DRAM and NAND prices are still rising in mid-2026
In early June, a single analyst note briefly knocked the memory trade off balance. Barron’s reported that Karl Ackerman expects DRAM and NAND average selling prices (ASPs) to peak in mid-2026 and begin falling quarter over quarter as early as 2027 — roughly a year ahead of his own earlier forecast. Micron’s stock fell on it, the financial press ran with it, and it set off a pointed discussion on X in which even the analyst-watcher who first surfaced the note said he didn’t buy the call. Bearish turning-point calls are common once a cycle runs this long. The more useful exercise is to set the forecast aside and check what the pricing data actually shows.
Where do memory prices actually stand right now, and what does the authoritative pricing data say about the next couple of quarters? Worth answering periodically, regardless of which analyst is bullish or bearish this week. Here’s a clear read as of early June 2026, separating what the market reports show from what any single forecast claims.
A quick distinction first, since the two move differently: contract prices are what large buyers — data centers, PC and phone makers — negotiate for bulk supply, usually quarter by quarter. Spot prices are the open-market rate for one-off orders, and they react faster. Watch spot to see where contract is heading.
Contract prices: still going up, just not as violently
The most recent quarter was extraordinary. Per TrendForce’s finalized survey published June 1, conventional DRAM contract prices rose 93–98% quarter over quarter in Q1 2026 — buyers signing fresh contracts at the start of the year paid roughly double what they paid three months earlier. (“Conventional DRAM” here means commodity memory — PC, server, mobile — as opposed to HBM, the premium stacked memory used on AI accelerators.) That single quarter helped push total memory industry revenue up 81% QoQ to $97 billion.
For Q2, TrendForce expects increases to continue, at a slower clip:
| Segment | Q1 2026 (finalized, QoQ) | Q2 2026 forecast (QoQ) |
|---|---|---|
| Conventional DRAM | +93–98% | +58–63% |
| NAND Flash | ~60% (prior quarter) | +70–75% |
Two things stand out. DRAM’s rate of increase is decelerating — still a big jump, but down from the Q1 record. And NAND is now rising faster than DRAM, the first time that’s happened this cycle. The driver is enterprise SSD demand from large-scale AI deployments, which is soaking up most production capacity and leaving client SSD and eMMC/UFS buyers — the flash that goes into phones and embedded systems — competing for what’s left.
What matters for the “peak” question is the direction. TrendForce has supply still tight, prices “continuing to trend upward,” suppliers locking large customers into multi-quarter long-term agreements, and meaningful new supply not expected until late 2027 or 2028. On the contract side — the prices that actually govern what data centers and OEMs pay — the data does not describe a top forming in mid-2026. It describes a market still tightening, with the rate of acceleration easing rather than reversing.
The spot market is flashing the first caution sign
Across late spring, the signal in spot pricing shifted — not to falling prices, but to buyers who’d stopped chasing them.
Through early May, buying stayed concentrated on DDR5 while mainstream DDR4 drifted; It is noted DDR4 1Gx8-3200 chips edging down 0.25% in one early-May read. By the last week of May, DDR5 spot gains narrowed amid sluggish trading, with most inquiries “only probing the price trend.” The June 3 reading is the telling one. DDR4 and DDR5 spot prices both kept climbing — the mainstream DDR4 1Gx8-3200 chip rose 3.57% in a week to $34.80 — but the latest survey flagged that suppliers’ higher quotes had “made buyers hesitant to chase further increases, thereby limiting actual procurement demand.” NAND spot stayed soft in parallel: 512Gb TLC wafers were essentially flat at $20.68, with “major buy orders nowhere to be seen.”
That nuance is the whole story right now: spot prices are still going up — this is not a downturn — but the demand behind them is starting to balk, and NAND spot has gone quiet. As one trader put it in the X discussion of Ackerman’s note, the signal to watch isn’t an outright price drop, which shows up late — it’s buying appetite fading first. Early June showed the appetite cooling while the prints stayed green.
What the supply side actually supports
The early-peak case leans on two things: new Chinese supply and demand destruction. Both are real. Neither is as immediate as a mid-2026 peak would require.
On supply, CXMT and YMTC — China’s two largest memory makers — are expanding, but capacity doesn’t arrive on a forecast’s schedule. Skeptics argued YMTC is more than a year from adding meaningful volume, which would push real relief well past mid-2026. The incumbents aren’t rushing to fill the gap either. Samsung and SK Hynix signaled late last year they wouldn’t pursue aggressive expansion, NAND makers cut output in the second half of 2025, and Samsung has reportedly weighed further NAND price hikes of 20–30% for 2026. TrendForce expects suppliers to lean on process migration rather than new wafer capacity this year, with meaningful new NAND supply not expected until late 2027 or 2028.
On demand, the destruction is genuine — global smartphone shipments are expected to fall about 14% in 2026 — but it’s concentrated in consumer segments. The AI-driven pull on high-capacity server DRAM and enterprise SSDs, the part of the market setting the price, hasn’t shown the same flinch. There’s also a definitional catch, raised in the same discussion: a forecast about “DRAM and NAND” ASPs usually means commodity memory, not HBM. HBM remains the tightest, most allocation-controlled product in the stack, and it keeps consuming wafer capacity that would otherwise produce commodity DRAM. A peak in commodity ASPs wouldn’t necessarily mean HBM has peaked.
It’s also why retail shelves feel the squeeze indirectly. When enterprise SSD and server DRAM orders absorb the bulk of supplier output, the leftover allocation for client SSDs and desktop memory shrinks — which is part of why DDR5 kits and consumer drives have been expensive and patchy even though the buying frenzy is mostly happening at the data-center level.
So where does the trend actually sit?
Putting the reports together rather than the forecasts: through Q2 2026, memory prices are still rising across the board — NAND now leading on contract, DRAM decelerating from a record Q1, and spot prices for DDR4 and DDR5 still ticking up into early June. The first genuine sign of strain isn’t in prices at all. It’s in behavior: buyers no longer chasing higher spot quotes, and NAND spot trading gone quiet, even as the headline numbers stay green.
The defensible read is that the parabolic phase is maturing, not that a peak has arrived. If the early-peak thesis proves right, you’d expect it confirmed first in spot momentum continuing to fade, then in TrendForce trimming its later-2026 contract forecasts, and only much later in actual contract declines. As of now, only the first of those is even arguably underway. Mid-2026 as the top looks early against the data; mid-2026 as the start of a plateau fits better. Forecasting memory tops is notoriously hard, and a warning that eventually ages well still looks premature at first — which is exactly why the pricing data deserves more weight than the headline.
What this means if you’re holding hardware
For anyone managing hardware rather than trading the stocks, the signal is narrower and more useful. The conditions that pushed new memory and storage to record prices also lifted resale values on used and decommissioned parts — and those conditions are still in force, just no longer accelerating.
Two practical notes follow from that:
- Secondary-market pricing on pulled server and desktop RAM and enterprise SSDs and drives tends to track the new-part market with a lag. If you’re sitting on decommissioned inventory, the window where values sit near cycle highs is worth watching rather than assuming it stays open indefinitely.
- The early softness is in demand behavior, not prices — so this is about timing and attention, not urgency. Nothing in the current data says values are about to fall off a cliff next week.
The next TrendForce spot updates and the Q3 contract survey, due toward late June, will tell us more than any single analyst note. If buyers keep refusing to chase higher quotes and NAND spot stays soft through the summer, the early-peak crowd has a case. If the Q3 forecast prints another big gain, the rally has more room than the headline suggested. We’ll revisit when those numbers land.