AI Memory Shortage Is Now Raising Consumer PC Prices

PC buyers are starting to feel the cost of the artificial intelligence boom in a place that has nothing to do with AI software: the parts inside ordinary computers. PC prices in Singapore have risen by nearly 40% over the past year, with memory, storage and graphics components among the areas under pressure. Reports from the region describe some memory prices rising far faster than complete PCs, while retailers are seeing weaker demand as buyers postpone upgrades. 

The interesting part is what sits behind those increases. This is not simply a temporary shortage of gaming components. AI data centers are consuming enormous quantities of high-bandwidth memory (HBM), server dynamic random-access memory (DRAM) and enterprise solid-state storage, and manufacturers are directing more capacity toward those higher-value products. That leaves consumer hardware competing for a tighter supply of conventional memory and storage.

AI servers are pulling memory capacity toward higher-value hardware

The memory industry has changed its priorities because AI servers need far more memory bandwidth and capacity than many traditional workloads. HBM is a specialized form of DRAM designed to move large amounts of data between processors and memory at very high speed, making it particularly important for AI accelerators. Conventional DDR5 memory, which is familiar to PC builders, is manufactured by the same broad memory industry and competes for manufacturing resources.

TrendForce reported in September that global DRAM industry revenue reached almost $154.73 billion in the second quarter of 2026, up 59.5% from the previous quarter. Its analysis says suppliers' inventories remain at historic lows and that additional supply is being directed mainly toward server applications. Consumer DRAM is expected to experience the strongest price growth because manufacturers have significantly reduced the proportion of supply going to that market. 

That distinction matters. The problem is not that memory manufacturers have stopped producing ordinary PC RAM. They are producing it in an environment where server and AI customers can absorb more expensive products and secure supply through large purchasing agreements. A PC manufacturer or memory-module seller therefore has less negotiating power when available capacity is tight.

HBM creates an unusually large demand for underlying DRAM

HBM also creates a less obvious supply problem. Building advanced HBM products requires DRAM dies, packaging capacity and more complex manufacturing processes. Micron said earlier this year that the industry's supply-demand balance had become substantially short of demand and highlighted the effect of HBM's production requirements on conventional DDR5 supply. The company described a three-to-one trade ratio between HBM and DDR5 in one part of its manufacturing analysis, meaning the transition toward HBM can consume considerably more underlying DRAM resources than an equivalent amount of conventional memory. 

Micron's subsequent results show how strongly its business has shifted toward AI infrastructure. In its fiscal third-quarter 2026 results, the company said HBM4 was already in high-volume shipments for a lead customer's platform and that it had begun shipping its next-generation HBM product while also expanding server-oriented memory and data-center storage. 

Those developments do not prove that every dollar of a PC price increase comes from HBM. They do show why consumer memory is competing against a rapidly expanding market with much larger capacity requirements and higher-value products.

Storage is facing the same pressure from a different direction

The squeeze is not limited to RAM. NAND flash is the non-volatile memory used to store data in solid-state drives (SSDs), phones and other devices. AI data centers need large quantities of enterprise SSD capacity because models generate and retrieve huge amounts of data beyond what can remain in the faster memory attached directly to AI accelerators.

TrendForce's September NAND analysis says demand from artificial intelligence and cloud services has pushed enterprise storage products into shortage, lifting average selling prices and industry revenue. At the same time, consumer electronics demand remains weaker because prices are already high. That creates an unusual split in the market: enterprise customers are buying more storage while ordinary consumers are becoming more reluctant to pay for it. 

For a desktop builder, the effect is straightforward. A 1TB or 2TB consumer SSD is not competing directly with a giant AI cluster for the exact same retail product, but both markets ultimately depend on NAND manufacturing capacity. When suppliers have stronger incentives to expand enterprise products, consumer drives can remain expensive even when retail demand is not particularly strong.

The result is already visible in consumer PC pricing

The Singapore market provides a useful real-world example because the change is large enough to be visible at the finished-PC level. The Straits Times reported that PC prices there had risen by about 40% in 2026, with memory identified as the largest source of component inflation and some RAM prices reportedly increasing by nearly 300% over a year. The same reporting described substantial declines in sales at some retailers as customers delayed purchases. 

A nearly 40% increase in a complete PC should not be interpreted as a 40% increase in the price of every component. A computer contains a CPU, motherboard, power supply, chassis, cooling system, display and other parts whose prices can move independently. But memory and storage sit inside almost every modern PC, so sustained increases in those components can eventually spread through the system price even when the processor itself has not become dramatically more expensive.

GPU prices add another layer to the problem

Graphics cards are also exposed to memory supply because modern GPUs use dedicated graphics memory. The relationship is not identical to the PC DRAM market: graphics processors have their own semiconductor manufacturing, packaging and board supply chains. Nevertheless, a shortage or price increase in graphics memory can raise the cost of producing graphics cards, particularly when demand for AI accelerators is already competing for advanced semiconductor capacity.

This is why a buyer can encounter several increases at once. System memory may be more expensive, SSDs may be under pressure, and a graphics card may carry its own supply and manufacturing premium. None of those increases needs to come from exactly the same shortage for the final PC price to move higher.

Memory companies are adding capacity, but factories do not appear overnight

The obvious solution is to manufacture more memory, and the major suppliers are investing heavily. Micron has said it is increasing investment in manufacturing and expects its supply-demand conditions for both DRAM and NAND to remain tight beyond calendar 2027. The company has also accelerated construction and expansion projects intended to add future capacity.

SK hynix has likewise reported strong AI-driven demand and said it has signed multi-year agreements with around ten key customers while expanding capacity. Those agreements give major customers greater supply visibility, but they also illustrate why consumer buyers cannot assume that new production will immediately translate into cheaper desktop RAM or SSDs. Much of the additional capacity is being planned around the products and customers generating the strongest demand. 

There is also a timing problem. A semiconductor fab requires cleanrooms, specialized equipment, process qualification and a lengthy production ramp. Even when a manufacturer announces billions of dollars in new investment, the resulting chips may arrive years after the original capacity decision. That is why today's shortage cannot be solved simply by increasing factory output next month.

PC buyers may need to change how they spend

For consumers, the practical response is not to panic-buy every component. The better approach is to identify which upgrade actually removes the current performance bottleneck. Someone replacing a five-year-old PC may benefit from a complete system upgrade, while someone with a modern processor may gain more from adding storage or memory rather than replacing the entire machine.

Capacity choices also deserve more attention. If memory and storage prices remain elevated, buying more than the workload requires can add a significant cost without producing a noticeable improvement. A balanced system with adequate RAM and a sensible SSD can be a better purchase than paying a large premium for capacities that will sit unused.

The next phase will depend on AI demand and new memory capacity

The important question is no longer whether AI is affecting PC hardware prices. The evidence shows that it already is. The harder question is how long the pressure lasts, because it depends on two moving targets: how quickly AI infrastructure continues to consume memory and storage, and how quickly manufacturers can add economically viable capacity.

Current industry data gives little reason to expect an immediate return to the unusually cheap memory and storage that consumers became accustomed to. TrendForce expects conventional DRAM contract prices to keep rising in the third quarter of 2026, although at a slower rate than earlier increases, while NAND markets remain split between strong enterprise demand and weaker consumer demand.

For PC buyers, that means the next few hardware generations may be shaped by a constraint that is easy to overlook when comparing processors and graphics cards: the world's most valuable computing hardware is competing for the same broad memory ecosystem. Until new capacity catches up with AI-driven demand, the cost of building an ordinary PC will remain partly tied to what is happening inside much larger machines.