[ This article is part of a series where I invited media industry leaders to share their thoughts on 2025 and what it means for the future. I lightly edited this for clarity. Larry O’Connor is the CEO of OWC.]
NAND / SSD and DRAM
Background: To understand Larry’s comments, you need to know there are two broad categories of RAM: low power (LPDDR) typically used in computers and high-power (HBM) typically used in servers and AI data centers. A production line creating RAM can make one or the other, but not both on the same production line. Given a finite number of production lines, manufacturers need to decide how many lines to devote to each type. That decision is generally made on which makes the most money.
DRAM and memory
2025 started with supply / demand that seemed in balance. DRAM was pretty stable, for those still with systems or backend servers where memory upgrades are still a thing. Starting in late May, that all blew up. Demand for the HBM (High Bandwidth Memory) memory needed for GPUs [used for AI and servers] was surging with GPU output and demand. We saw some of the first crazy jumps in memory costs in the spring, and they just kept coming. Capacity that was supporting the computer memory side was quickly being moved to meet [HBM] GPU demand, with a bump in demand for the memory also needed to be in-slot for those AI servers these HBM GPUs went into.
Result? Certain memory types are in short supply today due to much more limited production. In particular, LPDDR4 is impacted which is still widely needed for PCs and some older Macs of the past five years. It’s substantially scarcer than it ought to be. DDR5, used in modern PCs and servers, has more production support and it’s a necessity for the servers that these GPUs go into, but it’s a real challenge for all the DRAM makers to make enough due to the higher margin recognized on GDDR6 and GDDR7 HBM for the aforementioned GPUs. The challenge is that there can’t be GPU sales without HBM memory to support the servers those GPUs go into.
This supply / demand situation has resulted in memory costs being 4X to 10X higher than it was just six months ago. We’ve had these pricing blips before due to natural and unnatural disasters with a relatively short term impact and a fast return to normal. This time, though, feels like a train rolling ever faster with no signs at present of anything to bring costs down.
SSD and NAND
Background: NAND is a type of non-volatile memory, meaning it can retain its data when the device is powered off. “NAND” stands for “Not AND,” which describes how the storage works. Best if you don’t ask.
While the pricing of our higher-end SSD gear started be impacted from higher DRAM costs this past summer (such drives have 2GB to 8GB of memory onboard), SSD costs were pretty stable for most of the Spring and Summer, until we got to September. There was already very strong demand for AI Server and Data Center expansions, but it seemed like FOMO (Fear Of Missing Out) fever took hold as projections for AI demand started to look absolutely insane and beyond any satiation. Now costs are 60-400% higher vs. just 3 months ago. The retail market is currently behind in this pricing increase – but it’s going to quickly catch up. This is probably one of the first holiday seasons where SSD ‘deals’ during Black Friday and CyberMonday were actually priced higher than a month before.
If you hope that prices are coming down after the new year, it’s not going to happen. Prices will be higher come January – mark my words. The demand is just off the charts as a race for data center support is buying what seems like anything, everything, and at whatever price. I personally don’t like the current math there; especially considering the much higher cost to deploy the same compute/GPU capability in data centers at today’s costs vs. 6-9 months ago. But what do I know?
I believe prices may cool at some point – there isn’t enough power available (from data I’ve seen) in the next five years to meet the power demand for all the new data centers coming on line in the next three years.
For the Future
With that flow of information, supply is a major issue that’s likely to keep DRAM and SSD pricing up for at least the next 6-9 months. So much so, that we might be wiser to use the storage we have for up-coming projects. This crazy data center explosion isn’t just impacting us regular people’s power bills (that’s happening right now today too), it’s also threatening supplies used in automative and other consumer electronics (microwaves, refrigerators, all these smart devices that need EMMC and DRAM too).
Oh! I almost forgot Crucial. Micron, Crucial’s parent company, is a solid third in world market shares for DRAM and a little lower for NAND. But it was the top brand in the consumer channel when looking at major retailers like Best Buy, Amazon, etc.
