🔍 Read the full analysis: AI Is Reshaping Memory Demand—and Your Tech Budget on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
SK hynix expects memory demand to exceed supply through 2030, while Samsung and Micron have also signaled that tight supply will persist. AI data centers are competing for production capacity used to make memory for consumer devices, contributing to higher component costs; the duration and full effect on retail prices remain uncertain.
SK hynix expects memory demand to exceed supply through 2030, the South Korean chipmaker said Oct. 28, warning that high memory costs and supply limits could reduce shipments of PCs and mobile devices. The forecast links the AI data-center buildout to the cost and availability of memory used in consumer electronics, though the scale and timing of any retail price increases remain uncertain.
In its third-quarter release, SK hynix said supply was not keeping pace with market growth. It cited the technical difficulty of advanced manufacturing and the long lead time for new factories in forecasting that demand would remain above supply through 2030. The release did not specifically discuss phones or PCs; Chief Financial Officer Kim Woo-Hyun addressed consumer shipments on the earnings call, saying they were expected to decline because of rising memory costs and shortages.
The company said AI infrastructure demand would more than offset that expected decline. An SK hynix executive said customer requests for high-bandwidth memory (HBM) were more than twice what the company could currently supply. The company also said long-term agreements with 14 major customers cover more than half of its production capacity for 2027 and 2028. SK hynix described those agreements as typically five years long and said they include deposits.
Other suppliers have offered their own outlooks: Samsung has said tight supply may persist into 2028, while Micron has said it has no line of sight to a balance between supply and demand. Those are company forecasts, not guarantees. Samsung’s Oct. 29 results also showed a contrast between business lines: its memory business set a record, while its phone and network business lost money, according to the source material.
Infographic 01 · Where the memory goes
AI data centers get first call on memory.
High-bandwidth memory (HBM) for AI chips and the memory in your phone come out of the same factories. HBM needs far more wafer per gigabit, and its buyers sign years ahead.
The bakery problem
HBM share of wafers vs. share of the memory bits they produce
Same factories, same wafers: every wafer used for AI memory is missing from phone and PC memory.
Who gets first call
Long contracts lock up supply years ahead
- SK hynix: deals with 14 major customers, usually five years, with deposits; more than 50% of 2027–2028 capacity
- Micron: more than 75% of fiscal 2027 shipments already committed
- HBM demand: customers want “more than double” what SK hynix can supply today
- Cloud budgets: memory ≈ 47% of big cloud providers’ capex in 2026, 68% in 2027 (TrendForce estimates)
Three makers, three horizons, one customer’s view
- SK hynixThrough 2030Demand above supply (company forecast)
- SamsungInto 2028Tight supply, based on market research
- MicronNo “line of sight”Tighter in 2027 and 2028 than in 2026
- Acer (customer)Easing from 2H 2027Says suppliers talk up the shortage
HBM is a wedding cake, phone memory is bread. Same ovens. The cake customers pay more and order years ahead.Thorsten Meyer · my analogy
Sources: TrendForce, 2 Jun (wafer and bit shares, estimates); TrendForce, 25 Aug (capex shares, estimates); SK hynix, 28 Oct 2026; SK hynix call, 28 Oct; Micron call, 30 Sep; The Register, 1 Oct; Samsung call, 29 Oct; Acer via Tom's Hardware / DigiTimes. Wafer-per-gigabit ratio: my arithmetic. All horizons are forecasts.
Powered by Thorsten Meyer AI
AI Chips Compete for Memory Capacity
HBM is a form of dynamic random-access memory, or DRAM, designed to move data quickly beside AI processors. It is made using manufacturing resources that also serve other memory products, including DRAM for computers and phones. When producers assign more capacity to HBM, fewer resources may be available for ordinary memory. The extent of that trade-off depends on production choices and manufacturing yields, but AI demand is now competing with consumer electronics for a constrained supply base.
That pressure can affect shoppers through more than the price of a memory kit. The source reports that a 32-gigabyte DDR5-6000 kit cost at least $447 in the United States, compared with a historical low of $72 in Tom’s Hardware’s price index, last updated Oct. 26. This is a specific retail example, not a measure of all memory products or a guarantee that prices will stay at that level. Higher component costs can also prompt device makers to raise prices or adjust specifications, making upgrades and new phones more expensive or less capable at a given price.
For businesses, the same constraints can complicate procurement and product planning. Manufacturers may face higher bills or have to compete for allocations, while consumers may delay upgrades or choose lower-spec devices. The outcome will vary by product, supplier contracts and market conditions; SK hynix’s forecast signals prolonged pressure, not a precise price path for each device.
