Bernstein Analysis: Memory prices are still rising, but smartphones and PCs can no longer withstand it.

TL;DR
· Bernstein's June tracking shows 2QCY26 conventional DRAM sample contract prices rose ~74% QoQ, NAND ~60%.
· TrendForce's public forecast expects 3Q26 general DRAM to rise 13%-18%, NAND Flash 10%-15%.
· Samsung, SK Hynix, Micron still benefit, but LTA caps, smartphone cuts, and NAND competition will constrain the realized gains.

Bernstein's latest June memory tracking puts memory price increases at a very high level. Its sample-weighted data shows 2QCY26 conventional DRAM contract prices rose about 74% from Q1, while overall NAND contract prices rose about 60%.

This set of numbers has direct implications for investors: memory price increases will boost revenue and profit expectations for Samsung Electronics, SK Hynix, Micron, SanDisk, and other manufacturers, while also raising procurement costs for PC, smartphone, and server vendors. Over the past few months, demand from AI servers and cloud providers has pushed memory prices higher, but consumer electronics are already showing counter effects—smartphone makers are cutting production and reducing configurations, and PC and consumer segments are finding it harder to absorb sustained price hikes.

It's important to note that Bernstein's June tracking is based on a sample from its report, not the official final figures publicly released by TrendForce. TrendForce's March public forecast had given 2Q26 general DRAM contract price increases of 58%-63% QoQ, and NAND Flash of 70%-75% QoQ. As of July 3, TrendForce's public outlook expects 3Q26 general DRAM contract prices to rise 13%-18% QoQ and NAND Flash contract prices to rise 10%-15% QoQ.

In other words, memory prices are still rising, and the server shortage is not over, but the steepest part of the price increase may have already passed.

DRAM Sample Up 74% in Q2, Server DDR5 Still the Strongest Demand

DRAM prices remain strong, supported by servers.

Bernstein's June tracking shows spot prices continued the rebound from May. PC DRAM DDR4 and DDR5 chips rose 5.6%-11.5% MoM, server DRAM modules rose 6.1%-26.4% MoM, with DDR5 performing particularly strongly. More importantly, spot prices remain significantly above contract prices, meaning buyers who do not lock in supply through long-term agreements face higher costs in the spot market.

The breakdown of Q2 conventional DRAM contract price increases shows that the rise is not limited to a single category. PC DRAM rose about 49% QoQ, server DRAM about 67%, mobile DRAM about 80%, and consumer DRAM about 85%. Server-side increases are the most significant because they are directly tied to AI data centers and cloud provider capex, and better support memory makers' profit expectations.

DRAM spot and contract price summary; 2QCY26 conventional DRAM sample weighted average QoQ+73.9%; server DDR5 module spot +26.4% MoM.

US cloud providers are still prioritizing their supply. The report mentions that SK Hynix and Micron basically finalized related long-term agreements in April, while Samsung is still pushing for higher prices. Micron's LTA price ceiling is close to its Q2 level, while Samsung and SK Hynix ceilings may be higher. Chinese cloud providers' negotiations extend into Q3, with relatively less favorable terms.

This also explains why contract price increases will slow down, but that doesn't mean a near-term reversal is imminent. Server demand is still absorbing new supply, especially DDR5 and high-end server modules, which remain the strongest part of the overall DRAM market.

Server DDR5 RDIMM spot vs. contract price comparison; spot price far above contract price; June contract price +9.1% MoM; DDR5 premium over DDR4 widens to 22%.

NAND Is Also Rising, But Wafer Side Has Already Weakened

On the surface, NAND is also in a major upcycle, but its structure is more complex than DRAM.

Bernstein's sample shows NAND wafer spot prices fell 3%-4% MoM in June, while wafer contract prices only rose 0.3%-3.7% MoM. Looking at wafers alone, prices have clearly weakened compared to earlier periods. However, mobile NAND and SSD contract prices are expected to surge 70%-80%, pushing the overall Q2 NAND contract price up about 60%.

Not all NAND segments are rising equally. The wafer side has already shown weakness, while end-product and enterprise demand still supports the overall average price.

This also explains why Q3 NAND increases are unlikely to sustain Q2 levels. The high Q2 increases came more from mobile NAND and SSD price adjustments. Once smartphone makers reduce purchases and lower capacity configurations, NAND price pressure will become more apparent than with server DRAM.

