The most exciting period in technology since the dot-com era comes with a catch. Artificial intelligence is reshaping every layer of IT, from coding to operations, and the ripple effects are reaching into the physical world of silicon and supply chains.
"This is probably the most exciting time in technology since the dot-com era, but in good and bad ways," said Steve McDowell, founder and chief analyst at NAND Research, in a video interview recorded at the April 2026 Nutanix .NEXT event in Chicago.
"You can't talk about technology without talking about AI, whether that's the impact of AI on IT operations or coding or just societal," he told The Forecast.
AI is simultaneously a disruptor and an accelerator, creating scarcity in memory and storage while pushing IT teams to manage workloads across an increasingly complicated landscape of cloud, on-premises and neo-cloud environments. The result is a fundamental shift in how practitioners provision, allocate and manage computing resources.
AI infrastructure encompasses the compute, memory, storage and networking resources required to build and run artificial intelligence workloads, spanning on-premises data centers, public clouds and specialized AI cloud providers.
The supply chain squeeze is already measurable. According to Gartner, AI demand intensified the NAND flash shortage to 9% in 2026, with supply constraints expected to persist into 2027. The research firm projects a 293% price surge in 2026 followed by a further 30% increase in 2027, restructuring the traditional memory cycle through 2030. TrendForce reports that DRAM remains "very tight" at 20 weeks versus an 8-week balanced level, with hyperscaler AI infrastructure investment growing at 35% to 60% annually since 2022, far outpacing NAND capacity expansion of 8% to 15%.
The downside of AI is its appetite for physical resources. Memory and NAND flash share similar manufacturing processes and supply chains, and only a handful of factories worldwide can produce them. Memory manufacturers are doing what any rational business would: prioritizing high-bandwidth memory and DDR5 for AI clusters because those products command higher margins.
The unintended consequence is that mainstream memory and storage are harder to come by. Prices are climbing. Availability is shrinking. McDowell said he saw a report that the latest MacBook carried a five-month lead time because of memory constraints. Vendors are issuing price quotes valid for only 10 days as the memory market shifts beneath them.
"Things are more expensive, but there's also scarcity," McDowell said. "Which means as an IT practitioner, I have to do more with less."
The constraints are forcing a rethink. Three years ago, an IT team might buy a server with a terabyte of memory. That is impractical now. Instead, practitioners are looking for software solutions that let them separate workloads from hardware, placing them where resources are available, whether on-premises, in the cloud or at the edge. McDowell compared the moment to nothing he has seen in his career, with AI as a disruptor and supply constraints hitting at the same time.
"I don't know that in my history in the industry we've had a moment quite like this," McDowell said. "You have both AI as a disruptor and you have supply constraints. We see supply constraints in bubbles. This is going to last through the end of 2027 and maybe beyond."
AI agents are multiplying the impact of AI across IT environments. An agent is a small, specialized instance of AI that can operate autonomously. A coder might use one agent to write code, another to test it and a third to validate security. Agents talk to agents. AI talks to AI.
"It's a multiplier. It's a force multiplier for AI," McDowell said. "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."
The technology is powerful, but the question for IT practitioners is where the high-value plays are. The business of IT is to enable the business, not to play with technology. McDowell said practitioners are sorting through which agents, which providers and what cost structures make sense.
"My business as an IT practitioner is to enable the business, not play with technology," McDowell said. "Now technology is how I get there, but what's the right set of agents? What are the right providers? What's the cost look like?"
AI is also driving expansion of the edge, adding another layer of infrastructure to manage. The management plane, the software layer that orchestrates resources across distributed environments, is becoming critical in a way it never was before. In the days of pure virtualization, the goal was getting the most out of a single box. Now the challenge is getting the most out of a broadly distributed infrastructure spanning on-premises, cloud and neo-cloud providers like CoreWeave.
Fear is part of the picture. There has been widespread worry about AI displacing jobs, and McDowell said the irony is that the displacement is hitting technologists first. AI is doing the coding jobs, the technology jobs.
"I think they're scared," McDowell said of IT practitioners. "One of the great ironies is there's been a lot of worry and speculation around is AI going to put me out of a job? And the irony is the jobs that it's displacing right now are the coders. AI is doing the technology jobs, that's where we're seeing displacement more than others."
The path forward, McDowell said, is to treat AI as a tool, an enabler and an accelerator. It is not that organizations need fewer engineers. It changes the way they operate. Pushing out of a comfort zone is hard, but growth follows.
"You have to look at AI as a tool. It's an enabler. It's an accelerator. It helps you do more with less," McDowell said. "It's not that I need less engineers, but it changes the way they operate. And as humans, we like to live in a comfort zone and it's very hard, but once you push out of that comfort zone and grow, if you will, it becomes more natural."
There is a temptation, McDowell said, for technologists to chase the bright, shiny object. But the purpose of solving hard problems is to help somebody else do something that is not technology related.
"It is so easy as a technologist to get wrapped up in the bright, shiny object, the sexy, I want to go solve these hard problems," McDowell said. "But at the end of the day, why are we solving these problems? It's to help somebody else do something that's not technology related."
The supply constraints are not a bubble. McDowell said they will last through the end of 2027 and possibly beyond, potentially establishing a new baseline for memory prices. AI itself is not a bubble either, he said. It is a fundamental shift.
"It's a new normal, right? I mean, AI is not a bubble," McDowell said. "It is a fundamental shift. It is a fundamental shift. Absolutely."
