Conventional DRAM contract prices rose 90% to 95% in the first quarter of 2026 and a further 58% to 63% in the second, the steepest sustained memory price move in the industry’s modern history. The cause is not a supply accident but an allocation decision: memory makers prioritised high-capacity RDIMMs for AI servers and squeezed everyone else. The third quarter is finally moderating to 13% to 18% for server DRAM, though TrendForce already forecasts a further server shortage in 2027 with RDIMM bit supply growing only 15% to 20%. The figures below come from TrendForce contract pricing analysis and IDC market forecasts.
TL;DR
- Conventional DRAM contract prices rose 90% to 95% quarter over quarter in Q1 2026 (TrendForce)
- Prices rose a further 58% to 63% in Q2 2026 (TrendForce)
- Server DRAM is forecast to rise 13% to 18% in Q3 2026 (TrendForce)
- Every product category hit record quarter-over-quarter increases in Q1 2026 (TrendForce)
- The market remained undersupplied with suppliers revising quotations upward (TrendForce)
- Memory makers prioritised high-capacity RDIMMs for AI servers (TrendForce)
- PC and smartphone vendors were squeezed on allocation (TrendForce)
- A server DRAM shortage is already anticipated for 2027 (TrendForce)
- RDIMM bit supply is forecast to grow only 15% to 20% year over year (TrendForce)
- DDR5 profitability in 2026 was projected to surpass HBM3e (TrendForce)
- Long-term agreements cap increases for cloud service providers that secured them (TrendForce)
- IDC warns the PC market could shrink up to 9% in 2026 on memory costs (IDC)
- Total PC market value still rises to around USD 274 billion (IDC)
1. The Price Spike in Numbers
Sequential increases of this size are almost unheard of in a commodity component market. Contract prices rose 90% to 95% in the first quarter and another 58% to 63% in the second, and TrendForce described Q1 as producing record quarter-over-quarter increases across every product category simultaneously. Because these are sequential moves, the compounded first-half effect is roughly triple the starting point rather than the sum of the two percentages.
| Metric | Value | Source |
|---|---|---|
| Conventional DRAM contract price change, Q1 2026 | up 90% to 95% QoQ | TrendForce |
| Contract price change, Q2 2026 | up 58% to 63% QoQ | TrendForce |
| Server DRAM forecast, Q3 2026 | up 13% to 18% QoQ | TrendForce |
| Q1 2026 characterisation | record QoQ increases, all categories | TrendForce |
| Approximate compounded H1 multiple | roughly 3x | Derived from TrendForce figures |
| Market condition throughout | undersupplied | TrendForce |
| Supplier behaviour | quotations revised upward | TrendForce |
| Direction of the Q3 move | still up, sharply moderated | TrendForce |
The compounded row is arithmetic on the two sequential ranges and is illustrative rather than a TrendForce figure.
It is worth being clear about what a contract price is, because the number most people encounter is not this one. Contract prices are negotiated between memory manufacturers and large buyers, typically on quarterly cycles, and they set the floor for what a device maker pays per gigabyte. Retail module prices move separately, often later and by different magnitudes, because distributors hold inventory bought at older prices and because branded modules carry margin that dampens the swing. A buyer shopping for a memory upgrade in mid-2026 was therefore seeing the consequences of contracts negotiated months earlier, which is why consumer-facing price shock lagged the industry data by roughly a quarter. Anyone reconciling a retail price against a TrendForce figure is comparing two different markets. Source: TrendForce on Q1 2026 memory price increases.
2. AI Servers Took the Supply
This was an allocation outcome, not a manufacturing failure. Memory makers prioritised server applications, and specifically high-capacity RDIMMs for AI servers, which drove across-the-board increases and squeezed PC and smartphone vendors out of the queue. Cloud service providers locked in supply through long-term agreements, insulating themselves and concentrating the price pressure on everyone without one.
| Metric | Value | Source |
|---|---|---|
| Priority application for makers | server, specifically AI | TrendForce |
| Product prioritised | high-capacity RDIMMs | TrendForce |
| Segments squeezed on allocation | PC and smartphone vendors | TrendForce |
| Mechanism used by cloud providers | long-term agreements | TrendForce |
| Effect of those agreements | caps price increases for holders | TrendForce |
| Who absorbs the spot moves | buyers without agreements | TrendForce |
| Q1 2026 framing by TrendForce | makers prioritise server applications | TrendForce |
| Resulting price direction across categories | up across the board | TrendForce |
The long-term agreement dynamic is the most under-covered part of this story: the same buyers driving the demand shock are the ones most protected from its price consequences. Data centre context sits in our data center statistics. Source: TrendForce on memory makers prioritising server applications.
3. The Q3 Moderation
The second half looks materially different, and the reason is as much arithmetic as it is demand. Server DRAM contract prices are forecast to rise 13% to 18% in the third quarter, down from 58% to 63% in the second, with TrendForce citing weaker consumer demand and a high comparison base. A slower rate of increase on a tripled base is still an increase.
| Metric | Value | Source |
|---|---|---|
| Server DRAM forecast, Q3 2026 | up 13% to 18% QoQ | TrendForce |
| Prior quarter for comparison | up 58% to 63% | TrendForce |
| Cited reasons for moderation | weaker consumer demand, high base | TrendForce |
| Direction of AI server demand | still supporting prices | TrendForce |
| Effect of long-term agreements | capping increases | TrendForce |
| Consumer DRAM condition | still in shortage | TrendForce |
| Overall market state | undersupplied | TrendForce |
| Price direction, Q3 2026 | up, moderating | TrendForce |
Coverage that reads the Q3 moderation as prices falling is misreading a second derivative: the rate of increase slowed, the price level did not. A buyer negotiating a contract in the third quarter still paid more than one negotiating in the second, and considerably more than one negotiating a year earlier. Relief would require an actual decline, and nothing in TrendForce’s outlook forecasts one within the reporting horizon. Source: TrendForce on Q3 2026 memory contract prices.
