Tech Barometer – From The Forecast by Nutanix

Healthcare IT veteran Benjamin Urquhart says AI can transform patient care, but only if it fits clinician workflows and tames shadow AI before it scales.

Get tech leader insights to move faster and smarter.

Get more stories by subscribing to The Forecast.

Video transcript:

Benjamin Urquhart: AI, if done and implemented correctly, and I believe the healthcare environment, it has to be done thoughtfully. It can be amazing. It can be. I don’t want to get on the hype train and go, “It’s going to solve everything,” because we’ve had those solutions in the past. It’s going to solve everything and it never does. It might come close, but it’s going to miss some pieces that we just don’t realize. And AI has rapidly evolved and it’s been adopted at such a rate that in the 18 years I’ve been in the industry, I don’t think I’ve seen anything like it. In a sense, it’s scary. Because of that in healthcare, all the regulational pieces we have and understanding how our client data is kept, where it’s kept, what can access it, east-west traffic right inside with an AI agent can be extremely scary, very vulnerable in that aspect.

And so when we’re looking at it from AI and healthcare, in my mind, the biggest thing becomes doing it thoughtfully and then doing it in a way that works within the end user’s workflow already. We give them all these things all the time. We tool them to death, right? I’m sure you’ve heard that phrase. And tools and solutions are not always the way to do it. You can look at their best thing about AI and healthcare, I think we can get rid of a lot of the administrative burden. We can utilize it within their workflows to remove some charting capabilities, do the reporting, take care of these things, but then at the same time, also use it to help analyze that data that’s coming into these EMRs and everything like that to where, because it is complex data. Health data is much more complex than I think everyone doesn’t always understand.

But we can use AI to better analyze that, giving us better outcomes at the end of the day for our patient care. Because if we’re using it in a thoughtful manner to go, we’re going to remove some administrative burden and we’re also going to utilize AI to help assist you diagnose a client or work through in determination, well, that provider themselves now has a better understanding of what’s going on. They’re able to spend more time with the client, which gives us the outcome we want and that is helping people. That’s the way I view AI in our world right now in healthcare. It’s got to be done in their workflow. It has to be done very thoughfully.

Ken Kaplan: With the foundation that you’ve built and are growing, you could take on new capabilities.

Benjamin Urquhart: Yeah. Anytime companies grow or you have to evolve, it’s always kind of a mental check aspect and you kind of go, “Are we really going all the way in? Are we only going halfway in?” But being afraid of the technology as it’s evolved, if you are, we might want to look elsewhere. And I don’t mean to be mean or so blunt about that, but because tech, my dad, I’ve grown up around tech. My dad was a mainframe programmer in the banking industry when I was a kid. Going to the Saster test at midnight when I was like seven or eight, no idea what’s going on.

And now it’s great. The tech has evolved my entire life and everything. One of the things though too, people have evolved and our end users today, they grew up with smartphones. We maybe didn’t, but they did. And so they, for an end user and a lot of this younger generation I feel that are coming in, we need to change maybe how we communicate with them a little bit. It’s not just a technical solution to them. We need to prove that in value to their life because that’s the tech that they’ve had their whole life. It added a value. Our legacy work environment does not really add a value. We add burden, just the way that systems are designed. So adjusting that and making them work within that is an evolutional piece I think from just being in tech in general. We just have to kind of adapt how we’re explaining, how we’re talking and communicating more so than just about anything else.

Ken Kaplan: How comfortable are you and your team to say, “This is the year that we’re going to do these three things with AI.” How do you feel you’re able to move into the future?

Benjamin Urquhart: I feel cautious about it. It’s one of the things that I know we’re going to have to do. Every company is doing it. We have shadow AI now. They were talking about that, one of the keynotes today. And in the past we had shadow IT where people were just. Now we have agents where they’re going out and they’re, “Oh, hey, because we didn’t have something maybe locked down.” Well, now there’s this agent that no one knows about. So I think when we look at having to do AI, we’re going to have to do it. I think it’s the way of the world. I think if, like we said earlier, thoughtfully done, it can make massive improvements. But at the same time, we need to ensure that’s structured correctly for being in your industry, whether it’s healthcare, manufacturing, whatever, you structure it right before you just say, “Hey, we’re doing all of this.” Standardization.

You might have someone who goes, “I really like Claude. I’ve got someone over here, ChatGPT, someone who says they don’t like Copilot, but maybe Copilot’s part of your 365 ecosystem. Why would you come over here unless there’s an obvious reason to do so?” So right now it’s really more around trying to figure out in a strategic aspect, look at where we are, look at what fits our ecosphere within our organization and making those decisions on how to move forward and where to implement first.

Related:

Lean IT Team Scales AI at Alabama Healthcare Provider

Survey Shows Healthcare AI is Reaching Bedsides but IT Systems Need a Boost

Getting Beyond AI’s High Anxiety Period

How AI Is Revolutionizing Healthcare Tech

Healthcare Data Storage Trends and Challenges

In a video interview, Five Horizons CTO Benjamin Urquhart explains how hyperconverged infrastructure let a nonprofit healthcare provider scale from zero servers to over 50 virtual machines.

Get tech leader insights to move faster and smarter.

Get more stories by subscribing to The Forecast.

Video transcript:

Benjamin Urquhart: Nutanix for us was honestly a little bit of an easy call. Organizationally, we had no infrastructure in place as far as server backends, everything like that. Being in the nonprofit sector where we are, we’re a specialty healthcare provider, so we have a lot of grant limitations on how we can spend our resources. Because of that, we can’t really go, let’s go throw them in the cloud, let’s do this, let’s do that. We had to really look at it from a strategical aspect and go, well, what’s my ROI coming back in on this? What type of solutions are we trying to do that’s still going to allow us to be scalable and grow? We’ve opened up four new facilities in the last 18 months between Mississippi and Alabama right now. So being able to have something in house with Nutanix for us was great. It was HCI platform, something I’d never done, but let’s go.

It gives us our hardware, it gives us a hypervisor. It’s different, but let’s give it a shot. Licensing became much more easier or simplified. There we go. Nutanix allowed us to go from really not having an internal infrastructure to running over 53 VMs right now internally from different internal resources, whether they’re part of our research environment or our true day-to-day activity. So it’s been a great marriage from that aspect, I would say, like partnership. It’s allowed us to grow and be scalable, but also secure and understanding where our data is and how it lives within our ecosystem.

Ken Kaplan: From that time that you’ve got things going to the 57 VMs out there, what was that time span and how did things change?

Benjamin Urquhart: So it was a very unique instance actually. I was working for another organization very similar to five Horizons Self-Services and that particular organization was failing. Five Horizons and their board and our CEO, they came in. We had over 3000 clients in our coverage area for that particular organization. They bought all of our assets, everything like that. We retained certain staff members and we didn’t close the doors once in the fall of 2022. So from that angle, it was great because we have this infrastructure here that we had already started building with this company. Now we’re going to Five Horizons because they’ve acquired our assets and being able to keep the doors open for patient care.

