Memory Chip Prices Rise Nearly 20% in Q3 as AI-Driven Shortage Persists

Memory chip prices climbed nearly 20% quarter over quarter in the third quarter of 2026 as artificial intelligence demand continued to tighten supply, according to Bernstein. The increase keeps the memory market firmly in an upswing, although the pace of price gains is beginning to moderate.

Bernstein's latest assessment points to continued strength in both conventional DRAM and NAND, two of the most important categories of memory used across servers, PCs, smartphones and storage devices. The research firm expects supply constraints to support another round of price increases in the fourth quarter, with shortages potentially continuing into 2027.

For investors, the development is significant because memory pricing has a direct effect on the profitability of major chipmakers such as Micron Technology, Samsung Electronics and SK hynix. At the same time, higher memory costs are putting pressure on companies that manufacture PCs, smartphones, consumer electronics and other hardware.

Memory Prices Remain High Despite a Slower Pace of Growth

The nearly 20% increase should not be interpreted as another acceleration in the memory-price rally.

Bernstein's data indicates that conventional DRAM and NAND contract prices continued to rise strongly in Q3, but the increase was smaller than the extraordinary gains seen in the previous quarter. Earlier industry data from TrendForce had already pointed toward moderation, forecasting conventional DRAM prices to rise 13%–18% in Q3 and NAND Flash prices by 10%–15%.

That distinction matters.

The memory market is still experiencing a shortage, but customers are becoming less willing to absorb unlimited price increases. PC and smartphone manufacturers are dealing with higher component costs, while some are reducing procurement or shifting toward lower-capacity configurations.

In other words, the shortage remains severe, but the ability to keep raising prices at the same pace is beginning to weaken.

AI Is Changing the Memory Supply Equation

The biggest structural driver is artificial intelligence.

Modern AI data centers require enormous quantities of memory. High-bandwidth memory, or HBM, is essential for advanced AI accelerators, while conventional DRAM and enterprise SSDs are increasingly important for servers handling AI workloads and large-scale inference.

AI inference refers to the process of actually running trained AI models to generate responses, predictions or other outputs. As companies move from experimenting with AI to deploying it at scale, memory requirements continue to expand.

TrendForce said AI server demand remained a major source of memory demand in Q3, while suppliers continued shifting capacity toward server products. Its research also found that enterprise SSD demand was being supported by the expansion of AI infrastructure.

Samsung's second-quarter results provide another indication of the trend. The company's Memory Business reported a record quarter and said it was prioritizing server products because of strong AI demand. Samsung expected demand for server DRAM, enterprise SSDs and HBM to remain robust in the second half of 2026.

SK hynix has also highlighted the structural nature of the demand. The company reported record second-quarter results and said it had entered long-term agreements with around 10 key customers as it works to secure supply and address sustained AI-driven demand.

Why Memory Supply Cannot Increase Quickly

A key feature of the current memory cycle is that additional production cannot be switched on overnight.

Memory manufacturing requires expensive semiconductor fabrication facilities, advanced equipment, process qualification and significant time before new capacity produces commercially useful volumes.

At the same time, existing capacity is being redirected toward higher-value products.

Manufacturers have incentives to prioritize HBM and server memory because AI customers are willing to pay more for specialized products. That can reduce the amount of capacity available for conventional memory used in PCs and smartphones.

TrendForce said suppliers were continuing to prioritize AI-related applications, leaving some consumer memory segments under pressure. It also expects DRAM supply to remain tight as AI demand and HBM capacity requirements compete for wafer resources.

Micron is responding with significant capacity investment. In July, the company said it was increasing planned U.S. investment to more than $250 billion through 2035, citing surging demand for memory in the AI era.

These investments underline both the opportunity and the problem: manufacturers see substantial long-term demand, but the capacity needed to satisfy it takes years to build.

Memory Chipmakers Are the Main Beneficiaries

Higher memory prices generally improve the revenue and earnings potential of manufacturers when they can increase selling prices faster than their costs.

