The Old Assumption Is Breaking
For years, I approached personal technology with a predictable assumption: computers would get faster, storage would get larger, memory would get cheaper, and yesterday’s premium specifications would eventually become mainstream. Time generally worked in the consumer’s favor.
That assumption has become harder to make. Memory is expensive, storage costs are under pressure, and PCs are not necessarily becoming cheaper in the way many buyers expect. I can see the effect in my own habits: I am keeping technology longer. It is not because I stopped enjoying new hardware or because innovation became less interesting. Replacing functional equipment is simply becoming more difficult to justify.
Part of that story leads back to the enormous infrastructure buildout around artificial intelligence. AI is reshaping software and business, but one of its less glamorous effects is appearing much closer to home, in the price and availability of the technology the rest of us buy.
AI Lives in Hardware
We often talk about AI as though it were primarily software. We open Copilot, ChatGPT, or another service, type into a box, and an answer appears from somewhere in “the cloud.” But the cloud has always been physical. AI requires data centers packed with accelerators, servers, networking equipment, memory, and enormous amounts of storage.
The especially important technology is High Bandwidth Memory, or HBM. AI accelerators need tremendous memory bandwidth, while AI servers also consume conventional server memory and enterprise storage. Much of that demand depends on the same semiconductor manufacturing ecosystem that ultimately supplies PCs, SSDs, phones, and other everyday devices.
Manufacturers naturally want to serve a fast-growing, high-value market. Cloud providers, meanwhile, want to guarantee that hardware will be available for years of data-center expansion. When both groups act accordingly, production capacity and investment move toward AI and server products. That changes the economics of memory production, even for buyers who will never purchase an AI accelerator or operate a data center.
The Numbers Are Difficult to Ignore
This is more than anecdotal evidence from someone browsing upgrade prices. At the beginning of 2026, TrendForce reported that persistent AI and data-center demand had worsened the global memory supply-demand imbalance. Its forecast called for conventional DRAM contract prices to increase 90–95% quarter-over-quarter in Q1 2026, NAND Flash contract prices to rise 55–60%, and PC DRAM to increase by more than 100%.
The pressure did not disappear after the first quarter. By September, TrendForce projected another 10–15% quarter-over-quarter increase in conventional DRAM contract prices for Q4 2026 and another 15–20% increase for NAND Flash. It also said PC manufacturers were continuing to buy aggressively because they expected DRAM supply to remain tight in 2027.
J.P. Morgan Global Research estimates that DRAM prices will have risen more than 400% between the beginning of 2024 and the end of 2026. Just as important, hyperscalers are securing output through long-term agreements, sometimes extending five years or more. They are not shopping for memory the way an individual buys a DIMM. They are planning infrastructure years ahead and have the financial strength to reserve capacity accordingly.
| Market pressure | Where it appears | Practical effect |
|---|---|---|
| HBM and server-memory demand | AI accelerators and data-center servers | Manufacturing investment and advanced capacity favor high-value infrastructure products. |
| Conventional DRAM constraints | PCs, laptops, workstations, and servers | System builders and buyers face higher component costs and less predictable pricing. |
| NAND Flash pricing | SSDs and other storage devices | Storage upgrades and higher-capacity configurations become harder to treat as inexpensive additions. |
| Long-term hyperscaler agreements | Future production allocation | Large buyers can reserve capacity years ahead, leaving other market segments competing for supply. |
The Cost Moves Downstream
It would be tempting to call this only a RAM problem, but the impact is broader. DRAM affects PCs, servers, and workstations. NAND Flash affects SSDs and other storage. HBM and high-capacity server memory affect AI infrastructure. If manufacturers can earn greater returns from advanced server products, production priorities understandably shift in that direction.
TrendForce reported that DRAM suppliers continued prioritizing advanced-process capacity for high-performance server products in Q4 2026, while warning that PC DRAM supply could decline in 2027 as more capacity moved toward server applications. That decision eventually reaches ordinary buyers. A laptop customer does not need to understand an HBM stack; the customer simply sees a higher price. A small business may not follow DRAM contracts, but it notices that a desktop refresh quote is unexpectedly expensive. A PC enthusiast may not care which hyperscaler signed which agreement, but does care that memory and SSDs are no longer the inexpensive upgrades they once were.
AI is not the sole cause of every increase. Tariffs, currency changes, manufacturing costs, transportation, product positioning, redesigns, vendor margins, and normal semiconductor cycles all matter. Still, memory is one area where the connection between AI infrastructure demand and broader technology costs is increasingly difficult to dismiss.