It was almost like having where the starting line was on your infrastructure instead of completely starting over again like I had the three years before that. Oh hey guys, we’re halfway there. We have the cores. We’re changing domains. We were changing simplified things. So for us it’s just allowed that ability. The biggest push because we already had it was now we’re at the point where we’re looking at additional blocks and clusters because we continue to expand. Not necessarily a bad thing, but it allows us that ability to keep everything in a manner that we know we can do. We have a very lean team that really have to support not just our Nutanix environment, but our backup environment and everything else on top of it. And Nutanix has really simplified that. I mean they can spin up a new VM now in about 15 minutes from them, which is fantastic. There’s a lot less time and delay because of how HP operates.

Ken Kaplan: Now you’re like a veteran almost with HCI and what have you learned and what do you want to try next?

Benjamin Urquhart: So some of those things are coming from here right now this week and some of that is kind of going more into the VDI world maybe on top of the Nutanix platform running into HV, doing some more virtualizational pieces around our security and compliance that would help dealing with the industry we’re in. Looking at things like that, the containerization aspect, allowing us to spin up more things from an application aspect and having a better use case of our hardware equipment or allocation of whether we’re talking storage or our RAM compute, allowing us to trim line those into things like that is really, I guess evolutionary. It’s like we build an infrastructure to get our foundation and make sure it’s scalable. We build for today and tomorrow where we’re going, make sure it’s a foundational piece. Well now you also have to look at it and go, I have this foundation.

I need to kind of break it up though and let’s evolve it. Let’s allow that foundation to evolve and now we’re just building buildings out of our infrastructure stacks.

Related: 

In this video interview, Liqid Founder and CTO Sumit Puri argues the path to affordable inference runs through composable AI infrastructure that dynamically taps into pools of GPUs, CPUs and DRAM while optimizing usage to manage costs. He says Liqid allows a single server to access a pool of 30 AMD GPUs to meet performance-hungy AI needs.

Get tech leader insights to move faster and smarter.

Get more stories by subscribing to The Forecast.

Video transcript:

Sumit Puri, Founder & CTO, Liqid: Since we spoke to you guys last, one of the major trends that we’re seeing is the enterprises are finally past the initial exploration phase. Now they’re in the adoption and deployment phase of their journey of AI. And I think a lot of them are trying to figure out how exactly they are going to do this journey. And the first crossroads that they’re up against is, am I going to take all of my AI capability and move it into the cloud? Or am I going to own the infrastructure on-prem or in a colo that’s required to do AI? And what it all comes down to is tokenomics. At the end of the day, these companies have a limited amount of dollars and a limited amount of power in many cases that they can deploy. And what they’re trying to figure out is how do I get the most amount of tokens out of that very, very scarce resource?

And so those are the discussions that customers are having now. And the cost of these tokens is front in mind for them. If we notice what’s happening in the industry, large organizations are now saying, “Hey, we’re going to limit the amount of tokens that you as an engineer inside of our organization can consume because the cost of that is becoming very high.” And so the one great way to address that increasing cost is to own your own token machine. And so that’s one of the conversations that we are having with our customers is one way to reduce the cost of those tokens is bring the infrastructure, bring the GPUs on-prem or in a colo so it’s a one-time expense and you can consume all the tokens that you want out of that investment that you make. It is about the model that you are looking to deploy and you’re going to, in any environment, you’re going to have a variety of models.

[Related: Rise of AI Agents Forges IT Industry Partnerships]

You’re going to have big models, you’re going to have small models, you’re going to have medium sized models. The model actually, the size of the model will dictate the type of infrastructure that you need. If I’m running a very, very large model, I’m going to need a large quantity of GPUs in order to run that model. And so you have to figure out if I’m going to build a very big system, how do I get those large quantity of GPUs into play? One way to do it is independent scaling of resources. We come in and let customers say, “I don’t want to scale my compute as I’m scaling my GPUs. Allow me to just scale GPUs.” As an example, today we’re at the AMD show. The announcement that we’re making today is our ability to take a single server and scale up to 30 AMD GPUs to that single machine.

And the benefit of that is we can now start to run these very large frontier models on a single server infrastructure. These models are so big. Previously you had to run them in a cluster. Now we come in and say we can take those models that were previously clustered and consolidate them down to a single server, a single instance, back to that tokenomic story. That’s how we reduce the cost of deploying these models. Models will change over time. And so some cases you might want a very small model that will require, let’s say, a single GPU. Sometimes you will want a medium sized model that will require eight. And sometimes you will require a massively large frontier model, which will require dozens. You as a customer, it’s impossible for you to know ahead of time what size model I will need when. And so our vision is let’s take the guesswork out of that.

Let’s have a pool of servers, a pool of GPUs, and dynamically spin up GPU quantity to server based upon models. And the reality is the way that’ll be done is something called Kubernetes orchestration will be the way that they do it. Nutanix has a phenomenal way of going off and doing that through their NKP platform. We’re a big believer in NKP. And so we integrate closely with NKP. And so we say, let’s allow people to deploy these models, the variety of models that are supported with the perfect set of infrastructure so we can have this matching of infrastructure to model to not have any wasted resources, improve the tokenomics. In a resource constrained environment, utilization is critical. You cannot afford to have your GPUs at 30, 40% utilization because you can’t get enough. And so if we can implement technologies that drive utilization to 100%, we can do more with the same amount of infrastructure.

[Related: Demand Shifts to CPUs to Power Agentic AI]

So that becomes critically important and we do a lot of that today. We look at an environment and we say, “Hey, listen, this workload is only using those GPUs 30% of the time. Let’s set up policies that allow you to move those GPUs to different parts of the infrastructure to raise the utilization.” The second thing is Brownfield. We support brownfield environments. So imagine an environment where I can’t buy servers because memory’s limited, CPUs are limited, servers are very expensive. We come in with our solution, which is disaggregated pools of GPUs, and we attach them to existing infrastructure. And so now we’re giving new life, new capability to the infrastructure that customers already own. As inference becomes a thing in places like the enterprise, it will only be done in containerized environments. NVIDIA has created something called NIMS, NVIDIA Inference Microservices. AMD has created something called AIMS, AMD Inference Microservices.

And the reason is both of these companies realize the complexity related to kernel revisions, operating system revisions, library revisions is too complex for average enterprise customers to debug and solve, and it slows the rate of AI deployment. So if we can containerize all of these models, put them into a store, allow the enterprise customer to bring whatever model he wants, a vision model, a speech model, a text model, bring it down in whatever size of model that they need, 7B, 70B, 405B, and make all of these available containers that the customer can deploy and implement immediately, that’s how we’re going to accelerate the pace of deploying AI. The way that we approach that is we have a plugin for Kubernetes that when you say, “Hey, give me Llama 7B as an example,” we’ll take the container, we’ll bring it down, we’ll crack open the container, we’ll determine the exact amount of physical resources that that container needs.

Well, this needs two GPUs and we want to put it on server number four. We’ll pause that container, we’ll go into the background infrastructure, we’ll put two GPUs on server number four, and then we’ll take that container and we’ll launch it onto that server. We automate the entire process of deploying models down to a two minute instance. So you say, “Give me Llama 7B.” Two minutes later, you’re speaking to a chatbot with the exact amount of hardware resources required for that instance. And the moment that that server doesn’t need that GPU anymore, we can delete that instance, remove those GPUs, put them back into a free pool so your next model that comes along always has a pool of GPUs to draw from. Liquid started, the way we started was around GPU pooling. That was the first thing that we went off and we accelerated and we composed and we pooled and we shared.

