TrendForce will still forecast rising DRAM contract prices for the third quarter of 2027

Bottom line.

1. TrendForce's forecast increases for conventional DRAM contract prices have shrunk from a near-doubling in early 2026 to 10 to 15 percent for the fourth quarter, with the market still described as undersupplied.
2. Multi-year take-or-pay contracts and new fabs that mostly arrive in 2027 and 2028 weaken two of the three mechanisms that ended earlier memory cycles.
3. The main risk to the forecast is capacity moving back from HBM to conventional DRAM, which would probably show up in spot prices before it reaches TrendForce's contract forecasts.

Planning assumption. By July 31, 2027, TrendForce's first forecast for third-quarter 2027 conventional DRAM contract prices still projects a quarter-on-quarter rise (60 percent: likely).

Micron's gross margin is 86.8 percent while worldwide PC shipments fell more than a fifth last quarter, both because of the same shortage of memory chips. TrendForce, the market tracker whose quarterly price forecasts the industry watches, expects conventional DRAM contract prices to rise 10 to 15 percent this quarter, after a near-doubling in the first. This forecast puts a 60 percent probability on its first forecast for the third quarter of 2027 still showing a rise.

How tight the market is

Contract prices are what PC makers, phone makers and cloud operators pay under negotiated supply deals, and TrendForce publishes a forecast range for each quarter around the time it begins. Over the past year those forecasts have run far behind reality. In September 2025 TrendForce expected conventional DRAM prices to rise 8 to 13 percent in the fourth quarter of 2025 (TrendForce); they rose 45 to 50 percent (TrendForce). Its January forecast for the first quarter of 2026, a 55 to 60 percent rise, was revised to 90 to 95 percent four weeks later (TrendForce), and the outcome was roughly 93 to 98 percent, lifting DRAM industry revenue 81 percent in a quarter to $97 billion (TrendForce).

The forecast increases have shrunk since then: 58 to 63 percent for the second quarter (TrendForce), 13 to 18 percent for the third (TrendForce), and 10 to 15 percent for the fourth, with the market described as still undersupplied (TrendForce).

The suppliers' books show what that means. Micron reported revenue of $54.23 billion for its fiscal fourth quarter, which ended September 3, with a GAAP gross margin of 86.8 percent (Micron, SEC filing). That quarter ran 14 weeks rather than 13 (fiscal 2026 had 53 weeks), which inflates it relative to the 13-week quarter that follows (Micron 10-K). Fiscal 2026 revenue was $133.2 billion, up 256 percent. SK hynix reported a 76 percent operating margin for the second calendar quarter (SK hynix). Samsung's preliminary third-quarter operating profit was about 107.4 trillion won, against 12.17 trillion a year earlier (Samsung).

The cost lands on devices. Gartner counted 55.4 million PCs shipped in the third quarter, down 22.4 percent, and said DRAM costs had "surged five times year-over-year in the second quarter" (Gartner). In February, Raspberry Pi raised prices on its 2 GB and larger boards, citing "an unprecedented rise in the cost of LPDDR4 memory" (Raspberry Pi). TrendForce reports PC brands cutting SSD capacity in mainstream models to hold down costs.

The base rate: memory upcycles end, and they end the same few ways

Memory chips have a long record of booms followed by busts. In 2017 and 2018 DRAM sales grew 76.8 and 36.4 percent (SIA, SIA); the run ended after nine consecutive quarters of price growth (EE Times), and DRAM sales fell 37.1 percent in 2019 (SIA). The 2021 boom turned when customers found themselves holding "a higher level of DRAM inventory than what is considered healthy" (TrendForce, 2021). In 2022 inflation hit consumer demand while suppliers kept adding output (TrendForce, 2022), and Micron's fiscal 2023 revenue fell from $30.76 billion to $15.54 billion with a negative gross margin (Micron).

Across those cycles the turn came from three mechanisms: customers who had over-ordered stopped buying, supply kept growing after demand softened, or a macroeconomic shock cut demand. Micron's 10-K lists similar factors among those that can force it to cut output: "a weak demand environment, industry oversupply, inventory surpluses".

By that clock, a cycle of the 2016–2018 length would end around the third quarter of 2027, the quarter this forecast covers. TrendForce dates the sharp increases to the second half of 2025 (TrendForce), so the third quarter of 2027 would be roughly the ninth consecutive quarter of rising prices. The case for the main scenario is that the three ways cycles end are weaker this time, for structural reasons.

Why the usual endings are blocked, for now

Over-ordering is harder to unwind when it is contracted. Micron has signed 26 multi-year supply agreements, which it estimates at over 35 percent of its revenue through 2030; most carry floor and ceiling price bands, and customer financial commitments total $32 billion (Micron prepared remarks). Nearly all customer deposits are scheduled to be repaid between 2029 and 2031 (Micron 10-K). In earlier cycles, customers could simply stop ordering once their own inventory was high. Under these agreements, a customer who stops still owes for the contracted volume. Micron's 10-K concedes the reverse risk: purchases made only to meet those obligations "may result in elevated customer inventories" and an "oversupplied memory and storage market."

Supply cannot surge before the buildings exist. Micron's ID1 fab in Idaho is due to begin wafer output in mid-2027 and ID2 in late 2028, and the company warns that new fabs become "more meaningful starting a few quarters after initial output." SK hynix plans to open its first Yongin cleanroom in early 2027 (SK hynix). Micron expects industry DRAM bit shipments to grow "in the low-20s percentage range" in both 2027 and 2028, with the industry "supply constrained in both years."