I Can See It in My Own House
The market becomes personal when I look at the equipment I already own. I have two Dell Micro PCs that are about six years old. Traditionally, hardware of that age would have me thinking seriously about replacement. Yet comparable systems now approach roughly double what I originally paid. I also have a Mac mini, and a comparable replacement today is about 40% more expensive than what I paid for mine.
Those are not abstract semiconductor statistics. They change the decision. If a Dell still performs its job reliably, what am I really gaining by replacing it? If the Mac mini continues handling the workload I bought it for, is a newer processor worth a substantially higher price merely because a newer model exists? Increasingly, my answer is no.
The same realization shaped my recent thinking about a $299 Apple TV 4K. I still like Apple TV and prefer the experience, but my existing devices will remain connected until they fail, stop meeting my needs, or a future capability becomes compelling enough to justify replacement. The key question is no longer simply, “How much faster is it?” It is “What can the new device do that my existing device cannot?”
From Upgrade Cycle to Replacement Cycle
For a long time, an upgrade and a replacement were different decisions. An upgrade happened because something newer was attractive; a replacement happened because the old device no longer did its job. Cheap memory and storage blurred that line. When components were inexpensive, improving a machine simply because you wanted something faster did not always feel extravagant.
Higher prices restore the distinction. My threshold is increasingly straightforward: Does this device still reliably do what I bought it to do? If the answer is yes, the replacement conversation may be over. Maybe the six-year-old Dell stays. Maybe the Mac mini stays. Maybe the older SSD remains adequate. Maybe the PC gets another year or two.
There is an irony here. AI is accelerating technological progress, yet the infrastructure investment required to build it may lead consumers and businesses to slow their hardware replacement cycles. Standing still does not necessarily mean falling behind. A six-year-old PC serving a lightweight workload does not stop being useful because a newer processor exists. A functioning SSD does not care what the latest model can benchmark. A streaming device does not stop playing video because its replacement has a faster chip. Sometimes the most economical upgrade is no upgrade at all.
Useful Life Matters Again
Higher prices may force healthier discipline into technology purchasing. Instead of automatically assuming a three- or four-year consumer PC cycle, perhaps a system remains in service for five, six, or seven years when its workload and software support allow it. Rather than replacing the entire machine, we might replace a battery, add storage, or upgrade memory where the design permits it.
Business decisions are more complicated. Security support, warranties, operating-system requirements, reliability, and standardized configurations all matter. Keeping unsupported or insecure hardware indefinitely would be false economy. Performance also matters when the workload requires it. But none of those concerns require replacing a supported, reliable device simply because something newer exists. Rising acquisition costs make thoughtful lifecycle planning more important, not less.
There is a sustainability benefit as well. Manufacturing electronics consumes materials, energy, transportation, and capital. If higher costs persuade us to use a functional PC for six years instead of four, we extract more value from something already manufactured. The goal is not to resist necessary upgrades; it is to distinguish necessity from habit and get the full useful life from the technology we already paid for.
How I Am Changing the Way I Buy Technology
My approach today is more conservative than it was a few years ago, but it is also more intentional.
| Principle | What it means in practice |
|---|---|
| Buy for longevity | Purchase enough memory and storage to leave room for future needs, especially when neither can be upgraded later. |
| Separate want from need | Require a new feature to materially improve the way the device is actually used. |
| Maintain good hardware longer | Keep systems that remain fast enough, secure, supported, and reliable. |
| Value repairability | Give more weight to replaceable storage, memory, and batteries when complete systems are expensive. |
| Measure total useful life | Compare cost over the full service period, not just the initial purchase price. |
A $1,000 computer used for six years may be a better purchase than a $700 machine replaced after three. The cheapest device and the least expensive technology strategy are not necessarily the same thing.
The New Technology Economics
For years, technological progress encouraged replacement. Faster processors, larger drives, and falling component prices made the next purchase relatively easy to rationalize. The current economics increasingly encourage longevity instead.
That is certainly what has happened to me. I look at my six-year-old Dell Micro PCs, my Mac mini, and my existing Apple TVs differently. I do not see old technology merely waiting to be replaced. I see paid-for technology that still does its job. That is an important distinction.
If the AI boom continues consuming enormous amounts of memory, storage, and semiconductor capacity, some prices may remain elevated and some supply constraints may persist. Buyers will need to be more selective, plan longer, and demand clearer value from each upgrade. The smartest purchase over the next few years may sometimes be the one we do not make.