That is particularly important for companies with exposure to several memory categories.

Micron, for example, produces DRAM, NAND and high-bandwidth memory products. Its business therefore has exposure to both traditional memory pricing and the faster-growing AI infrastructure market.

SanDisk is more heavily focused on NAND and storage, making NAND pricing particularly important for its earnings outlook.

Samsung and SK hynix have broader portfolios, but memory remains a major earnings driver for both companies.

The financial impact can be substantial because memory is a cyclical industry. During periods of tight supply, pricing power can expand margins rapidly. When supply catches up, however, prices can fall sharply and profitability can deteriorate.

That makes today's strong pricing environment valuable—but potentially difficult for investors to extrapolate indefinitely.

Higher Memory Costs Are Becoming a Problem for Hardware Makers

The other side of the story is downstream pressure.

PC manufacturers, smartphone companies, gaming-device makers and other electronics producers have to pay more for memory components. If they cannot pass those costs to consumers, their profit margins can shrink.

TrendForce has already warned that record memory prices are pushing consumer customers toward affordability limits. Higher component costs can ultimately result in higher retail prices or lower device shipments.

Smartphones are particularly exposed because memory is an important component cost and the market is already highly competitive.

PC manufacturers face a similar challenge. Companies can reduce memory configurations, delay upgrades or raise prices, but each strategy carries a potential cost in terms of demand.

This creates an unusual situation in which a strong memory market can be excellent news for semiconductor suppliers while simultaneously becoming a headwind for much of the broader technology hardware industry.

Will Memory Prices Keep Rising in Q4?

The current evidence suggests that prices are likely to remain elevated in the fourth quarter, but the rate of increase may slow.

TrendForce's latest forecast calls for conventional DRAM contract prices to rise 10%–15% quarter over quarter in Q4 2026, while NAND Flash prices are projected to increase 15%–20%. Enterprise SSDs are expected to be the strongest NAND category because of continuing cloud-service-provider demand for AI infrastructure.

Bernstein likewise expects the shortage to support further price increases in Q4.

However, several forces could eventually limit the rally.

Long-term supply agreements can include price caps, while high prices are already encouraging PC and smartphone manufacturers to become more cautious. Bernstein expects memory prices to continue rising into 2027 but sees less room for sharp increases as these constraints become more important.

TrendForce also expects the memory market to become more balanced over time as new capacity and process improvements increase supply, particularly in NAND.

What Investors Should Watch

The most important indicators over the next few quarters will be memory contract prices, AI server demand, HBM capacity allocation and supplier capital expenditure.

Investors should also watch whether higher memory costs begin materially reducing PC and smartphone shipments. A slowdown in those markets could eventually weaken demand enough to offset some of the AI-driven strength.

For memory companies, the crucial question is how long elevated pricing lasts. Strong prices combined with disciplined capacity expansion could produce powerful earnings and cash-flow growth. But if manufacturers add capacity too aggressively, the industry could eventually move from shortage to oversupply.

That is the classic risk in a cyclical semiconductor market.

Bottom Line

Memory chip prices rose nearly 20% in Q3 2026 as AI infrastructure demand continued to outpace available supply. Bernstein expects shortages to remain supportive of prices into Q4 and potentially through 2027, although the pace of increases is likely to moderate.

The immediate winners are memory manufacturers such as Micron, Samsung and SK hynix, while PC and smartphone makers face increasing component-cost pressure. The bigger investment question is whether AI demand remains strong enough to keep memory supplies tight before new production capacity catches up.

For investors, the direction of memory prices matters—but the sustainability of those prices matters even more. The next major signals will come from Q4 contract pricing, AI server demand, HBM allocation and semiconductor companies' capacity-expansion plans.

Follow the blog for more semiconductor, AI, technology and global stock-market updates.

This article is for informational and educational purposes only and should not be considered investment advice.