The NAND shortage is not ending soon—it is getting worse. Silicon Motion, the independent developer of SSD controllers with roughly 30% market share in client storage, is warning that NAND supply conditions will deteriorate significantly in 2027 rather than improve. The company’s forecast cuts against earlier industry hopes that the memory crunch would ease, revealing instead a structural supply-demand imbalance driven by relentless AI infrastructure expansion.
Key Takeaways
- NAND shortage expected to intensify in 2027, with shortages potentially lasting until 2028.
- All NAND production capacity for 2026 is already sold out; new factories won’t open until late 2027.
- Silicon Motion’s SSD controller sales grew 35–40% year over year in Q4 2025 on strong demand.
- NAND product prices surged 4 to 10 times between August 2025 and March 2026.
- AI infrastructure like NVIDIA’s NVL72 cabinets require massive storage—1.16 petabytes per unit.
Why the NAND shortage is structural, not cyclical
This is not a temporary supply hiccup. The NAND shortage stems from a fundamental mismatch between production capacity and demand driven by AI data centers. Silicon Motion CEO Wallace Kou stated that the shortage in 2027 will be even more severe than in 2026, with relief unlikely before the second half of 2027 or 2028. The company expects memory and SSD shortages to persist into 2028 due to continued AI-related demand. What makes this different from past shortages is that the constraint is not temporary—it reflects how much infrastructure AI requires.
All NAND production capacity allocated for 2026 is already sold out. New fabrication plants, the only real solution to easing supply, will not open until late 2027 at the earliest. This means the shortage will intensify before any relief arrives. NAND prices have already reflected this scarcity, with some products surging 4 to 10 times in cost between August 2025 and March 2026. These are not small fluctuations—they are crisis-level price movements.
How AI infrastructure is consuming storage at an unprecedented scale
The root cause is simple: AI systems require staggering amounts of storage. A single NVIDIA NVL72 cabinet, used in large-scale AI deployments, requires 1.16 petabytes of operational storage. That is not a typo. One cabinet. Multiply that across the thousands of cabinets deployed globally by cloud providers, and the storage demand becomes incomprehensible to traditional supply chains. Silicon Motion is one of only two approved suppliers for NVIDIA’s BlueField-3 DPU boot storage, and the company began shipping significant volumes to NVIDIA in Q4 2025, with forecasts of approximately $50 million in boot drive revenue for 2026.
This is why Silicon Motion’s SSD controller sales are hitting record levels despite the shortage. The company’s Q4 2025 controller sales increased 35–40% year over year, and management forecasts a record revenue year for 2026. The shortage is not hurting the company—it is enriching it. Controller makers profit when NAND is scarce because demand for their chips remains strong while NAND prices spike. Phison, another SSD controller vendor, has also seen strong revenue growth tied to controller shipments. But for everyone else—PC makers, server manufacturers, consumers—the shortage is a cost problem.
What this means for the broader storage market
The NAND shortage is reshaping the storage hierarchy. Client SSDs—the drives in laptops and desktops—are being squeezed as data center demand takes priority. This is a structural shift, not a temporary allocation issue. PC makers and consumer device manufacturers will face higher component costs and potentially longer lead times throughout 2026 and into 2027. The shortage will not ease until new capacity comes online, and that will not happen until late 2027 at the earliest.
For data center operators and AI infrastructure companies, the shortage is a cost of doing business—expensive, but manageable because the storage value they extract justifies the premium. For everyone else, it means paying more for less. The industry is in a race: can AI demand eventually plateau, or will it keep accelerating faster than new fabs can open? Based on current trajectories, demand is winning.
Will new NAND capacity solve the problem?
New fabrication plants are the only structural solution to the NAND shortage. However, these facilities take years to build and billions to finance. The earliest new capacity is expected in late 2027, which means the shortage will likely persist well into 2028. Even when new fabs open, it takes time for them to ramp production to meaningful volumes. This creates a multi-year supply crunch that will reshape pricing across client storage markets.
How long will the NAND shortage last?
Silicon Motion expects relief to arrive no sooner than the second half of 2027 or 2028, depending on how quickly new capacity comes online and ramps. Some forecasts suggest shortages could persist into 2028 entirely. The shortage is not a six-month problem—it is a two to three year problem baked into the supply chain.
Is the NAND shortage affecting SSD controller makers positively?
Yes. SSD controller vendors like Silicon Motion are thriving because demand for controllers remains strong while NAND supply is constrained. This drives up controller prices and shipment volumes simultaneously. However, the broader storage industry—from NAND manufacturers to PC makers to consumers—faces mounting cost pressures as a result.
The NAND shortage is not a temporary market disruption. It is a structural imbalance between AI infrastructure demand and production capacity that will intensify before it improves. New capacity will not arrive until late 2027, meaning shortages will likely persist into 2028. For storage buyers, the message is clear: expect higher costs and tighter supplies for at least the next 18 months.
Edited by the All Things Geek team.
Source: Tom's Hardware
