The price of storing a byte is about to go up. Significantly.
After years of steadily declining costs that made terabyte-class solid-state drives affordable for everyday consumers, the SSD market is entering a period of turbulence not seen since the NAND flash shortages of 2017. The forces at play this time are different — and arguably more complex. A convergence of U.S. tariff policy, surging AI-driven demand for storage, and preemptive buying by manufacturers spooked by supply uncertainty is creating conditions that could push SSD prices up by 20% or more before the end of 2025.
The warning signs have been flashing for months. According to The Verge, NAND flash memory — the fundamental building block of every SSD — has been subject to tightening supply conditions since late 2024, when major manufacturers including Samsung, SK Hynix, and Kioxia began shifting production capacity toward high-bandwidth memory (HBM) chips used in AI accelerators. That reallocation didn’t cause immediate pain. But it planted the seeds for what’s unfolding now.
The trigger was tariffs.
When the Trump administration imposed sweeping new tariffs on Chinese-manufactured goods in early 2025, the ripple effects reached deep into the electronics supply chain. SSDs and their components weren’t exempt. Many consumer drives sold in the United States are either assembled in China or contain NAND flash packaged at Chinese facilities. The tariffs, which in some categories reached 145%, created immediate cost pressure on importers and OEMs. Some absorbed the hit temporarily. Others couldn’t.
As The Verge reported, the situation was compounded by a wave of preemptive purchasing. PC makers, data center operators, and even smaller system integrators began stockpiling SSDs in anticipation of further price increases and potential supply disruptions. This panic buying — rational from each individual buyer’s perspective — accelerated the very shortage it was meant to hedge against. A classic self-fulfilling prophecy in the supply chain world.
The numbers tell the story. TrendForce, the Taiwan-based market intelligence firm, projected that contract prices for NAND flash would rise between 13% and 18% in the second quarter of 2025 alone. That’s the wholesale price that drive manufacturers pay for raw flash chips. By the time those increases work through the supply chain — with each intermediary adding margin — retail SSD prices could climb substantially more. Some analysts expect 2TB consumer NVMe drives, which had fallen below $100 in late 2024, to creep back toward $130 to $150 by fall.
And it’s not just consumers who’ll feel the pinch. Enterprise storage — the massive arrays of SSDs that power cloud computing, AI training clusters, and corporate data centers — faces even steeper increases. Enterprise-grade NAND commands premium pricing, and the customers buying it tend to be less price-sensitive but far more volume-hungry. When hyperscalers like Microsoft, Google, and Amazon place orders, they don’t buy thousands of drives. They buy millions. That kind of demand, layered on top of an already tightening market, creates enormous pressure on available supply.
The AI factor deserves particular scrutiny. The explosive buildout of GPU clusters for training and running large language models has created a secondary demand surge for fast storage. Every AI training run requires enormous datasets to be fed to processors at high speed. That means NVMe SSDs — lots of them. Nvidia’s own reference architectures for AI servers specify multiple high-capacity SSDs per node. When a company like Meta or xAI orders tens of thousands of AI servers, the SSD requirements alone can consume a meaningful fraction of quarterly NAND production.
This isn’t hypothetical. SK Hynix, one of the world’s three largest NAND producers, has been openly prioritizing its advanced packaging and memory production for AI-related products, including HBM chips that sit on Nvidia’s H100 and H200 GPUs. That strategic pivot is rational — HBM commands margins that commodity NAND flash simply can’t match. But every wafer allocated to HBM is a wafer not producing the 176-layer or 232-layer TLC NAND that goes into your laptop’s SSD.
Samsung, the world’s largest memory chipmaker, finds itself in a similar position. The company has been aggressively expanding its HBM production to recapture market share lost to SK Hynix in the AI memory race. Samsung’s V-NAND production hasn’t been cut, but it hasn’t grown at the pace that would be needed to offset the demand surge from both AI infrastructure and tariff-driven stockpiling.
Kioxia — formerly Toshiba Memory — and Western Digital, which share NAND fabrication facilities in Japan, represent another critical node in the supply picture. Their joint venture fabs in Yokkaichi and Kitakami produce a significant share of global NAND output. Any production hiccup at these facilities, whether from equipment issues, natural disasters, or deliberate output management, can move global pricing.
So where does this leave the average buyer?
In the near term, the advice from most industry watchers is straightforward: if you need an SSD, buy it sooner rather than later. Prices at retail haven’t yet fully reflected the wholesale increases that took effect in Q2 2025. Retailers and e-commerce platforms are still working through inventory purchased at older, lower prices. Once that stock is depleted, the replacement inventory will carry higher costs that will be passed to buyers.
