The AI PC Episode: The Real Cost of Waiting

Almost every machine you can buy today is technically an “AI PC.” Which is exactly why the label can’t tell you what to buy. This episode skips the hype and asks the question IT leaders are actually wrestling with: why now, and what happens if you wait? Because two things just landed on the same calendar — the end of Windows 10 support, and an AI-driven memory shortage pushing hardware prices up every quarter — and together they’re quietly taking the refresh decision out of your hands.

In this episode:

• Why “AI PC” has stopped being a useful filter — and what to look at instead
• The hidden tax you’re already paying on the machines that still turn on
• How Windows 10’s end of support and the memory squeeze changed the math on waiting
• A practical way to make the refresh call: match the machine to the workload, not the org chart

Featuring Ivo van Selst, Senior Technical Solutions Architect for AI and Data at Softchoice, a World Wide Technology Company, and Brandon Boone of Dell Technologies’ Client Solutions Group. 

This episode is brought to you by Dell Technologies. 

Explore Dell’s AI PC lineup — Dell, Dell Pro and Dell Pro Max — and find the right fit for your fleet at softchoice.com/technology-partners/dell

#AIPC #ITLeadership #DeviceRefresh #Windows11 #MidMarketIT #EndUserComputing #TheCatalyst

Show Notes & Resources

Referenced in this episode:

• Windows 10 end of support (October 14, 2025) — Microsoft: support.microsoft.com/en-us/windows/windows-10-support-has-ended-on-october-14-2025
• The global memory shortage and its impact on device costs — IDC: idc.com/resource-center/blog/global-memory-shortage-crisis
• DRAM prices and the AI memory squeeze — IEEE Spectrum: spectrum.ieee.org/dram-shortage
• AI-capable PCs to surpass half of global shipments in 2026 — Counterpoint Research: counterpointresearch.com/en/reports/ai-advanced-pcs-to-surpass-half-of-global-shipments-in-2026
• The productivity cost of aging PCs — Intel research (PCs 4+ years old lose ~21 hours per year)
• Dell’s AI PC portfolio — Dell, Dell Pro, Dell Pro Max: dell.com/en-us/blog/dell-transforms-ai-pc-portfolio-for-anywhere-productivity
• Workplace lifecycle services — Softchoice: softchoice.com/solutions/modern-workplace/managed-services/lifecycle-services

Guests:

• Ivo van Selst, Senior Technical Solutions Architect for AI and Data, Softchoice — linkedin.com/in/ivo-van-selst-4579a96
• Brandon Boone, Dell Technologies, Client Solutions Group — dell.com

Key moments:

• Cold open — the hidden tax: the laptop nobody complains about
• Act One — what an AI PC actually is, and why the label misleads
• Act Two — the false economy of deferring a refresh
• Act Three — Windows 10, the memory squeeze, and losing control of your timeline
• Act Four — match, don’t guess: segmenting your fleet by workload

The Catalyst by Softchoice is the podcast dedicated to exploring the intersection of humans and technology. 

Transcript

Katey 0:01

There's a laptop somewhere in your building that takes four minutes to wake up. Somebody's using it right now. And they've stopped noticing the wait. They get up, they get coffee, they come back, and maybe by the time they sit down, it's ready. It boots. It opens the same three apps it always opens. On a spreadsheet, it looks fine. It's paid for. Nobody's complaining. And that's the problem.

SPEAKER_02 0:32

I think businesses generally underinvest in their hardware solutions for their employees. So if you lose 10 minutes a day and the device has a three-year lifespan, that's a lot of cost. There's a total false economy around that.

Katey 0:47

A false economy. Money you think you're saving that you're actually spending, just somewhere you're not looking. This is an episode about a decision almost every IT leader will face in the next year, whether they've budgeted for it or not. It's dressed up in one of the hottest new buzzwords in tech: the AI PC. You've heard it. You might have even tuned it out. But stay with us anyways, because underneath the buzz are real questions with real deadlines. What an AI PC actually is? Why waiting to buy one may cost more than actually buying it? And how to make the call? By fleet on a mid-market budget before the supply chain makes it for you. From SoftChoice, a worldwide technology company, this is The Catalyst. I'm Katie Tikasing. Today, the machine on your desk is quietly becoming an AI machine. And choosing the right one is trickier than the label makes it look. We're calling it the AI PC episode. Act one, the label that means nothing. Let's start with what actually changed inside the box, because it's smaller and stranger than the marketing makes it sound. For most of the PC era, your computer had two engines. Here's Brandon Boone, a subject matter expert at Dell Technologies. He works on the client solutions group, and here's what he sees.

SPEAKER_01 2:32

We went from a PC having a CPU and a GPU to now getting a system with a dedicated NPU, which is a neural processing unit. So the PC went from a two-engine machine to really a three-engine machine now.