Now the next thing that we have just announced is memory pooling. So we are the first company to provide an end-to-end solution around DRAM pooling built around CXL. So now I can have my pools of GPUs and now I can have my pool of DRAM. The two most expensive things in the data center can now be dynamically allocated by workload. The days of over-provisioning your server with excess memory, those days are done now. And so we must find methods to precisely provision the exact amount of memory the application needs so we can reduce our memory costs inside the data center. And by doing that, it’s back to that tokenomics. Yeah. Well, the company is surely in its growth phase right now. I’m a technologist at heart. I’m super lucky to have built out an amazing team. We’re very lucky to have a new CEO on the team.

A gentleman named Rick Hagberg, who’s just an industry veteran who’s been phenomenal. We’re raising additional capital. We’re growing the team. We’re bringing new executives on board. We are having our moment right now and we’re super excited and blessed to be part of this industry. The way I feel about it is when we go around and we tell customers about our vision, which is stop putting static resources inside of a server. Let’s take all of the GPUs, pull them inside of the rack, add a push of a button, dynamically allocate any GPU to any server that you want. You won’t lose any performance. You’ll scale up as big as you want. We’re a driverless solution. We can drive to 100% utilization. When we tell this story, no one ever says, “Well, that’s a terrible idea.” That’s not the reaction that we get, right? The reaction we get is, “Wow, this is incredible. This is obvious. Obvious is the one that we get more than anything else. This is obviously the right answer of how this should be done. Why isn’t everybody doing it this way?” And me, that’s my mission as the founder here is to make sure eventually everybody is doing it this way.

Financial services firms are pouring money into AI but face soaring token costs, shadow AI risks and data sovereignty pressures. The 2026 Nutanix ECI reveals the gaps.

Get tech leader insights to move faster and smarter.

Get more stories by subscribing to The Forecast.

Podcast transcript:

Jason Lopez: Banks, hospitals, government agencies, organizations in most sectors are all building AI into how they operate and they’re hitting a wall. The compute infrastructure wasn’t built to run old systems as well as the new ones at scale, and that means running systems securely without breaking. Nowhere does that strain show up more than in banking and financial services.

Sean O’Dowd: I guess my head immediately goes to, this is going to get uglier and scarier before it gets better.

Jason Lopez: Sean O’Dowd is the head of financial services solutions at Nutanix. This is the Tech Barometer Podcast. I’m Jason Lopez. On this podcast, we talk to O’Dowd about the 2026 Enterprise Cloud Index or ECI. It’s Nutanix’s annual global research survey on IT trends. When we asked him how financial services are doing overall, he compares what’s happening to things like the shift from human labor to mechanized factory production in the late 1800s or the way nuclear technology reshaped geopolitics or even how space exploration, which had a big front end investment, ultimately proved transformative.

Sean O’Dowd: So I say it’s going to get uglier before it gets better because as we all appreciate when you look at these awe technologies, the disruption, the market structure changes, especially in financial, this has the potential to really upside a lot. I think you’re going to see those frictional and structural pains play out over the next 10, 15 years like you did in these other major industrialization cycles. So it’s going to get a lot uglier before it gets better. But I still have a lot of conviction that there will be a lot of positive outcomes here. And we have to be thinking about not just the technology, but how do we govern this? How do we police it? What do we want for ourselves? These are big questions that will get addressed whether we want to or not. They will percolate up.

Jason Lopez: O’Dowd has been an observer of the banking industry for nearly 30 years. He’s seen the ups and downs and lately identifies macro factors which map to his earlier comment. Things will get uglier and scarier before they get better.

Sean O’Dowd: Volatility is good for a lot of financials, so are where rates are. So from a business standpoint, these guys, they’re profitable. They’re posting strong earnings. Regulations are in their favor. So business is good. Being a bank I think is good business right now. However, it’s increasingly costly. So the big things that I continually look at are what are the executives thinking about and obviously the tech behind it.

Jason Lopez: The cost pressure is real money. According to Forrester, financial services as a whole is projected to spend nearly half a trillion dollars on technology in 2026, about 17% of total US tech spending. Findings from the Nutanix ECI report provide more insight into how this is playing out. Shadow IT is one of O’Dowd’s ugly scenarios. Suppose a loan officer asks ChatGPT to summarize data in a complex customer file. Suddenly data like a social security number sits outside of the bank’s firewall. It’s just one of a myriad of weak links. According to Nutanix’s ECI report, 86% of financial sector executives believe shadow AI tools introduce severe business risk.

Sean O’Dowd: Yeah, for me, it’s expected, but also surprising given the FinServ industry has some of the most mature risk management discipline across sectors. They manage credit risk, market risk, op risk with a lot of rigor. The fact that two-thirds of them are discovering AI tools are deployed with zero oversight. I think that gap between how seriously they take every other category of risk and how exposed they are on that government as this percolates up top and as oversight sees and hears, they are just turning the screw on that operational oversight.

Jason Lopez: AI is pulling most of the tech world toward the public cloud, but financial services isn’t fully going along. 79% of FinServ’s IT leaders call data sovereignty a top priority and enough caution that it’s capping public cloud use at just 62%. Odowd says the reason isn’t really about any one company. It’s about what happens if too much of the financial system ends up depending on the same few points of failure.

Sean O’Dowd: We’re really worried about concentration risk with a few IT players. We’re also concerned about sovereignty. So how do we protect the financial infrastructure so that there’s not disruptions? Yeah, it’s not really a play against AWS and so forth, but it’s just ensuring that things like payment transaction doesn’t fall away or trading markets aren’t disrupted from this thing. So that’s the big one.

Jason Lopez: Banks are pouring money into AI, but it isn’t clear if it’s paying off just yet. O’Dowd talks to CIOs across the industry regularly and notices some common blind spots, including managing the cost of using AI.

Sean O’Dowd: There’s fatigue in the way that it can present actual risk to adoption long term. And I think that’s moving away from what we saw last year was build it and they will come. That fatigue is just forcing it on employees. Cost and management obviously is huge. When you have a tool that’s generating a hundred thousand in savings, but you spend a million on tokens to get there, what’s the true value of that?

Jason Lopez: One of those cost structures is tokens. The per use currency that AI models run on and one that’s quietly exploding.

Sean O’Dowd: On token usage, it’s up 500% in one year. JP Morgan even talked to the fact that they have employees that are spending more on tokens than their own salary. The other thing that’s come to bear is not just the cost, but how to track it. So there’s just a massive lack of transparency into the bill that they are paying on a monthly basis. A lot of these guys just don’t have that granularity and able to say, “Hey, it’s costly, but it’s worthwhile.”

Jason Lopez: The lack of transparency isn’t just a financial headache. It’s a symptom of exactly which banks have their AI infrastructure under control and which don’t. The ECI report cites the experience of Fairway Home Mortgage, which has streamlined mortgage processes and preserved high governance standards while rolling out agentic AI. O’Dowd points to the democratization of data and information.

