I am going to try really hard to make this article not overly negative. The urge I have to start this one off with something like “well the 2020’s couldn’t possibly get worse” or “tech companies are doomed” is very strong. For real, who had DDR5 kits increasing over 250% on their bingo card? Not me, that’s for sure. And the annoying part is that this isn’t just a gamer problem, it’s an anything-with-memory-chips-in-it problem. I guess this just stings a little worse for me because I was planning on upgrading my personal gaming rig in 2026. But I digress, this article is for work and not my hobbies.
I am writing this not to be negative or just complain, I want to help educate and prepare others where I can. Plus, I really do think there is some silver lining here. If we plan for it instead of getting surprised by it, we can make smarter system choices, give better guidance to customers, and avoid some quote revisions.
Quick note: I’m using “RAM” as shorthand here, but I’m also talking about the broader memory market, DRAM and NAND included.
Why is RAM expensive now?
If you didn’t already know, you could probably guess. The big elephant in the room of 2025 is the culprit, AI. The simple way to explain it is this: AI is hogging all the memory.
On the manufacturing side, there’s a big industry push toward higher-margin memory products that feed AI servers, especially high-bandwidth memory and server DRAM. When capacity and advanced process nodes get prioritized for that world, the DDR5 RAM that is used in modern PC suffers, and so do every other piece of technology that uses a memory chip.
On the pricing side, TrendForce is basically saying the bulk, behind-the-scenes pricing for DRAM could jump 55 to 60% in Q1 2026 vs Q4 2025, and server-grade memory could be even worse, over 60%. And it’s not just DRAM, TrendForce also expects NAND flash pricing to climb hard too, which matters because SSDs and flash storage are everywhere in our world now.
Consumer pricing has followed the same general storyline, even if it is messy and model specific. Tom’s Hardware tracked examples where mainstream DDR5 kits went from normal pricing to extreme pricing through late 2025, and while some kits are showing signs of leveling off recently, the new “level” is still inflated.
Prices will go up
This matters for broadcast and AV because memory is not just a PC part. Sure, production workstations and servers get the most obvious price hikes, because you’re literally buying DIMMs. But even when you buy a piece of dedicated hardware, a switcher, encoder, decoder, camera, intercom system, LED processor, multiviewer, control box, whatever, there’s still a very good chance it has DRAM in it somewhere. It might be DDR4, LPDDR, soldered memory, a buffer next to an FPGA, or memory hanging off a SoC. You just don’t see it because it’s buried inside the product.
So, if the whole memory market gets more expensive, it’s not only PC builds that get hit. You get a little memory tax across the board. Some manufacturers will eat it for a while, some will quietly adjust pricing, and some will change configurations, like less memory than they used to ship with, different storage sizes, or different models getting prioritized first.
That’s why I started with DDR5, it’s the most visible example, but the real story is broader: if DRAM and NAND climb, the cost pressure shows up in a lot more places than just a workstation quote.
What to do about it
If you are building systems, selling systems, or advising customers, don’t panic. Be deliberate.
The first move is treating memory as a real cost driver. That includes the obvious stuff like system RAM in workstations and servers, but it also includes the less obvious memory that’s buried inside gear, the DRAM in an encoder, the flash in a control box, the storage in a replay unit, and so on. You don’t always get to pick those parts directly, but their pricing still influences what manufacturers charge and what stays in stock.
The next piece is sizing things right, and I mean that in a practical way. I am not saying to buy less RAM and hope. I’m saying measure what the workload does, then build with enough headroom. Plenty of systems are overbuilt because we default to maxing everything out, and that habit hurts more when memory is the expensive part. At the same time, underbuilding is worse, because memory pressure issues always show up at the worst possible moment, which is usually when you’re live.
The silver lining
Now for a little sunshine through the storm. (featuring my in-laws cat)
This pricing mess might push a small, healthy correction in workflow design. Not a retreat from IP, just a reminder that new is not always best for every job.
Baseband is good ol faithful. SDI and HDMI are straightforward, stable, and easy to troubleshoot. They do not care about memory pricing, they do not care about driver updates, and they do not randomly start dropping frames because a background service decided it needed 12GB of RAM today.
I am not saying we all go back to 2012 and pretend IP never happened. I am saying the best-looking systems in 2026 might be hybrid on purpose. Baseband where reliability and simplicity matter most, IP where flexibility and scale actually pay off.
If memory prices stay high through 2026, which plenty of reporting suggests is at least possible, I would expect more people to land right there in the middle.
Baaaaaa (RAM)
If you take nothing else from this, take this: memory pricing is now a planning problem, not just a purchasing problem. The teams that budget early, spec carefully, and build with intention are going to have a much easier year than the teams that treat memory like a checkbox at the end.
If you want help checking a build, or you are trying to decide when to go IP-heavy versus when to keep things baseband, reach out to our team. That’s what we do.