Video transcript:
Steve McDowell: This is probably the most exciting time in technology since the dot-com era, but in good and bad ways. I mean, the things we're looking at right now is certainly you can't talk about technology without talking about AI, whether that's the impact of AI on IT operations or coding or just societal. I mean, it's a big topic and it's impacting everything that I cover. That's causing some rippling effects. There's supply chain problems with memory and flash memory. So there's a scarcity that's impacting how IT operates. That's the bad side. And then there's just the usual, how do I manage my workloads across cloud, on-prem, and with AI, these neo clouds? So IT infrastructure is getting very complicated and you really have to step back and think about what's the best way to tackle these problems if I'm an IT guy. And this comes after years, I don't want to say it's been stagnant in IT world, but kind of natural evolution. And now we're in kind of a revolution period.
Ken Kaplan: How do you perceive agents and what that's doing to the mentality of the IT folks who have to be ready and manage these things?
Steve McDowell: Sure. So it's all an extension of how do I use AI? And all an agent is simply is a small version of an AI someplace. So 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. So it's a multiplier. It's a force multiplier for AI. And I think where we are right now is there's so much that we can do with that technology. What are the high value plays? Certainly as an IT guy, because my business as an IT practitioner is to enable the business, not play with technology. Now technology is how I get there, but what's the right set of agents? What are the right providers? What's the cost look like? There's so many things, but a genic AI, that's the accelerator for how I'm going to use this.
Ken Kaplan: You get into the mind of the IT guys pretty well and you often speak for them when we talk. Just get a sense for how you think they feel and what they're seeing.
Steve McDowell: I think they're scared. One of the great ironies is there's been a lot of worry and speculation around is AI going to put me out of a job? And the irony is the jobs that it's displacing right now are the coders. I mean, AI is doing the technology jobs, that's where we're seeing displacement more than others. So if an AI can do my job, so there's a lot of fear and some's founded, some is not. I think you have to look at AI as a tool. It's an enabler. It's an accelerator. It helps you do more with less. It's not that I need less engineers, but it changes the way they operate. And as humans, we like to live in a comfort zone and it's very hard, but once you push out of that comfort zone and grow, if you will, it becomes more natural. But no, I mean you have to put yourself in the mind of the IT guy because the other side of me, my whole life I've worked for tech companies. It is so easy as a technologist to get wrapped up in the bright, shiny object, the sexy, I want to go solve these hard problems. But at the end of the day, why are we solving these problems? It's to help somebody else do something that's not technology related.
Ken Kaplan: Let's talk about the supply chain constraints and what's happening there. Have you seen anything like it before?
Steve McDowell: No. So I alluded earlier. The downside of AI is it's consuming a lot of physical resources. It's very memory intensive and there's only a handful of factories in the world that can build memory. And with memory, I put NAND flash, so flash storage and memory is very similar manufacturing processes and supply chain. And the memory companies are doing what I would do if I owned a memory company. I'm going to focus on these high value AI clusters, this high bandwidth memory DDR5 because it's high ASP, high margin. So they're doing their normal business. I am not saying a bad word about the memory guys. They're doing what I would do. But kind of the unintended consequence of that is mainstream memory, mainstream storage is harder to come by. It's driving up prices, it's driving up availability. And I think I saw yesterday that the latest MacBook, it's got a five month lead time because of memory. Vendors are given price quotes that are only valid for 10 days because the memory market is shifting. So what that does to the IT guy, yes, things are more expensive, but there's also scarcity, which means as an IT practitioner, I have to do more with less. And how do I do more with less? And then I start looking at solutions that allow me to separate my workload from the hardware. We used to call that virtualization, but the containers, virtualization, number of techniques, but it's all about managing my workloads now to take advantage of my existing infrastructure. Whereas even three years ago, I might buy a server with a terabyte of memory, that's impractical now. So now I'm buying a server that's not as equipped or maybe I'm not buying that server. So it requires me to think differently about how I provision those servers. So maybe I have an old server and that's a perfect place to go put my mail or my SharePoint. So it forces me to think differently about allocating resources and that's new. That's new for IT. I don't know that in my history in the industry we've had a moment quite like this with you have both AI as a disruptor and you have supply constraints. We see supply constraints in bubbles. This is going to last through the end of 2027 and maybe beyond. I think maybe we might even be establishing a new bar for normal for memory prices. So how do I deal with that as an IT guy? I'm going to go look for software solutions, allow me to manage that workload both on-prem so I can put the workload where it belongs, but also offload the cloud. And there's not a lot of solutions that allow me to bridge on-prem to cloud. And with AI, even Neo-Cloud, so I might have something on CoreWeave and something on Google Cloud and then the core of my data center. And the other thing we haven't touched on is AI is also driving expansion of edge. So I have this whole new infrastructure expansion and how do I manage that? So that management plane is becoming critical in a way that it never was before. In the days of pure virtualization, I'm just trying to get the most out of a box. Now I'm trying to get the most out of a very broadly distributed infrastructure and that's a whole different skill set. And that's where Nutanix I think excels and where you're finding great success.
Ken Kaplan: It sounds like we've arrived at this place where you're able to store some data here or over here and have this system grab it from there or there. It does feel like there's this complicated web that maybe wasn't functioning that way a few years ago, but that's where it is now and it will probably continue like that.
Steve McDowell: Oh yeah. It's a new normal, right? I mean, AI is not a bubble. That comes up a lot in these kind of conversations. Now, certain segments of AI may be, right? I mean, how much training capacity do we really need or whatever? And I think where we are right now and we're trying to figure out is where does inference live? There's going to be many answers to that, depends on what I'm trying to do. But it is a fundamental shift. It is a fundamental shift. Absolutely.
Jason Lopez 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.
Ken Kaplan contributed to this story. He is Editor in Chief for The Forecast by Nutanix. Find him on X @kenekaplan and LinkedIn.