Sean O’Dowd: If you go back 10 years ago during the Hadoop and big data wars, there was a lot of promise there. But as we all know, AI took what was almost like a data consolidation exercise within these organizations across all types of data to actually being able to really unlock value at speed with it. Yes, intelligent applications, yes, omnichannel. Yes, better intelligent information driving it and automating workflows. But cloud didn’t prove to be the easy button, although it’s accelerated. Fast-forward, you find yourself, I think a bank CIO having to manage the dual architectures that you’ve built or even more between on-prem, cloud, co-location. So it’s just gotten entirely way more complex. We all know this. The data problem, they’ve done a good job, I think over the past five, 10 years trying to build out those databases, consolidate and catalog that information. It increasingly becomes harder with AI. And this is where the chief data officer really I think are key. And that is because they’ve spent a lot of time trying to catalog it in big data for big data elements and machine learning. The data tagging, what I’m seeing is because foundational models become so expensive, they want to use open source models. But in order to use that, they’ve got to do what they’ve done previously, which is build up that data hierarchy, data catalog, data tagging in order to make those cheaper, sometimes lower quality, intelligent foundational models that are open source.

Jason Lopez: Knowing why AI is expensive is one thing, but building infrastructure that can actually handle its demands is another. ODOWD explains why banks are getting better at managing existing and new applications, which are typically built with cloud native containers.

Sean O’Dowd: AI workloads, they’re dependency heavy, they’re expensive to run, they’re bursty, and they’re increasingly modular. And containers helps address those pressure points all at once. So AI adoption I think is pulling that container adoption along with it. I think what’s different, at least the way I’m seeing it from my seat, is what’s the mix of new applications being built and what’s the mix of legacy applications that are changing within the bank? There’s a real come to Jesus conversation right now in terms of which ones to tackle. We increasingly see consolidation in the market as a means of survival. There once was 40,000 banks in the US, now there’s 10,000. To compete, you’ve got to outspend and out maneuver. You’re just going to see fewer banks doing it well. Those that can spend, those whose management can steer through. Those that get larger and consolidate are kind of going to win here.

Jason Lopez: Sean O’Dowd is Nutanix’s head of financial services solutions. You can read the 2026 Enterprise Cloud Index at nutanix.com/enterprise-cloud-index. There you’ll find industry-specific reports for financial services, healthcare, and the public sector. This is the Tech Barometer Podcast. I’m Jason Lopez. Tech Barometer is produced by The Forecast. And you can find more podcasts like this and articles about the tech industry and the people in tech at theforecastbynutanix.com. That’s all one word, theforecastbynutanix.com.

As agentic AI rewires enterprise IT, Cisco and Nutanix are turning a three-year strategic partnership into a full-stack answer for infrastructure under pressure.

Get tech leader insights to move faster and smarter.

Get more stories by subscribing to The Forecast.

Video podcast transcript:

Jason Lopez: AI demand, security, and legacy infrastructure are a few of the challenges in enterprise computing. Instead of rebuilding from scratch, IT looks to networking to make systems work.

Jonathan Gorlin: It’s a complex time to be in IT and technology, and I’ll say there are probably three trends that we’re looking at right now, plus I’ll give you a bonus trend. So first of all, what we’re seeing is, of course, with the proliferation of new types of applications, cloud native, AI, of course, is top of mind. So how do we get IT to grapple with all of these newer applications that are coming while still having to maintain all of the older, more traditional applications, right? It’s not an either or, it’s an and, always with IT. It’s this and this and this. How do we manage all of that? So that’s one big trend. The second trend that I’ll say is just how do we simplify? And our customers are telling us they no longer want the pieces and parts to build an architecture. They really want a full-stack, validated turnkey solution that they can deploy at scale. And so really simplifying everything in terms of how you manage this environment is really top of mind for customers. And then third is really what we see about with edge and this proliferation of data at the edge. How do we capture that? How do we turn that into value? So we have some exciting announcements between the two companies related to that as well. And then my bonus one is something that I think we’re all dealing with right now, which is just supply chain. And how do we get a hold of hardware? And how do we deal with just this incredible churn in the system of how do we get equipment? How do we continue to grow our environments? And so we’re really trying to address all of these problems with our joint partnership.

Ken Kaplan: Do you see the supply chain issue as just another challenge on top of many other challenges? Or is it something you’ve seen happen in the past? Put your perspective.

Jonathan Gorlin: Yeah, I would say this is not an unfamiliar territory. IT always has to deal with unknowns and challenges. And that’s what makes it such an exciting space to be in. But this is certainly maybe to a greater extent than what we’ve seen before. And of course, it’s changing day by day. I get asked constantly, you know, Jonathan, what do you think? When is it going to end? And I wish I had the magic crystal ball. But we will get through it on the other side. And I think what’s important also is we’re helping customers maximize the existing infrastructure that they have. And we can get into that a little bit later to really help hold us over while we get through this period of just a little bit of instability, I would say, in hardware availability.

[Related: Hybrid Multicloud IT Teams Benefit from Cisco and Nutanix Partnership]

Ken Kaplan: Customers want a full solution, which could include a variety of different technologies. And you’re helping them find that right connection. What are they looking for now to get through this period? Are there some things that they’re asking you specifically?

Jonathan Gorlin: Yeah, absolutely. I would say really it comes down to how can I continue to run my environments and do our job and answer to all of our stakeholders in the event where it would take me longer to procure new hardware. And so we’re doing some specific things within the partnership. Actually, the timing is perfect. We’re here at .NEXT. Nutanix and Cisco have been leaning in heavily into external storage partnerships as well. So the timing couldn’t be more perfect. And what that means is even in a time where it’s harder to get memory and NAND flash, we can actually leverage those existing investments from some of our storage partners. We can bring that in and extend the life of existing compute environments. And we’ve gone so far as to even go back and qualify some of the UCS B200 blades that we used to have. And we have a huge install base of those. So we wanted to unlock for customers and say, how can we use what’s on your floor now? How can we bring it into the easy operational model that Nutanix offers for lifecycle management? And so we’ve actually delivered on that. And at the show, there are going to be some new announcements about more storage partnerships. So we’re just continuing to lean and qualify more hardware, being able to reuse as much as you possibly can.

Ken Kaplan: So we’re thinking storage, we’re thinking compute. What’s the role of networking in that?

Jonathan Gorlin: Yeah, certainly networking is fundamental when you’re talking about storage, right? Storage has to traverse the network. So I’d say there’s really maybe two areas that I would look at here. One is when you’ve got distributed scale-out storage, you have to have a high-performing network. And that’s something that we do really well with the integration between our two companies. If you remember, I mentioned one of the challenges is customers don’t want to build it themselves. They want a turnkey solution. And with the Nutanix tooling integrated with Cisco Intersight, which is how we manage our compute, we actually integrated the two management tools together. What that allows us to do is out of the box, we pre-configure not just the compute and the hypervisor and the storage, but also the networking, the east-west high-speed fabric. And that’s really what UCS brought to the game 15 years ago, like what made Cisco Compute and UCS unique. It was about integrating the network into that. And so we’ve got a best practices, out-of-box configuration for the network, which is what we need to ensure high performance. And then at the same time, the requirements, I talked about those applications coming in, now AI is just demanding more and more. There’s an insatiable demand for bandwidth and compute and GPU resources. And so I think the need is going to continue to grow, but we’re here and we’re ready for it.

Ken Kaplan: At some point, everyone’s going to do this, but right now people are kind of trapped. Just describe what the challenge is that these kinds of partnerships are overcoming.