Katey 2:45

A third engine, the NPU, a neural processing unit. A chip whose whole job is to run AI directly on the device instead of shipping it to the cloud and waiting for it to come back. That's the actual hardware story. And here's where the label gets confusing. Microsoft has a name for a machine that clears a certain bar, a Copilot Plus PC. To earn the label, a machine needs an NPU that can do at least 40 tops, 40 trillion operations per second, plus 16 gigs of memory and a 256 gig drive. The 40 tops number is a gate for unlocking a set of Windows features: live captions, background blur, a search tool called recall. But it's not a measure of how much a real AI machine can run. For comparison, a high-end graphics card does more than 1300 tops. So the number on the box tells you a machine crossed a line. It just doesn't tell you how far past the line it actually is. Which is why the person on the sales side will tell you the label is close to useless. And it's about to be on everything. Analysts at Counterpoint Research expect AI-capable PCs to go from around 39% of shipments in 2025 to roughly 59% in 2026, the majority of everything sold. The register put it more bluntly. Before long, you won't be able to buy a computer that isn't AI enabled. So here's the strange place we've landed. This is Evo Van Selt, a senior technical solutions architect for AI and data at SoftChoice. His whole job is translating this stuff for the people who have to buy it. For a lot of your fleet, it's already here. Nobody announced it. It just arrived. Quietly in your last refresh. But here's the catch. Technically, an AI PC and the right machine for your team in the next four years are not the same thing. One is a label almost everything qualifies for now. The other is a decision. And the label can't make it for you. So if the name on the box won't decide it, and it won't, something else has to. And it turns out the thing that should drive the call isn't on the spec sheet at all. It's on the balance sheet. And it's the cost of the machine you already have. Act 2, the hidden tax. Refreshing a fleet of laptops is nobody's favorite line item. It's expensive, it's disruptive, and the machines you'd be replacing still turn on. So it's the easiest thing in the budget to push for one more year. And on paper, pushing it works. You don't spend the money. The savings are real and they do show up immediately. Brandon has watched a lot of companies do exactly this, and he's watched what happens next.

SPEAKER_01 6:36

Performance drag just becomes normal at that point. So your employees are losing time and productivity every day.

Katey 6:42

Performance drag becomes normal. That's the quiet part. Nobody files a ticket that says, my laptop is three seconds slower than it should be 40 times a day. They just adapt. They wait, they work around it, and the cost of all that waiting never shows up as a cost. It shows up as a slightly slower company.

SPEAKER_01 7:05

You are losing money behind the scenes that maybe you're just not seeing on the front end.

Katey 7:10

Now, a salesperson at a hardware company saying, Your old hardware is costing you, is exactly what you'd expect a salesperson at a hardware company to say. So don't take it from Dell. Take it from the numbers. According to research from Intel, a business PC that's four years old and older loses about 21 hours of productivity a year and breaks about one and a half times as often as a newer machine. 21 hours. That's most of a work week per person per year evaporating into spinning wheels. Evo puts it this way.

SPEAKER_02 7:58

So if you lose 10 minutes a day and the device has a three-year lifespan, that's a lot of cost. There's a total false economy around that.

Katey 8:08

10 minutes a day. It sounds like nothing. Over the three-year life of a machine across the team, it's the false economy Evo's talking about. You save $50 on the front end and pay for it slowly in the one resource you can't buy back. Your people's time. Now here's the thing. This math has been true for years. Slow computers have always cost more than they look like they're going to cost. So why is it that this is the year it stops being something you can safely ignore? Because this year, two other things are happening at the exact same time, making the stakes much, much higher. Act 3. The squeeze. Two forces closed on IT leaders at once. The first one has a date on it. On October 14th, 2025, Microsoft ended support for Windows 10. Not recommended you upgrade it ended. No more security updates. And a lot of older machines simply can't move to Windows 11. They're missing the security hardware that Windows 11 requires. Which means for those devices, the upgrade isn't an upgrade. It's a replacement. That turned a soft, we should refresh eventually, into a hard deadline. And it landed squarely on our people. Analysts note that for small and mid-sized organizations where cost drives the timing, this migration will stretch into 2027. In other words, it's happening right now and it's happening to you. That's force one. Force two is why the timing is cruel. And Evo has a name for it.

SPEAKER_02 10:04

So GP workloads, which is AI, runs in video RAM. And that's been the whole RAM apocalypse.

Katey 10:12

The RAM apocalypse. Here's what it means. The same AI boom driving demand for these machines is devouring the memory that goes inside them. The world's memory makers have pointed their factories at the high-end chips that feed AI data centers. Because that's where the margins are, and left everyone else fighting over what's left. The scale of that shift is hard to overstate. According to IDC, data centers are expected to consume around 70% of the world's memory chips in 2026, up from 20 to 30% just a few years ago. And when supply gets that tight, prices move. Counterpoint research clocked memory prices climbing 80 to 90% in a single quarter. IDC expects the cost of PCs and phones to rise 10 to 20% by the end of 2026. And to hit the cheaper machines the hardest. Brandit sees it from inside the supply chain. And his warning isn't really about price, it's about who is in charge of your timeline. Wait too long, and the question stops being, what do we want and when? And it becomes, what can we get and what will it cost? The factories are the constraint now, not the budget.

