowell. He noted that the latest MacBook laptops are carrying five-month lead times due to memory constraints, and some vendors are issuing price quotes valid for only 10 days because the market is shifting that fast.
The root cause is economically rational, even if the consequences are disruptive. Only a handful of factories in the world are capable of producing memory and NAND flash, and those manufacturers are directing output toward higher-margin products that serve AI clusters.
"The memory companies are doing what I would do if I owned a memory company," McDowell said. "I'm going to focus on these high-value AI clusters."
The unintended consequence is that mainstream enterprise memory has grown scarcer and more expensive at precisely the moment IT demand is accelerating.
The AI Deployment Era Has Shifted
That hardware squeeze is landing at the same time as enterprises are confronting a more fundamental shift in how AI itself is deployed. The first era of generative AI centered on training, the kind of compute-intensive work that consumed enormous resources at hyperscalers and AI labs.
The current era is about inference: the practical application of trained models to real business processes. And that work, McDowell said, is landing squarely at the feet of IT practitioners who had no hand in building those models. From inference, the technology is evolving rapidly into agentic AI: networks of specialized models working in coordination across tasks.
"All an agent is simply is a small version of an AI someplace," McDowell said, who has over 25 years of experience in the IT industry. "I have agents talking to agents. I have AI talking to AI. If I'm a coder, maybe I have an agent that's writing code. I have another agent that's testing it. And I have another agent that's validating the security. It's a force multiplier for AI."
The challenge for IT practitioners, he said, is identifying the right high-value applications rather than simply reaching for the nearest bright technical object.
The Jobs AI Is Displacing May Surprise You
The workforce implications of this force multiplier carry an unexpected irony. Much of the anxiety around AI focuses on non-technology roles, but McDowell said the jobs facing the most visible displacement right now are in technology itself.
"The jobs that it's displacing right now are the coders," he said. "I mean, AI is doing the technology jobs."
He framed that observation not as a reason for despair, but as a prompt to adapt: AI changes how engineers work, not simply whether they work.
A Lesson From the Broadcom-VMware Playbook
For IT leaders sorting through various large language models, McDowell advised IT teams to evaluate AI providers the same way they evaluate any enterprise technology vendor: “On economics, trust and fit for purpose,” he said.
Emerging standards are reducing switching costs: Model Context Protocol (MCP) for agentic interfaces and OpenTelemetry for observability are both gaining traction, McDowell said. This means that code written for one AI platform will increasingly run on another. That need for interoperability across IT infrastructures matters because infrastructures and vendors change and those changes can impact business decisions.
"We saw with Broadcom and VMware, and we're still feeling those ripple effects," he said, referring to the impact of changes that impacted customers after Broadcom acquired VMware. "You don't want that to happen again."
He said Broadcom bought VMware to treat the software platform as an annuity, not to build the next generation of leading-edge technology. New application development on VMware, he said, has effectively come to a halt.
“VMware, to me, it's kind of a legacy workload management system that's sitting in the infrastructure,” he said.
“It's going to be there probably for as long as you and I are in the business, because IT guys, you know: if it ain't broke, don't fix it. Until the switching costs and the paying costs cross, they're just going to run it.”
Quantum Computing's Inflection Point Is Closer Than IT Thinks
On the longer horizon, McDowell sees a different kind of inflection point: the arrival of quantum computing.
He’s spending significant time now tracking quantum computing, which he sees not as an evolution of AI but as an independent trajectory aimed at solving problems that classical compute architectures simply cannot. Commercial adoption is already beginning in oil and gas and drug research. If the analogy holds, the surprise may arrive faster than most IT teams expect.
"That's where AI was in 2019," McDowell said. "Quantum's going to go through that same kind of evolution."
The aha moment, he said, is coming.
Related:
AI's Defining Decade: Nutanix Leaders on Inference, Agents, and the Road Ahead
Nutanix CEO Sees Multi-Year Opportunity After Broadcom's VMware Acquisition
2025 Enterprise Cloud Index Shows What's Driving AI and Cloud Native Tech Trends
Ken Kaplan is Editor in Chief for The Forecast by Nutanix. Find him on X @kenekaplan.
Jason Lopez contributed to this story. He is executive producer of Tech Barometer, the podcast outlet for The Forecast. He’s the founder of Connected Social Media. Previously, he was executive producer at PodTech and a reporter at NPR.
© 2026 Nutanix, Inc. All rights reserved. Nutanix, Inc. is not affiliated with VMware by Broadcom or Broadcom. For additional information and important legal disclaimers, please go here.
Related Articles