Jonathan Gorlin: Yeah, I think traditionally there was kind of a way that Nutanix and other server vendors would kind of come together and qualify. It was typically more of like you take Nutanix software and you qualify it on a server platform and like off we go, we’re off to the races. And that worked well when we had abundance of hardware and we didn’t necessarily need quite as much flexibility. But three years ago when we formed this strategic partnership between Cisco and Nutanix, we were very thoughtful about challenging the engineering teams to how do we get away and break the chain of having this inflexibility. And the engineering teams, boy did they deliver, right? And so we thought about this and now it’s paying dividends for customers so we can go back to existing customers, we can help them navigate the situation that we’re in. And we’re not going to be in this situation forever, right? This will pass and we’ll get back to normal but customers are not going to want to go back to an inflexible model. So I think we’re there. We’ve spent the time to invest there and that’s the compute side and then on the storage side, the Nutanix teams have done a phenomenal job of working with some of the biggest storage partners out there. And at the show here we’re talking about Everpeer as well as NetApp. So these are going to be huge for our customers because those are synonymous. We have a CI converged infrastructure business with both of those partners and it’s huge with Cisco UCS. And so we’re really excited. NetApp is the newest member of the family that we’re going to be talking about through the rest of this year. But also FlashStack and Everpeer is shipping and available now. We have a number of very excited customers using that.

[Related: Rapid Spread of Hybrid Multicloud IT Drives Need for Simplification]

Ken Kaplan: You want your networking security capabilities to be able to plug into somebody else’s technology. So security, being able to plug into that and controls from another partner, is it more intense these days than ever?

Jonathan Gorlin: It is. And we’re definitely, just like Nutanix has done for the last several years now, we’re really embracing an open ecosystem. It’s really important. I think gone are the days where you’ve got proprietary stacks and you decide you do it your way and you don’t interoperate. I think those days are behind us now. That’s another trend that we’re seeing. And so we’ve embraced this for many years now with our UCS platform. We call it a platform because really, we want to support the best of breed solutions in the data center, right? We’ve got a phenomenal hardware platform and a management layer of insight of how we manage this, as well as the fabric that interconnects it. But we need to support the best of breed, including the Nutanix stack, both for traditional VMs and HV, as well as the Nutanix Kubernetes platform for newer applications. But we also need to support our other partners, right? Microsoft, Red Hat, Broadcom and VMware. We have to support everybody. And it actually works out as a net positive for customers. Because the way we support our hardware platform is we actually take a agnostic approach. We don’t build specific hardware appliances. Instead, what we do is you standardize on a compute platform, and then we allow you to retrofit into any solution. So this allows customers, and this is what our customers really love about the joint Nutanix-Cisco solution, you can take existing servers that you have that maybe are running VMware today, or maybe running Windows Server today, and you can bring those into a Nutanix environment with no compromises. So that’s really critical, and we’re not yet seeing that from other compute vendors. And so that’s really part of the secret sauce between the two companies.

Ken Kaplan: How has the view from a networking perspective changed with cloud native, with the demand for AI, AI factory type infrastructure? How has that changed your world?

Jonathan Gorlin: Yeah, well, I definitely say that networking is just a key, when you talk about these distributed systems, whether we’re doing training or inferencing, how quickly you can access the data is everything. How do you make sure that GPU utilization is the maximum it can be? These are not inexpensive parts and components that IT is investing in. How do we make sure we keep those components at peak efficiency, number one, but also just in terms of how AI has evolved, we’re moving from LLM chatbots to now agentic AI, that’s actually changing things quite a bit in terms of the pattern in the network. So it used to be kind of a request response kind of traditional model that we’ve had for quite some time, but now when you unleash all of these agents, and really we should be thinking of agents now as almost extensions of our workforce, right? Now we’re going to see just an insatiable demand for bandwidth and we have to continue, all these agents are going to be working on our behalf and the only way that they can get stuff done is by communicating with each other and communicating with our important data for being able to make decisions. So I’d say the network is definitely crucial and the connective tissue to make all of this work, but also we shouldn’t forget about security, and security has to be paramount in this new world that we’re in. So doing network and then baking the security into the fabric of the network is really crucial and that’s really what we’re thinking about at Cisco. With these agents now, we have to protect agents from the world, can imagine from prompt injections and leaking confidential information, we have to be careful about that, but we also have to think the other direction. How do we protect agents? How do we protect the world from agents? The other way around, right? And so you can imagine now when decisions are being made in an autonomous way, where is the accountability? So we have to make sure we’re very thoughtful about how we do this in a very secure way, we put the right guardrails in place and the right way to detect when there are situations like this. So these are areas that Cisco is really spending a lot of time thinking through.

In this Tech Barometer podcast, Eightfold AI CMO Navneet Singh makes the case for using AI to improve routine human resources processes.

Get tech leader insights to move faster and smarter.

Get more stories by subscribing to The Forecast.

Audio podcast transcript:

Navneet Singh: The current system is actually already deeply biased. The recruiter who sees 200 resumes by Friday afternoon, he or she is not evaluating the last 50 with the same rigor as the first 10. And this is not a character flaw. It’s just human judgment, just degrades with volume and fatigue.

Jason Lopez: Navneet Singh is the CMO for Eightfold. It’s a company that provides AI-based HR tools, which they describe as talent intelligence. This is the Tech Barometer Podcast produced by The Forecast. I’m Jason Lopez. Today we’re going to look at AI-based hiring. The gist of what Navnit tells us in this story is you can’t just dump data into a typical AI engine and have it do the job of hiring. It won’t turn out well. So one wrinkle Eightfold starts with is to focus on skills rather than resumes. A resume focus often overlooks the intangible capabilities of people which Eightfold’s AI recruiting and tracking platform takes into account. It also matches people to open roles inside a company. Singh says it takes a lot of work to get AI recruiting like this right.

Navneet Singh: It’s not whether AI is safer than humans. Typically, HR and recruiting has been done by humans. The real question is, is it fairer than humans, especially at scale? And the honest answer is that well-designed AI is.

[Related: IT Sovereignty Moves Beyond Location]

Jason Lopez: His view is one that scientists hold about the unreliability of human perception. Scientists in other fields like biology or astrophysics use tools to measure findings. Tools that work the same regardless of what kind of day the scientist is having. It works the same in the hands of each scientist. A well-built tool such as an algorithm in the hands of an HR person can do the same thing.

Navneet Singh: It can apply the same standard to candidate one as it does to candidate one million. The bar doesn’t move. The evaluation doesn’t drift. So that’s where critically every decision can be logged, it’s auditable. You can actually see bias if there is one and correct it.

Jason Lopez: But here’s where AI falls short. Yes, it makes calculations and machine decisions which are consistent, but ultimately not human decisions. Just as AI can identify what a joke is, analyze it and describe it. If you wanted to write a joke and the output to be actual humor, it requires a human in the loop. Singh draws a similar line in a human resources setting.

Navneet Singh: Where it should not be used is fully autonomous decisions. It should surface signals. Humans should make the judgment. Performance management, terminations, leadership assessments in terms of cultural nuances. So just simple principle is that AI executes on volume. It should surface signal. Humans own the decisions that carry any consequence.

Jason Lopez: That line matters because the scale he’s talking about is enormous. He points to Amazon. I

Navneet Singh: Was listening to Amazon, right? It was saying that last holiday season had to hire 250,000 people. It’s just a mind-boggling number.