SPEAKER_01 11:55

Most of those factories and the fabs are already sold out for this year and next year, right? This is going to go through 2028, which is two years from now, right?

Katey 12:04

And building more capacity doesn't happen fast. Evo, who is more optimistic than most about how this ends, is quite blunt about how long it's going to take to get there.

SPEAKER_02 12:16

It takes years to build a fab, right? And so this friction will consist for a long time. But eventually things will normalize.

Katey 12:26

Eventually, which brings us to the most honest part of this episode. Maybe you're not convinced you need an AI PC. Maybe you think the term is mostly marketing. And that's a fair place to stand. And you don't have to settle it to see the problem. Because the reality doesn't care what you call it. Windows 10 is done. Memory is getting scarcer and more expensive by the quarter. And the performance drag that we talked about is already on the clock. Whatever name you put on the machine that answers all three, the point is the same. A new kind of device is becoming necessary sooner than you might have expected. So the question isn't whether to move, it's how to move without making an expensive mistake. Act four. When a refresh finally becomes unavoidable, the instinct is to standardize. Pick a machine, buy a thousand of them, and you're done. It feels efficient. It's how you make an expensive mistake at scale. Brandon's advice for the person listening to this on a Monday morning is to start somewhere completely different.

SPEAKER_01 13:45

Don't start by asking, you know, what laptop to buy necessarily. Start by segmenting your workforce by workload. And once you do that, the refresh kind of stops becoming abstract and starts becoming a matching exercise to those users.

Katey 13:59

A matching exercise. Not what should we buy? But who is this for? The person who lives in email and Teams and a browser needs a totally different machine than the person who's building models. And the mistake is buying one machine for both in either direction. You overspend on the first and you strangle the second. And matching isn't only about the box. Especially now. With people splitting the week between home and the office, the laptop is just one piece of the setup.

SPEAKER_01 14:38

That's the mistake some companies make is just refreshing the device itself, but not all the peripherals that go with it, right? And you you know you buy this great computer, but then you're underinvesting in the workspace around it.

Katey 14:51

The camera nobody can quite hear you through. The one cramped screen, the dock that drops connection mid-call. Refresh the laptop and ignore the desk around it. You've solved half the problem and left the half your team actually feels. And whatever you match them to, buy for the life of the device, not for the app running today.

SPEAKER_01 15:17

You can't just buy for the applications today, right? Because of how quickly AI is moving.

Katey 15:23

Buy for today only, and Brandon has seen exactly how this story ends.

SPEAKER_01 15:30

Two years down the line with all these new applications, they came back to us like, we made a mistake, our our users are unhappy.

Katey 15:37

We made a mistake. They bought for the moment, the moment moved, and they paid twice. Once for the wrong machines, and again to replace them early. The whole point of the exercise is not to have that conversation, which is where Evo pushes back on the idea that there's a single right answer for everyone.

SPEAKER_02 15:59

There's not a right decision and a wrong decision from a technology perspective. It's more about how the technology applies to your business.

Katey 16:06

How it applies to your business, not the vendor's roadmap, yours. What your people actually do, what your data rules actually require, what you can actually afford this year. The refresh isn't a spec contest, it's a fit problem. And the cost of getting the fit wrong isn't a bad quarter. It's years of drag that you've signed yourself up for.

SPEAKER_02 16:35

People don't understand the implications of the steps that they're taking today and the chaos that it can lead them to in the future.

Katey 16:48

So here's where we've landed. The name on the box won't make the decision for you. Nearly every machine is an AI PC now. So the label alone can't tell you which one actually fits your team. The old laptop still turns on isn't free. It's charging you a hidden tax in your team's time. This is the year the tax comes due because a Windows deadline and a memory squeeze landed in the same calendar year. And the way through isn't to standardize and hope. It's to match the machine to the person and buy for the life of the device before the supply chain decides your timeline for you. Evo puts the whole thing in one line.

SPEAKER_02 17:36

When you have questions like this, I think getting good advice matters.

Katey 17:46

This episode of The Catalyst was made with support from Dell Technologies. Thanks to the team at Dell for helping us tell it. If everything you've heard has you looking at your own fleet a little differently, here's where to start. Dell has built its AI PCs into three simple tiers: Dell, Dell Pro, and Dell Pro Max. So the question stops being which laptop and starts being which user. Each one runs AI on the device through a dedicated NPU, a longer battery life, sensitive data that stays on your machine, and hardware that's ready for whatever comes next. You don't have to figure out the match alone. SoftChoice, a worldwide technology company, works with organizations to size the right devices to the right people and on a mid-market budget. Visit softchoice.com forward slash technology-partners, forward slash Dell. Or click the link in the description to start the conversation. The catalyst was reported and produced by Tobin Dalrymple and the team at Pilgrim Content. Editing by Ryan Clark with support from Philippe Dimas, Joseph Bayer, and the marketing team at SoftChoice. Thank you to our guests, Evo Van Selt and Brandon Boone. And to our sponsor, Dell Technologies. This is the Catalyst.