High-bandwidth memory is one reason the shortage reaches ordinary DRAM, because it crowds out conventional capacity; TrendForce attributes the sharp rise in conventional DRAM contract prices to cloud operators' aggressive extra orders for general-purpose servers (TrendForce). HBM, the stacked memory beside AI accelerators, needs far more wafer area per bit: TrendForce estimates it takes about 22 percent of the top three suppliers' DRAM wafer input by the end of 2026 for about 9 percent of DRAM bits, rising to 30 percent of wafers by the end of 2027 (TrendForce). Its blended HBM price is projected to rise 121 percent in 2027 (TrendForce), and Micron says it has contracted "the vast majority" of its 2027 HBM supply at significantly higher prices. TrendForce expects capital spending by the nine largest cloud operators to pass $1.5 trillion in 2027, up 60 percent, a figure TrendForce says partly reflects higher component prices and supplier prepayments (TrendForce), and server CPU shipments to grow by more than 40 percent (TrendForce), which pulls ordinary server DRAM along with the accelerators.

The most likely course through mid-2027

The increases keep shrinking without turning negative. TrendForce's first-quarter forecast, due in early January, shows a smaller rise than the 10 to 15 percent expected now. Long-term agreements make prices adjust slowly in both directions: TrendForce already notes that some suppliers' server prices lag the market because of terms "stipulated in long-term agreements." PC DRAM is likely to stay especially tight, because TrendForce expects that "PC DRAM supply could decline in 2027 as more capacity shifts to server applications." Micron's guidance for its fiscal first quarter, $61.5 billion plus or minus $1.5 billion with sequential revenue growth each quarter of fiscal 2027 and "a more moderate rate of price increases", describes this path.

By mid-2027 the first new cleanroom space is producing, but at low volume, and it is ramping into contracts already signed. TrendForce's forecast for the third quarter, published around the start of July as in 2026, then shows a small rise: low single digits, possibly with a range that starts just above zero.

The rivals, and what would let them win

A forecast of flat or falling prices for the third quarter is the main rival, at 35 percent. The most likely route there is a reallocation of supply rather than a drop in demand. Micron's 10-K warns that "if demand for HBM weakens and suppliers shift capacity from HBM to conventional DRAM, this could result in a significant increase in conventional DRAM supply." The conditions for such a shift already exist: TrendForce reports that HBM profitability fell below that of DDR5 server modules in the first quarter of 2026, although Micron says 2027 HBM contracts at higher prices are "narrowing the gross margin gap with conventional DRAM"; and that GPU and ASIC makers are considering less HBM per device, citing supply rather than weak demand. Add demand pulled forward into 2026 by buyers fearing higher 2027 prices (TrendForce), consumers at what TrendForce calls their "affordability limit", and China's CXMT with capacity of about 300,000 wafers a month with plans that could take it above 600,000 (TrendForce, citing Reuters), and a flat third quarter is plausible.

The remaining 5 percent covers TrendForce publishing no comparable conventional-DRAM forecast for the quarter by July 31, 2027.

The strongest argument against the main scenario is the one this cycle's record makes about its forecaster: TrendForce has erred, and erred low, by wide margins. The direction of that error favours the main claim, but it also means that a turn, when it comes, may show up in spot markets before it reaches TrendForce's contract forecasts.

Dated register

These secondary predictions are published with the forecast and will be checked against the sources named; only the headline claim is scored automatically.

1. SK hynix reports a third-quarter 2026 operating margin of at least 75 percent on October 28, 2026 (80 percent).
2. Micron's fiscal first-quarter 2027 revenue, reported in December 2026, is at least $60.0 billion, the bottom of its guidance range (90 percent).
3. Micron guides fiscal second-quarter 2027 revenue above its fiscal first-quarter actual (80 percent).
4. TrendForce's first forecast for first-quarter 2027 conventional DRAM contract prices has a range whose lower bound is below 10 percent (60 percent).
5. Gartner reports fourth-quarter 2026 worldwide PC shipments down year on year (90 percent).
6. Micron states in its fiscal fourth-quarter 2027 results, around late September 2027, that ID1 has begun wafer output (70 percent).

What to watch

The weekly spot market is the earliest signal. The mainstream DDR4 8Gb spot price was $46.357 on October 20, up 0.84 percent on the week (TrendForce), after a period TrendForce described as "moving sideways" (TrendForce); sustained weekly declines would come before any turn in contract prices. Micron's days of inventory, 129 at the end of its fiscal fourth quarter and up nine days (Micron) and attributed by Micron partly to build-ahead before older nodes are retired, would show whether stock is building up at the supplier. Any change in the HBM layer count per accelerator, announced by Nvidia or its rivals, would show whether wafers are about to move back to conventional DRAM.

The unresolved fact that would most change this probability is how much HBM capacity can be moved back to conventional DRAM, and how quickly. If the suppliers can convert that capacity within a few quarters, the shortage could end soon after HBM demand eases. If conversion is slow, prices keep rising until the new fabs are running.

What this means

For PC, phone and server buyers, the planning case is that memory stays expensive through mid-2027, with smaller increases rather than relief. Contracted supply is worth more than spot flexibility while that holds, and product specifications set in 2026 to absorb memory costs are unlikely to need reversing before the second half of 2027.

For investors, the suppliers' margins depend less on the next quarter's price than on whether HBM capacity can move back to conventional DRAM quickly. That conversion speed, and the weekly spot price, are the earliest signs that the cycle has turned.

For policymakers counting on new domestic fabs, the first output arrives in mid-2027 and ramps over several quarters, so it cannot ease this shortage before then.