[Related: Heading Off Data Harvesting Before Q-Day]

Jason Lopez: And he says that’s exactly the kind of high volume hiring where AI is already being deployed. It’s also being regulated more. The EU’s AI Act classifies hiring algorithms as high risk, contending they require transparency and human oversight. New York City has its own laws on the books. Singh says companies shouldn’t wait for regulation to force the issue.

Navneet Singh: Internal guardrails should be, for example, audit samples and audit bias before deployment. People should know when they are being evaluated by AI. There should be human review gates that should be very clear to candidates as well as the hiring people or committee. So data governance, data should reside in the country or state based on the laws and regulations. And all of these should be accountability of the vendor. When you buy an AI system, you need to understand what it was trained on, what bias audits were conducted. So all of that is really critical.

Jason Lopez: And this is what it looks like as Singh walked us through it.

Navneet Singh: Imagine an AI agent that you can deploy and candidates can interview with the AI agent at any time the candidate wants. You don’t have to do scheduling. You don’t have to look for when the hiring manager is available or the recruiter is available. Candidates apply an interview at any time on demand.

Jason Lopez: The agent doesn’t just conduct the interview. It takes the notes that hiring managers used to have to chase down with a full recording of it that anyone can go back and review. And he says a similar model extends into an employee’s career after they’ve been hired.

Navneet Singh: Our AI has been trained on 1.6 million skills that have been used over a billion career trajectories. We already have the data that if you have skills A, B, and C, you are much more likely to be able to acquire skills X.

[Related: Agent Gateway Enforces AI Token and Traffic Control] 

Jason Lopez: For companies wondering where to begin, his advice is to not do a big rollout, but to start with a small project.

Navneet Singh: Start with a use case where the risk potentially is lower and the ROI is high. So as an example, we internally used to go to 10 colleges at universities for interns. Now we are able to go to 10 times that because we can deploy the AI interviewer agent to screen candidates. And just screening, it’s just the first step. Humans are still involved in the next. You are just expanding the talent pool.

Jason Lopez: He’s also blunt about what he thinks companies get wrong.

Navneet Singh: Don’t just invest in the general purpose AI because it’s not been deeply trained in HR and hasn’t been audited for bias. And then you can’t automate accountability. Humans still have to be accountable. Humans still have to use human judgment and any decisions of consequence must be made by humans.

Jason Lopez: As Singh returned to the idea of that guardrail of humans deciding what matters, he pointed out that most companies will be greatly challenged to even get there.

Navneet Singh: The MIT report, which said that only 5% of projects succeed in AI, the way to success is accountability at the exec level, but people who source AI, they should be people who are prosumers, which is people who are at the ground level, who are already using AI in their daily lives. They should be sourcing the enterprise technology for AI.

Jason Lopez: Navnit Singh is the CMO for Eightfold, which makes AI-based HR tools. This is the Tech Barometer Podcast produced by The Forecast. I’m Jason Lopez. Thanks for listening. The Forecast is a technology news publication from Nutanix covering a wide range of tech stories as well as people in technology. You can read more stories or listen to other podcasts by going to theforecastbynutanix.com. That’s the forecast by Nutanix, all one word.com.

As AI shifts from training to inference and call-and-response chatbots give way to autonomous agents, AMD’s Brayden Mahdavi explains why enterprises are rebalancing the CPU-to-GPU ratio and turning to nimble neo-clouds for cheaper, faster compute.

Get tech leader insights to move faster and smarter.

Get more stories by subscribing to The Forecast.

Video transcript:

Brayden Mahdavi: I really sit at the intersection of three different groups. I sit between cloud providers, so this is your kind of typical cloud service provider, then your infrastructure providers, and then end customers within our enterprise. My role really gives me a lot of energy because I hear a common theme across all of these groups, and that is customers want options. I’m really seeing it across three main layers. So the first is really the infrastructure layer, right? This is your silicon choice and optionality across different silicon providers. And then you have within the cloud service provider layer, right? So my traditional tier one hyperscalers, but now we’re seeing this new emergent AI specialized cloud provider often referred to as neo-clouds. And the third layer is really your virtualization layer. As many enterprises know, licensing structures have changed as of recent, and that is creating a lot of challenges and a lot of ingenuity around how we reinvent our stack to better serve our business needs.

If many enterprises can put their whole fleet on this tier one cloud, albeit Amazon, Google, Microsoft, and that business is critical for AMD. It’s not going away anytime soon. Given some of the challenges that we’re facing in 2026, capacity and supply are huge challenges. We are really hungry at a market level for a new offering, right? And this is where you are seeing these emergent AI cloud providers that are specialized historically maybe in the GPU space, but many of our partners are seeing with the shift to agentic AI, a bigger need for a CPU. They’re so nimble and lean that they can help productize our technology in a way that is favorable for our end customers who are facing challenges around total cost of ownership and other performance challenges. We are seeing a major industry shift from the classic call and response AI that was your traditional ChatGPT to now agent calls.

And agents can call more and more agents to do more autonomous tasks for us. They bring us a lot of value. What we don’t see is that they demand a lot of compute resources. What we are going to see very likely over the next few years is a shift back to parody, if you will, of the CPU to GPU ratio. So where our enterprise customers were typically buying as many as eight GPUs to one CPU, we’re seeing that ratio come down much more to a one-to-one ratio because our CPUs are able to handle much of the orchestration and much of the agent calls, the tool calls that are required for these types of workloads. The last few years from that ChatGPT moment in 2023, when a lot of people realized this is a real invention that is upon us today, the shift has really been many training cycles within the frontier lab environments.

We’re doing a lot of forward passes and back propagation at scale to train up these models, whereas in the next couple of years, it will be very heavily favored towards inference. Now we want to see the outputs of these models. And as we get to that shift, the demand for compute just continues to soar through the roof. Our customers are starting with their workload need and working back to solve what the right infrastructure stack is. We are seeing more optimization in that process, whether it’s through a FinOps practice or whether it’s through just a straight up business need. It’s at a time where we are moving into optimization and how do I lower my total cost of ownership? And between AMD and our partnerships with Nutanix and with our AI cloud providers, we are very confident that we are going to be able to help our end customers accomplish their business requirements.

In this video interview, HyperFrame Analyst Don Gentile explains how data storage is shifting from a passive to an active participant in AI, raising a defining question: bring compute to the data, or data to the compute? He says months of supply chain pressure is pushing enterprises to squeeze more from existing infrastructure and lean on private cloud and neo-clouds.

Get tech leader insights to move faster and smarter.

Get more stories by subscribing to The Forecast.

Video transcript:

Don Gentile: With the advent with GPUs now, and we need to constantly feed those GPUs, the concept of performance and the AI data pipeline, it really a point of contention is in the storage. The bottleneck happens there. And so you have to start to think about how is that storage media, whatever that McFlash, for example, going to feed those GPUs to make sure that they’re not starved, that they’re constantly being fed the pipeline data. So that’s a big shift.

Ken Kaplan: IT teams, are they having to use a variety of different types of storage technologies today or can they get it all done simply with one kind of storage?

Don Gentile: That is a question for every organization to deal with, right? In some cases, an organization might want to reuse their existing storage for cost performance reasons. They might have the ability to bring on new type of storage platforms. And so it’s a little bit of a purpose built exercise there. Certainly you have three tiered environments where you’ve got your hot, your medium, your cold storage for cost and performance reasons as well. There can be offsite archives, data archives, which is the least expensive and also air gapped for data protection reasons. So there’s a variety of different storage platforms that exist out there. The evolution then has been about thinking about how are we going to feed those AI pipelines and how does storage become an active participant in the AI process?