4. What It Does to Everything Downstream
Memory is in every computing device, so a tripling of its cost propagates fast. IDC has warned the PC market could contract by up to 9% in 2026 because of memory pricing, with even its moderate scenario showing a 5% decline, while total market value still rises to around USD 274 billion. Fewer machines at higher prices is the shape of the whole downstream effect.
| Metric | Value | Source |
|---|---|---|
| Worst-case PC market contraction, 2026 | up to 9% | IDC |
| Moderate-case contraction | 5% | IDC |
| Cause cited | skyrocketing RAM pricing | IDC |
| Total PC market value, 2026 | around USD 274 billion | IDC |
| Direction of market value | rising despite unit decline | IDC |
| Segments squeezed on memory allocation | PC and smartphone | TrendForce |
| Underlying driver | AI infrastructure demand | IDC, TrendForce |
| Effect on bill of materials | rising | IDC |
The circularity is worth naming: AI demand is the reason NPU-equipped machines exist and the reason fewer machines of any kind will ship. Downstream detail sits in our AI PC statistics and PC market statistics. Source: IDC’s 2026 PC market warning.
5. 2027 Is Already Forecast Short
The most consequential figure in this dataset is about a year that has not started. TrendForce anticipates a server DRAM shortage in 2027, with total RDIMM bit supply growing only 15% to 20% year over year against faster growth in server CPU shipments. Supply expanding more slowly than the systems it must populate is the definition of a structural shortage rather than a cyclical one.
| Metric | Value | Source |
|---|---|---|
| Anticipated 2027 condition | server DRAM shortage | TrendForce |
| Forecast RDIMM bit supply growth | 15% to 20% YoY | TrendForce |
| Comparison | lagging server CPU shipment growth | TrendForce |
| Nature of the imbalance | structural, not cyclical | TrendForce |
| DDR5 profitability, 2026 | projected to surpass HBM3e | TrendForce |
| Driver of that inversion | tight conventional DRAM supply | TrendForce |
| Current market state | undersupplied | TrendForce |
| Q3 2026 price direction | still rising | TrendForce |
DDR5 out-earning HBM3e is the clearest signal that conventional memory, not the exotic high-bandwidth tier, became the binding constraint of this cycle. The industry spent two years building capacity and narrative around high-bandwidth memory for AI accelerators, on the reasonable assumption that this was where scarcity and margin would concentrate. What actually happened is that every AI server also needs a great deal of ordinary system memory, and that demand landed on the same fabs serving phones and laptops. The bottleneck turned out to be in the commodity tier nobody was defending, which is a useful reminder that capacity planning tends to over-index on the component everyone is talking about. Silicon supply context sits in our semiconductor industry statistics. Source: TrendForce on DDR5 contract prices and 2026 profitability.
Summary: RAM and DRAM by the Numbers
| Metric | Value | Source |
|---|---|---|
| Contract price change, Q1 2026 | up 90% to 95% QoQ | TrendForce |
| Contract price change, Q2 2026 | up 58% to 63% QoQ | TrendForce |
| Server DRAM forecast, Q3 2026 | up 13% to 18% QoQ | TrendForce |
| Approximate compounded H1 multiple | roughly 3x | Derived |
| Q1 2026 category coverage | record increases across all | TrendForce |
| Priority application | AI servers | TrendForce |
| Product prioritised | high-capacity RDIMMs | TrendForce |
| Segments squeezed | PC and smartphone vendors | TrendForce |
| Protection mechanism for hyperscalers | long-term agreements | TrendForce |
| Reasons for Q3 moderation | weak consumer demand, high base | TrendForce |
| Anticipated 2027 condition | server DRAM shortage | TrendForce |
| Forecast RDIMM bit supply growth | 15% to 20% YoY | TrendForce |
| DDR5 profitability versus HBM3e, 2026 | projected higher | TrendForce |
| Worst-case PC contraction, 2026 | up to 9% | IDC |
| Moderate-case PC contraction | 5% | IDC |
| Total PC market value, 2026 | around USD 274 billion | IDC |
| Market condition throughout | undersupplied | TrendForce |
Methodology and Sources
- Contract price movements by quarter, supply allocation, long-term agreement dynamics, the Q3 moderation, DDR5 profitability, and the 2027 shortage forecast all come from TrendForce press analysis (Q1 2026 outlook, server prioritisation, Q2 2026 contract increases, consumer DRAM shortages, Q3 2026 outlook, long-term agreements and 3Q26 server DRAM, DDR5 and HBM3e profitability).
- Downstream PC market effects come from IDC forecasts as reported contemporaneously (Tom’s Hardware, The Register).
- Data watch: TrendForce figures are contract prices negotiated between suppliers and large buyers, not spot prices and not retail prices, so consumer memory module prices move on a different schedule and by different magnitudes. All quarterly percentages are sequential, and adding them together overstates the cumulative move while ignoring compounding understates it; the roughly 3x figure here is illustrative arithmetic rather than a TrendForce estimate. The Q3 2026 figure and the entire 2027 outlook are forecasts, and TrendForce has revised its memory outlooks repeatedly through this cycle. IDC’s PC contraction range of 5% to 9% is wide enough that downstream unit effects remain genuinely uncertain. Rows marked as derived are arithmetic on published figures.
- Last updated: August 2, 2026. We update this roundup quarterly as TrendForce publishes new contract pricing analysis.