The situation has drawn comparisons to the hard drive crisis of 2011, when catastrophic flooding in Thailand wiped out a significant portion of global HDD manufacturing capacity. Prices doubled virtually overnight and took more than a year to normalize. The current SSD situation isn’t as acute — there’s no single catastrophic event — but the combination of tariffs, demand shifts, and supply reallocation creates a more structurally persistent problem. Floods recede. Trade policy and AI demand curves don’t.
There’s a geopolitical dimension worth watching too. China’s domestic memory industry, led by Yangtze Memory Technologies Co. (YMTC), has been making significant strides in NAND flash production. YMTC’s 232-layer NAND is technically competitive with offerings from Samsung and SK Hynix. But U.S. export controls imposed in 2022 and expanded since then have restricted YMTC’s access to advanced chipmaking equipment, constraining its ability to scale production. And the new tariffs make it harder for YMTC’s output to reach American consumers even through indirect channels.
The irony is thick. U.S. policy simultaneously restricts Chinese memory production (through export controls on equipment) and penalizes the import of memory products from China (through tariffs). The net effect is to reduce the total supply available to American buyers from multiple directions at once.
For enterprise buyers, the calculus is more nuanced. Many large cloud providers have long-term supply agreements with NAND manufacturers that partially insulate them from spot-market volatility. But “partially” is the operative word. When underlying costs rise 15% or more, even contracted prices get renegotiated. And smaller enterprise buyers — mid-market companies running their own servers, managed service providers, regional data centers — typically don’t have the purchasing leverage of a Google or an Amazon. They buy closer to market pricing and feel the swings more acutely.
The storage industry’s response has been mixed. Some SSD brands have already begun adjusting MSRPs upward. Others are holding prices but quietly reducing promotional activity and channel incentives, which effectively raises the street price without changing the sticker. A few have introduced new SKUs with slightly lower specifications — fewer DRAM cache chips, lower endurance ratings — at the old price points, a form of shrinkflation adapted for the tech world.
There are countervailing forces that could moderate the price increases over time. NAND manufacturers have been investing heavily in increasing the number of layers in their 3D NAND chips — moving from 176 layers to 232 and beyond. More layers per chip means more storage capacity from the same silicon area, which improves cost efficiency. Samsung has announced work on 300-plus-layer NAND. These advances will eventually bring per-gigabyte costs down again. But “eventually” might mean late 2026 or 2027 before the benefits fully materialize in consumer pricing.
Another potential relief valve: if the tariff situation changes. Trade policy under the current administration has been volatile, with tariffs announced, paused, modified, and reimposed in rapid succession. A negotiated reduction in tariffs on electronic components would immediately ease cost pressure on imported SSDs and NAND flash. But betting on specific trade policy outcomes is a fool’s errand.
The broader tech industry should be paying attention to what the SSD market is signaling. Storage costs are one of those foundational inputs that affect everything from the price of a laptop to the cost of running a cloud workload. When storage gets more expensive, it doesn’t just mean consumers pay more for a new drive. It means higher costs for cloud storage services, more expensive AI training runs, pricier gaming consoles, and tighter margins for every company that depends on digital infrastructure. Which is to say, every company.
The last time NAND flash experienced a sustained price increase cycle, in 2016-2017, it contributed to higher PC prices and delayed the adoption of SSDs as the default storage medium in budget laptops. The industry had only recently crossed the threshold where SSDs became standard equipment in most new PCs. A significant price reversal now could slow the ongoing transition away from hard drives in cost-sensitive segments like Chromebooks, budget desktops, and entry-level servers.
Not everyone is pessimistic. Some analysts argue the current supply tightness is manageable and that the worst-case scenarios assume all negative factors persist simultaneously — maximum tariffs, maximum AI demand, maximum stockpiling — without any offsetting adjustments. Markets adapt. Manufacturers will eventually shift capacity back toward NAND if margins justify it. Tariffs may be negotiated down. Panic buying will subside once inventories reach comfortable levels.
All true. But the timing matters enormously for anyone making purchasing decisions right now.
For IT procurement teams planning second-half-2025 infrastructure refreshes, the message is clear: budget for higher storage costs and lock in pricing where possible. For consumers building or upgrading PCs, the window of sub-$100 2TB SSDs may be closing. And for the industry at large, the SSD price squeeze is a reminder that the physical infrastructure underpinning the AI boom has real costs — costs that are rising, that are interconnected with global trade politics, and that will ultimately be borne by someone.
The question is who, and how much.
WebProNews is an iEntry Publication