Ken Kaplan: Storage seems like it’s a very dynamic environment. There’s just a lot of evolution around that. Why did that happen and why do people want new types of storage capabilities?

Don Gentile: Maybe that started about 10 years ago. I could probably put a pin on it and say, if you start to think about where the workloads are going and you start to think about the advent of AI, which ChatGPT had not happened yet, but that moment was going to happen. So a number of companies started thinking about how do we need to design for those future workloads? And so you can rethink storage as more of a substrate that is managing and coordinating and governing across within a data center from cloud to on – prem and of course across geographies as well. And so things like global namespaces emerged where you have to start keeping track of where all that data sits. And then the ultimate question I think for companies is going to be, do you bring the compute to the data or do you bring the data to the compute?

And so that also will influence your storage decisions.

Ken Kaplan: How is storage evolving when we have more activity at the edge?

Don Gentile: Right. Well, if it’s a real time inferencing kind of experience that you have to have, you don’t have the time for that round trip. And so vehicles are part of that autonomous vehicles. And so if the decision has to happen in the moment, there’s no round trip to the cloud for that inferencing to happen. So you have data collection, you have data analysis, and then you have the action from that happening all at the edge. And in some cases you can aggregate, you can collect data from the edge and you can bring it to a central location. It really is going to depend on the application.

Ken Kaplan: You have a perspective from HCI and now storage and security. How have you seen those things evolve?

Don Gentile: Well, I think where companies are looking now is more toward private cloud instances. It’s a very rational way of thinking about how to evolve for flexibility and for cost reasons and for future proofing. We can’t always predict where things are going to go, so you need to be able to add new technologies or ad scale. And so private cloud gives us that opportunity to be able to scale and add technology as it occurs. So it’s more future-proofing.

Ken Kaplan: You’re talking about people who do want to still manage some of their infrastructure. They might use cloud and what they own. What happens when they decide to migrate to the cloud or start all in the cloud?

Don Gentile: I think it’ll be a combination of things. So there will be situations where you want to get more out of your existing infrastructure because you’re not able to add the capacity at the pace that you plan to. It might also mean abandoning or postponing some projects where you say that we’re just not going to be able to get to that right now because we don’t have the capacity. Now you also have the neo-clouds where you have the ability to scale out to a third party to access that capacity in the cloud. That’s why they exist. And they have the access to the current GPUs that are coming out of Nvidia and AMT. And so you can look to the NeoClouds as the next scalable option for enterprises that are not able to access that capacity on site. So that’s another option.

Ken Kaplan: In some ways they’re forced to do these things. The capacity is not there. It’s an interesting time.

Don Gentile: Or they may need to postpone existing projects. Some may say we’re getting more value out of this. We’ll postpone that. We’ll keep focusing on this. And if we’ve got existing infrastructure that we can repurpose, and vendors are helping with that too. There are programs, there are software solutions as well to help people to get more capacity.

Ken Kaplan: Is that driving people to get more out of what they have? Because they will need to keep adding storage. It seems to be a given.

Don Gentile: Well, from a storage perspective, we have a supply chain issue and that is really pressing for a lot of organizations. So that impacts performance, it impacts cost. And so a lot of organizations need to be planning for that. Not just the enterprises, but the vendors as well. And so there are vendors like Dell and others who have long-term contracts and access to NVIDIA GPUs and all that ad infinitum, but there are cost pressures on top of that too. And you see storage vendors and memory vendors that are starting to elevate their prices. And we have 18 to 24 months supply chain constraints. And so that really could change people’s plans about how fast they’re able to adopt AI, how fast they’re able to extract value of AI because now they have to say, “Well, I have to get more value out of my existing resources.” There’s no magic store I can go to, to purchase more capacity like that. And so the memory that GPUs, all these assets are suddenly constrained and enterprises and vendors are rethinking their plans and saying, “How do I adapt to that supply chain issue?”

As artificial intelligence innovation outpaces the early days of public cloud, CIOs face the burden of parsing through noise to build a cohesive infrastructure stack, says analyst Scott Sinclair, practice director at Omdia, in a video interview.

Get tech leader insights to move faster and smarter.

Get more stories by subscribing to The Forecast.

Video transcript (edited):

Scott Sinclair: I think something that is absolutely crucial when you think about artificial intelligence, there is no one single company that is the one-stop shop for everything you need from artificial intelligence. Not even Nvidia, even though they have great technology and they do a bunch of great things, it requires a number of different tools and technologies. Not only that, but we’re also seeing rapid innovation in this space. Everyone’s talking about agents right now. Nobody’s really talking about agents a year ago. So in next year, there’s probably going to be something else that we’re all talking about that no one’s talking about this year. So the challenge is that it’s requiring partnerships A, because nobody can offer a full stack, everything you possibly need, infrastructure code, tools, everything. Number one, but then also the fact that organizations want to partner because they also need to stay in touch with all the new innovation that’s happening to make sure they don’t get left behind.

And so if you think about that from a business standpoint, from a consumer or an IT leader, a CIO standpoint, on one hand, that’s great because what you want is more partnerships, it provides better validation to solutions. Hey, look, these companies are working together. They pre-validated. They’ve made sure it works. So it reduces some level of risk. That’s helpful. The other part, though, is that it does make it complicated because every organization has an AI story, and they all sound very similar. And where the burden falls, unfortunately for right now, it falls to the CIO and their team to parse through the noise and understand, okay, yes, every company has an AI story. What is your building block, so to speak, that you’re being delivered? How does that fit with the rest of the other building blocks? Who do you partner with, and what’s the best way to leverage you?

[Related: The Rapid Rise and Future of Neoclouds]

And these are the types of questions that we see being asked over and over again, and organizations are just going to have to keep asking them. We see a lot of excitement in, okay, how can I put in agents? How can this help manage my IT operations environment? But at the same time, we sit and say, “Well, there are a lot of tools that are already developed to do a lot of AI, do a lot of automation for infrastructure, as well as optimization.” So right now, I think we’re still in a ‘we need to learn’ phase. I see a lot of interest in things like knowledge sharing. How can we basically put all our training manuals into AI and make that easier to search? We see a lot in terms of how do I use AI to do things like work with customer support and trouble tickets, things like that.

Those are early areas where people have seen adoption. I have seen organizations say, “Well, look, maybe we can do some sort of agent-based model to do troubleshooting and put in, identify possible root cause of issues, and some sort of remediation.” I think we’re still very early in that stage, but honestly, this is a stage that even I, as an analyst, am actively researching to see where organizations are. I think this is going to be a rapidly evolving space, and it’s going to be where different organizations just have a different level of risk appetite to how they want agents to work and how quickly they’re able to adopt them. Put in perspective how fast people have to learn things nowadays. As I think about that question, because it’s something I think about quite a bit, I want to compare it to the early days of public cloud adoption, because it was very much AWS was out there, you saw movement from Google, you saw moving from Microsoft.