Popular right now
Ninja Flip Toaster Oven & Air Fryer, 8-in-1, Stainless Steel, SP151
Vitamix Propel Series 750 Professional-Grade Blender, 4 Programs, Black
Vitamix 5200 Professional-Grade Blender, 64 oz. Container, Black
As an affiliate, we earn on qualifying purchases.
How HBM Affects Consumer Memory
Two broad types of memory help explain the connection. DRAM temporarily holds information a device is using, while NAND flash stores files and apps after power is off. PCs often use removable DRAM modules, while phones and thin laptops typically use memory soldered onto their boards. HBM is also DRAM, but it stacks chips and connects them to serve data-intensive processors at high speed.
Market researcher TrendForce estimates that HBM takes a larger share of wafer capacity than its share of the resulting memory bits. Its figures cited in the source put HBM at about 22% of wafers started by the three biggest makers by the end of 2026, producing about 9% of memory bits; for the end of 2027, the estimates are about 30% of wafers and 13% of bits. These are TrendForce estimates; the claim that HBM uses roughly three times as much wafer per gigabit is the source author’s calculation from those figures, not a direct company statement.
The source also cites TrendForce estimates that contract prices for ordinary memory rose 93% to 98% in the first quarter of 2026 and continued rising in subsequent quarters. That is a reported contract-market trend, distinct from retail prices and from the particular memory-kit example above. SK hynix’s forecast arrives as suppliers invest in capacity, but new production takes time to build and qualify.
“Supply is not keeping pace with the growth of the market.”
— SK hynix, in its Oct. 28 earnings release
Prices and Shortage Duration Remain Unclear
The 2030 outlook is SK hynix’s forecast, not an independently verified timetable for shortages or retail price increases. Samsung’s projection into 2028 and Micron’s comments are also supplier assessments. Actual conditions will depend on how quickly new capacity comes online, manufacturing yields, customer orders and whether AI investment continues at its current pace.
There is also a dispute over how suppliers describe the market. The source says Acer has accused memory suppliers of talking up shortages to protect margins. That is an allegation from a large customer, not an established finding. Suppliers reporting record results also have a financial interest in the market outlook. The available material does not establish that forecasts are misleading or quantify how much of the price pressure comes from AI demand rather than other factors.
Reported changes to device specifications—including reduced storage, cameras or memory in some mainstream models, and the reuse of recycled memory chips by some phone makers—are attributed in the source to TrendForce and Counterpoint. The material does not identify every affected model or establish that these changes are widespread across all brands.
Watch Supply and Device Pricing
Investors and customers will look to upcoming earnings calls and company updates for revised supply forecasts, production plans and contract details. The next practical test is whether new manufacturing capacity and improved output can narrow the gap suppliers describe, particularly for HBM and standard DRAM. The companies have not provided a single industry-wide date when supply will catch up.
For consumers, the relevant indicators are retail memory prices, device makers’ announced prices and specifications, and shipment data from research firms such as IDC. The source reports that IDC expected the average smartphone to cost $581 in 2026, up 27.6%, and cited declines in third-quarter PC and smartphone shipments. Those figures describe the market outlook available in the source; they do not by themselves show that memory costs caused every price increase or shipment decline. Further reporting will be needed to distinguish the effects of memory costs from other market factors.
Key Questions
Why is AI demand affecting memory used in phones and PCs?
AI servers use high-bandwidth memory alongside processors. HBM and consumer memory draw on overlapping manufacturing resources, so allocating more capacity to HBM can leave less available for other memory products.
Did SK hynix confirm that PC and phone prices will rise through 2030?
No. SK hynix forecast that memory demand would exceed supply through 2030 and said higher costs and shortages could reduce PC and mobile-device shipments. It did not give a retail price forecast for those devices.
What is HBM?
High-bandwidth memory is a type of DRAM made from stacked chips and designed to move data quickly near AI processors. It is different from the standard memory modules or soldered chips used in many consumer devices.
How certain is the shortage timeline?
The timeline is uncertain. Through 2030 is SK hynix’s forecast; Samsung has pointed to tight supply into 2028, and Micron has said it cannot see a supply-demand balance. Future output, demand and prices may change.
Could device makers respond without raising prices?
They may change product specifications, seek alternative supplies or absorb some added costs, but the effects will differ by company and product. The source reports examples of specification reductions and recycled memory use, but does not establish how broadly those measures are being adopted.
Source: ThorstenMeyerAI.com
