Silicon Motion says retail SSDs are nearly gone as OEMs divert NAND to AI servers
Nelson Duann links the retail slump to OEM NAND shortages and explains why module makers are ending up in PC OEM drives.

Silicon Motion VP Nelson Duann says the retail SSD market has almost disappeared, with controller shipments increasingly landing in SSDs for PC OEMs. For decision-makers, the bottleneck is shifting where inventory and pricing pressure show up, not just how fast storage gets cheaper.
Silicon Motion VP Nelson Duann says the retail SSD market has “almost disappeared.” And he ties it to a specific supply chain reality: the controllers Silicon Motion sells to module makers are “now largely ending up in SSDs that are shipped to PC OEMs,” because PC OEMs cannot obtain enough NAND directly from memory manufacturers.
Duann also says this shift has been happening “since late last year and into this year.” Translation for anyone shopping for storage: module maker SSDs that would normally reach the retail channel for gaming PCs are increasingly being pulled into OEM builds. That means the inventory pressure ends up where consumers actually experience it, with retailers facing tighter availability and higher prices at the same time.
To unpack why this matters, you have to understand how SSDs get from chips to your PC. NAND is produced by memory makers. OEMs and system builders then need NAND, plus controllers and other components, to assemble storage. Traditionally, OEMs have been able to bulk-buy NAND directly from memory manufacturers and build SSDs (or work through supply paths that effectively source components at scale). But Duann’s explanation flips that flow: when OEMs cannot obtain enough NAND directly, they increasingly source SSDs from module makers instead. Module makers, in turn, get Silicon Motion controllers and assemble drives.
So the “retail SSD market” is not only a pricing story. It is a channel allocation story. If OEMs take more of the module makers’ output, the portion left for retailers shrinks. And Duann’s phrasing suggests it is not a modest rebalancing. He is describing a situation where the retail market has become so small that it is, in practical terms, nearly gone.
Why are OEMs suddenly more constrained? Duann links the change to “increased demand for AI servers.” AI workloads are data-hungry, and servers are often the highest-priority buyers in a supply scramble. When AI servers absorb NAND and related components, other buyers compete for what remains. Even if NAND supply exists somewhere in the broader ecosystem, OEMs may still struggle to get it on terms that are workable for their production schedules. The result is a sourcing workaround: buying completed SSDs from module makers rather than attempting to secure enough NAND directly.
This is where the story gets uncomfortable for retail shoppers, but also strategically relevant for executives. Duann’s comments imply that supply chain pressure can move like a pinball. Retail shelves do not just reflect global production. They reflect allocation decisions, contracting, and how much of the module maker pipeline is captured by OEM procurement. If that pipeline is increasingly booked for OEM shipments, retail availability becomes a downstream casualty even when upstream NAND makers are active.
The source notes that “NAND makers are doing better than they ever have done” and that SSD companies are “purchasing memory from them years in advance.” That combination matters because it suggests the pain is not evenly distributed. Upstream participants may be monetizing scarcity. Downstream players, meanwhile, may be constrained by how quickly they can turn component purchases into retail-ready inventory at a price that makes sense for end-users.
It is also happening alongside the broader memory stress that people have been calling the RAMpocalypse. The article connects the dots with Team Group, which recently told consumers, “If you need memory, we recommend purchasing it as soon as possible.” The author admits this is partly typical product marketing, but argues it also reflects a genuine reality: when supply is unstable, waiting can be an expensive strategy. Even if Team Group was speaking mainly about RAM, the storage implication is clear. If NAND is being re-routed toward AI servers and OEMs are unable to secure enough NAND directly, both SSD and RAM purchasing can get harder and more expensive for consumers.
Now zoom out to second-order implications for boards and operators. If retail channels are effectively starved, retailers will raise prices or run out, and consumer sentiment can sour quickly. That can feed back into volumes for gaming PCs, especially if storage upgrades become a budget bottleneck. Meanwhile, OEMs may appear to be winning procurement leverage, not because demand for PCs has exploded, but because their supply path is being built around whatever module maker output remains. For companies that sell components, build storage products, or plan inventory, this is a reminder that demand shifts toward AI can rewire the channel map for everything else.
Finally, consider what this means for decision-making right now. Duann’s comments describe a market that is not merely “tight,” it is being reallocated. If you are planning inventory, forecasting revenue, or negotiating supply terms, the key question is not only how much NAND exists, but where it is going: AI servers, OEM SSD builds, module maker pipelines, and retail availability. In a world where one channel can nearly disappear, the companies that plan for allocation risk rather than just price risk are the ones that keep their footing when the bottleneck moves.
This story's Key Insights and Take-aways are locked.
Create a free account to unlock Executive Actions for one credit.
Register to UnlockAlways free for Executives Club members. Join the Club
More in Technology

UC Davis proves an ALS brain implant can speak with 99% accuracy for 3,800 hours
A Nature Medicine study reports nearly 2 million words, 56 words per minute, and independent work.

Roman Imankulov stopped an npm backdoor with an AI read-only scan
A fake recruiter baited a Python developer into cloning a repo, but a Codex agent flagged a prepare-hook trap.

Databricks unveils Lakehouse//RT and LTAP to erase the serving tier for agents
Two new products aim to cut latency and pipeline complexity by unifying transactional writes and live lakehouse reads.
