Prices for memory chips are forecast to soar in 2026. but what just as bad/even worse it looks like chip customers are looking at serious supply shortages during the year.
Higher prices and supply shortages could be enough to put earnings at risk at some high multiple cloud and AI companies.
TrendForce and other market trackers project conventional DRAM prices up roughly 55–60% quarter‑on‑quarter in the first quarter of 2026, after a 45–50% jump in the fourth quarter of 2025. For all of 2026, analysts see DRAM prices possibly roughly doubling versus mid‑2025 levels.
NAND flash contract prices (SSDs, eMMC, UFS, etc.) are expected to increase at a double-digit rate quarter over quarter in the first quarter of 2026. Full‑year 2026 projections generally call for total NAND price increases of about 20–30% versus 2025, driven by demand growth (20–22%) outpacing capacity growth (15–17%).
The biggest price increases and supply shortages are likely for high-bandwidth chips used in AI data‑centers.
Micron Technology’s (MU) AI‑oriented memory capacity is effectively sold out for 2026, especially in High Bandwidth Memory (HBM). Executives state that Micron is “completely sold out for 2026” on AI‑oriented memory, with HBM capacity fully booked through calendar 2026 and broader 2026 memory production described as sold out. Other major memory manufacturers have effectively sold out their 2026 production capacity for key products, especially high‑bandwidth and advanced DRAM. SK Hynix has said its HBM capacity for 2026 is fully booked, and multiple reports note that its DRAM, NAND, and HBM production capacity is sold out through 2026 due to large AI orders (notably from Nvidia and hyperscalers). Samsung’s HBM3/HBM3E capacity for 2026 is already sold out, with management highlighting that even after boosting HBM capacity, demand still exceeds supply and is crowding out traditional DRAM.
The biggest hit from this will fall on AI data‑center hardware and then on mass‑market PCs, laptops, and smartphones, with knock‑on effects for many other device categories.
Some, but not all, of these price increases will get passed on to customers. Cloud providers are seeing meaningful cost pressure from higher memory prices and are starting to pass part of that through in the form of higher prices, especially for memory‑intensive services, while absorbing some hit to margins. Hyperscalers will mostly pass higher DRAM/HBM costs into cloud and AI‑service pricing, squeezing margins for customers running large‑scale training or inference on rented infrastructure. Analysts expect PC DRAM costs and system prices to rise sharply, with typical laptops either getting more expensive or shipping with less RAM (e.g., 8 GB instead of 16 GB at the same MSRP).
IDC and other market analysts see PC and server price increases of roughly 15–20% as likely in early 2026, primarily due to memory. Mobile DRAM and NAND costs have surged (often 70–100% versus last year), pushing memory to around 20% of a smartphone’s bill of materials and adding roughly 25% to costs on average.
the price increases for memory chips will lead to margin compression at the infrastructure layer, at least near term.
The squeeze is much worse for smaller cloud and colocation providers, which lack scale and long‑term supply contracts; reports warn of significant margin pressure and potential consolidation among mid‑tier players. AI training and inference clusters are deeply affected, as advanced memory chips dominate their cost structure; higher memory cost raises per‑hour pricing for top‑end GPU instances and model‑serving platforms.
The question for investors is whether the effects on earnings will be big enough to worry investors. I’d watch earrings for Microsoft (MSFT) on Wednesday, January 28, 2026; Apple (AAPL) on Thursday, January 29, 2026, and then in February Advanced Micro Devices (AMD) and Alphabet GOOG).