If you don’t think this AI demand on NAND/SSD and DRAM/Memory forcing prices higher is real, Micron just made it worse. While Crucial has had questionable pricing strategies since inception vs. margins its parent company demanded from integrators and enterprise – the current market conditions were enough for Micron’s board to announce an abrupt end the Crucial consumer division a few weeks ago. Crucial is in full wind-down mode, completed by next by February.
Now that Micron won’t be producing RAM for Crucial, it can focus its capacity on HBM RAM. When I talk to different channel partners, the exit of Crucial is one of the most significant they have to contend with as we go into 2026.
That missing RAM will be a significant missing piece.
VIDEO
Video is still of huge importance and a driver, I look at camera technology and that is driving the camera media / CFxpress we manufacture to new heights.
OWC brought out top performing and top reliability camera media over the last 12-16 months and we’re not going to stop. This faster media has also required faster media readers. We now have hardware that can keep up with all this new tech and that tech is saving us so much time with faster ingest, faster duplication, and faster processing. Saving time time means more budget can be used to ensure the best creative minds are getting that content where it needs to be when it needs to be there/
I’d argue that while Thunderbolt 4 hardware is plenty fast, even a plain M4 or M5, or an earlier M-series – but the higher end Thunderbolt 5 systems pack more punch to enable other tool optionality + workflows that only Thunderbolt 5 can make so simple and fast.
Today’s M-Series Macs are so stupidly fast – that we have hardware today that’s ahead of what 99.5% of pro creators can get benefit from. It’s a golden age in some respects; you don’t need to wait for a faster system, instead you can put dollars into improving workflow or even your own marketing/visibility, etc. It’s been more than a decade since Apple really was pushing the envelope for creators specifically, but the technology available today is beyond what we might have dreamed about five years ago. The computer is not the choke point so much at all today.
2026 is, for OWC, more about reliability, connections, and storage. More of the same from us – but with more ports and/or more speed. Higher capacities. Ongoing reliability and ruggedness improvements from setups for a single creator to our Jellyfish that supports small teams to entire production studios of hundreds. The feedback you give us as industry experts, users, our customers, drives us in all areas. O
ur goal is to ensure not just the best hardware, but the best accompanying software that’s under-the-hood driving the total experience forward. The unspoken hero behind our great hardware is the software that OWC creates – it elevates our hardware beyond the ordinary. Enhancements and improvements that are ongoing will continue to bring even more benefit to our customers.
We see such a deluge of knockoffs and wanna be’s out there. They are not OWC. It’s the unseen efforts, engineering, and software that sets us a world apart. I was saddened to see Roomba fail. They were the consumer electronics robotics innovator. they lost because of substandard copycat/knockoff crap that ruined their margins and even perception and expectation. Consumers can be easily swayed and the result is a lot of consumer complaints on those Roomba copycats – and the death of the innovator.
OWC builds solutions that work and are designed by a team that cares about the customer we all ultimately serve. We’re fortunate that our customers can see and recognize the differences and know the OWC difference. I am grateful that we have this opportunity and a discerning customer that understands that a look-a-like doesn’t mean it is alike. We’re always here to earn trust and stand above.
AI is a real threat to all of us. It’s threatening the supply chain right now, but I expect that to settle down in the next six-to-nine months as certain unavoidable realities (like access to power) set in.
[One financial guru noted: When companies with multi-trillion dollar valuations ‘let’ private equity get on the hook for 10 billion dollars (and more) to build data centers with crazy favorable terms to the AI companies vs. the ones spending the cash, someone clearly knows the ROI risk/benefit and is happy to let the FOMO financial crowd compete to toss their dollars out there.]
Anyway; we will survive and thrive not by fearing the capabilities of these new AI tools, but by embracing and elevating our own capabilities with them. Whether we are talking hardware knockoffs, or creative projects, the ultimate customers want and will pay for the quality that only real talent brings to life.