[Related: Dual Native Plugs Containers into VMs or Bare Metal]

Okay, how do we modernize? Everyone had to move to the cloud. That was the thing. We all need to move to the cloud. But even that, even though it represented a fundamental paradigm shift in the way in which we think about apps and infrastructure and even IT operations, it was somewhat contained because at the end of the day, there were still a handful of major players. There wasn’t just the proliferation of hundreds, if not thousands, of different startups of different tools and different technologies, and everyone’s doing your own thing. And also, cloud in a sense was a way of optimizing things we already had, right? It was a way of, instead of doing our own data center, we can offload the work to somebody else. Okay, look, I can access it not from a component level. I can do it from an SLA level. So the language and the ideas stayed very similar.

I think with AI, one of the fascinating elements is the area of what’s unknown is much bigger than it probably has ever been in a long time. I can’t think of a timeframe where there’s just been this combination of scale of unknown, as well as rapid evolution and innovation happening all at the same time. So yeah, to me it’s unprecedented.

Data harvest is already underway as bad actors await the description power of quantum computers. In this Tech Barometer podcast, technology experts Marco Graziano, Steve McDowell and Don Gentile explain why IT organizations must implement quantum-resistant encryption now, not later.

Get tech leader insights to move faster and smarter.

Get more stories by subscribing to The Forecast.

Transcript:

Marco Graziano: All infrastructure today, inter-company communication and on the internet, is possible because you can protect this communication channel with encryption. It’s all secure because of encryption.

Steve McDowell: I can take a key that would take, you know, however many hundreds of years on a GPU today and solve it in hours on a quantum computer.

Marco Graziano: Overnight, communication becomes insecure.

Don Gentile: When you think about the bad actors, they have access to harvest now, decrypt later, which means they’re collecting your data today and it can be broken with tomorrow’s quantum computing decryption capabilities.

Jason Lopez: Three technologists, Marco Graziano, Steve McDowell, and Don Gentile are saying pretty much the same thing. Encryption, as we know it, is facing an existential crisis. According to Marco Graziano, it’s closer than most of us think. This is the Tech Barometer podcast. I’m Jason Lopez. Graziano’s path started decades before anyone was talking about quantum computers. Early in his career, Graziano worked at Olivetti. It’s the Italian company whose 1965 Programma 101 is widely considered the world’s first personal computer. Its design would go on to influence Steve Jobs and the original Apple. Olivetti was the kind of place that launched many computer science careers, shaping Silicon Valley and beyond. Olivetti alumni like Graziano are still at it. From pioneering streaming media, home security and blockchain technologies, he now pushes the limits of AI with his whole system approach.

[Related: Agent Gateway Enforces AI Token and Traffic Control]

Marco Graziano: In the time of AI, specialization is over. Today you need to be a generalist, not a specialist, because your specialization will never be as good as AI. I remember in the 80s and 90s, in computer companies, there were people who dedicated their whole career to SCSI firmware, to BIOS, to Ethernet firmware. Their whole career, they wouldn’t be able to do anything else outside that area of expertise. And that expertise was so deep that it was amazing. These jobs don’t exist anymore. Today, my recommendation to a young generation of engineers is think systems, think outside the specialization. Become an inventor. Go back to the Renaissance. What pieces fit together? How do they fit together? What does working mean? What is the end goal?

Jason Lopez: This background on Marco’s view of technology and innovation helps to understand his concern about quantum computing.

Marco Graziano: The moment I realized that this was an issue is when I read about Harvest today and Decrypt tomorrow, because that’s not obvious. There is a risk today, if quantum computers are not available, for communication that we are exchanging today, because this communication can be collected, harvested, and become unprotected one day soon. So that realization made me move in working in this.

Jason Lopez: And some of what’s being harvested right now isn’t sitting in a well-guarded data center.

Marco Graziano: So we know that devices will hit Q day. We know that communication that’s being collected today by bad actors will become visible to them sometime in the next few years. And devices carry critical information, some of them very critical infrastructure, for instance. Think about utility devices, smart grid, and so on. These are systems and devices that are commissioned somewhere in the middle of the desert sometimes or outside the cities, in a distribution station or so on, and forgotten, but yet they carry critical infrastructure information.

Jason Lopez: It’s not just remote infrastructure. Graziano says the exposure runs through nearly every institution people already trust with their most sensitive information.

Marco Graziano: We can imagine banking and financial institutions, insurance, medical information, it’s all at risk.

Jason Lopez: The reason it’s all at risk comes down to speed. Today’s conventional machines could take hundreds of years to break a key. But it may take minutes with a quantum computer. So that is the difference.

[Related: The Rapid Rise and Future of Neoclouds]

Marco Graziano: That is what we are looking at.

Jason Lopez: Graziano isn’t vague about when that difference starts to matter.

Marco Graziano: It’s going to be broken. So all this communication that we believe today being secure and protected is actually going to be broken sometime soon in the next few years. Maybe five, maybe ten. But it will happen. The technology behind quantum computers is progressing very rapidly and accelerating in the last few months. So now prediction is between 2029, 2030 all the way to 2035. But it’s going to be in this time frame. Quantum computers will not be for the masses. They will be used to solve specific problems, which we don’t have full grasp yet. But it’s clear that it’s not going to replace even special purpose computers anytime soon. But one of the things they’re good at is in breaking cryptography.

Jason Lopez: Steve McDowell of NAND Research describes the technology in similar terms.

Steve McDowell: They’re very expensive. They’re very esoteric. There are only a handful of companies building them. But they’re real. And we’re starting to see some commercial adoption in the oil and gas industry first. We’re already seeing that. But five years from now, I look at quantum. That’s where AI was in 2019. That’s where we are with quantum.

Marco Graziano: They are capable of performing millions of computations at once, in an instant. And today computers, you know, they do well, but not that well.

Jason Lopez: Though the risk level is high, Graziano says it can be addressed.

Marco Graziano: The good news is that the technology and the software to protect it, encrypt the channels in presence of quantum computers, is available. NIST and other world organizations have put together a consortium and came out with software libraries and standards, a different level of encryption infrastructure from protocols to algorithms to keys, key exchange protocols. The whole encryption infrastructure has been, in a way, redesigned.

[Related: Everything Happens in a Prompt Box]

Jason Lopez: But installing those fixes isn’t as simple as pushing an update.

Marco Graziano: The transition is not going to be painless, because you’re touching the core of your systems. You’re touching a very important component, which you cannot break, but it can break. And you need to manage legacy and new, and be able to switch between the two.

Jason Lopez: And Graziano says patience for a slow transition won’t last forever.

Marco Graziano: One day, organizations will decide to reject communication with endpoints that are not quantum resistant, because at some point the risk is going to be so high that you cannot take any chance.

Jason Lopez: Mitigating that risk opens up an innovation path we can barely predict.

Steve McDowell: If I’m doing drug research, I’m doing seismic processing, oil and gas, anything that relates to the physical world is going to revolutionize, you know, weather and climate modeling, just because of the nature of how it operates. It’s going to disrupt a lot of industries, and we’re going to find use cases that we’re not thinking of now, right? Those are the low-hanging fruit.

Jason Lopez: The voices you heard in this podcast were of Marco Graziano, a Silicon Valley-based computer scientist, Steve McDowell, analyst at NAND Research, and Don Gentile at HyperFrame Research. This is the Tech Barometer podcast, I’m Jason Lopez. Tech Barometer is produced by The Forecast. For more insights into technology and the people in the tech industry, you can find more stories at theforecastbynutanix.com.