NAND wafer and eMMC/UFS contract price trends; TLC wafer contract price +3.7% MoM in June, QLC +0.3% MoM, but mobile NAND and SSD drive 2Q overall QoQ ~+60%.

For NAND-related companies like SanDisk and Kioxia, this divergence is especially important. Prices are still trending upward in the near term, but if wafers weaken first, subsequent profit elasticity depends more on enterprise SSDs, mobile contract execution, and inventory digestion speed.

Q3 Will Still Rise, Just Slower

Q3 is not the end of the price increase, just a slower pace.

TrendForce's July 3 public outlook expects 3QCY26 general DRAM contract prices to rise 13%-18% QoQ and NAND Flash contract prices to rise 10%-15% QoQ. Compared to Q2, the increases have clearly narrowed.

Pressure comes from both ends.

On one side, AI servers are still taking supply, and US cloud providers get priority, supporting prices in the near term. On the other side, consumer electronics have begun to adjust. Smartphone OEMs are cutting production plans and memory usage, with low-to-mid-range models more affected because they rely more on mature products like LPDDR4 and find it harder to pass cost increases to consumers.

TrendForce previously publicly estimated that global smartphone production in 2026 will decline 10% YoY to about 1.14B units, with the decline potentially widening to over 15% in a pessimistic scenario. When smartphone production drops and memory down-speccing occurs simultaneously, the room for mobile DRAM and NAND price increases becomes constrained.

PC faces similar pressure. In June, PC DDR4 and DDR5 chip spot prices continued to rise, but if end demand cannot bear higher costs, procurement pace may slow, and channels will be more cautious in restocking.

PC DDR5 chip spot vs. contract price trends; from end-2025 to June 2026, spot prices rebounded rapidly from lows, briefly corrected in March-April, then rose again.

Another constraint comes from long-term agreements (LTAs). LTAs help memory makers lock in orders and cushion the impact of price declines, but they may also limit further price increases. Especially when some agreements set price ceilings, even if spot prices rise higher, it may not be fully reflected in the manufacturer's current quarter average selling price.

Therefore, the Q2 sample contract price increase cannot be simply equated with each company's actual ASP increase. Product mix, HBM share, LTA price caps, and quarter-end final settlement will all affect the final revenue recognition of Samsung, SK Hynix, Micron, and others.

Vendors Still Benefit, But Smartphones and LTAs will Cap Realized Gains

Memory vendors remain the most direct beneficiaries.

According to Investing.com's June report, Bernstein raised Samsung Electronics' target price to 440k KRW, SK Hynix to 3.3 million KRW, Micron to $1,300, and maintained "Outperform" ratings. SanDisk's target price was also raised to $3,000, maintaining an "Outperform" rating, citing new long-term agreements and price floors. Kioxia maintained an "Underperform" rating.

The ratings themselves are not the focus of this article. More importantly, in the current price environment, the market is still willing to assign higher profit expectations to leading DRAM and some NAND vendors.

But short-term price increases cannot be directly extrapolated to long-term risk-free upside.

Demand destruction on the consumer side has already appeared. Smartphone makers are cutting production and specifications, with low-to-mid-range models more sensitive. The faster memory prices rise, the more likely end vendors will respond by reducing capacity, switching to older generation products, or delaying procurement.

Supply will also continue to ramp up. In Bernstein's model assumptions, memory prices will gradually peak in 2H CY27 and normalize in CY28. The current shortage is more a result of strong AI server demand, lagging supply adjustments, and long-term agreement lock-up, not a permanent supply-demand imbalance.

Chinese vendors' competition in NAND remains a long-term pressure. Compared to DRAM, NAND is more vulnerable to new capacity and price competition. Once demand slows combined with supply increases, price declines may start first in the weaker NAND segments.

The near-term theme of this memory cycle remains clear: server and AI demand keeps DRAM prices elevated, while NAND was also pulled up by mobile and SSD prices in Q2. But starting from Q3, pressure from smartphones, PCs, and consumer segments will significantly slow the pace of increases. For investors, "prices are still rising" and "the strongest phase of increases has already passed" are two things that need to be viewed separately.

Click to learn about job openings at BlockBeats

Welcome to join the BlockBeats official community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Twitter Official Account: https://twitter.com/BlockBeatsAsia