The 2026 PC Build Cost Crisis: 32GB Memory Breaks Price, SSD Shortage May Last Years, RTX 5090 Tops $5,000, and the 5080 SUPER Delayed#

Zixing | September 24, 2026

Hi everyone, I’m Zixing.

Last month I broke down the monitor price hike in detail, and readers kept asking in the comments: monitors only went up 7%, so why have memory, SSDs, and GPUs - the real “big-ticket items” - risen in ways that make no sense? I went through a round of search data, and this week’s market can honestly be described with one word: crisis - 32GB memory retail prices have multiplied several times over, 2TB SSD prices are pushing $500, and the RTX 5090 has surged past twice its MSRP.

This article won’t discuss scalpers or “hoarders”; instead, it pulls the view back to the source of the costs. The core of this price surge is not speculative hype but AI infrastructure “siphoning” memory and storage capacity - a genuine reallocation of production capacity. Below I lay out the math in five steps.

Table of Contents#

1. Memory: 32GB Kits Up About 4.9x in a Year
2. Storage: The NAND Shortage May Last Years
3. GPUs: RTX 5090 at a 135% Premium, 5080 SUPER Delayed
4. AI Siphoning: The Real Driver of This Price Surge
5. Build and Upgrade Decisions by Budget
6. Zixing’s Wait-or-Buy Decision Framework

1. Memory: 32GB Kits Up About 4.9x in a Year#

Start with the most painful item: memory. It is the single product with the most extreme increase in this round.

1.1 Price Trend of Mainstream 32GB DDR5-6000 Kits#

From $110–140 in Q3 2025 to about $392 in August 2026, a mainstream 32GB DDR5-6000 kit’s retail price has nearly tripled. On 3DCenter’s DDR5 price index, it comes to about 486% of the July 2025 baseline - nearly 4.9x in a year.

Note that 3DCenter’s index reflects German retail figures - a directional reference, not a global average; prices in China are also affected by channels and exchange rates and don’t fully match overseas prices. But the trend is the same.

1.2 China Market Comparison (2026-08)#

The China market makes the point even more clearly. 32GB DDR5 went from about RMB 900 to about RMB 4,000, an increase of over 300%; 16GB single sticks/kits climbed from 350–400 yuan to 1,500 yuan; 1TB SSDs surged from about 410 yuan to 950–1,000 yuan.

For most players, memory and SSDs used to be the two items you “cared about least” when building a PC; now they are the two components with the steepest increases.

1.3 Why Memory Increased So Dramatically#

Memory rose so sharply because wafer supply was substantially pulled away. The three mainstream manufacturers (Samsung, SK hynix, Micron) have allocated about 70%–90% of advanced capacity to high-margin HBM (the high-bandwidth memory used in AI chips), squeezing consumer DRAM. At the upstream contract-price level, TrendForce data shows general DRAM contract prices rose 58%–63% quarter over quarter in Q2 2026, and multiple firms expect full-year DRAM increases of 125%–280%. Wafer prices passed through to retail, producing the numbers above.

2. Storage: The NAND Shortage May Last Years#

Beyond memory, storage is the second hardest-hit area.

2.1 SSD Price Trends#

NAND flash prices have roughly doubled since the end of 2025. 2TB NVMe SSDs rose from $120–150 a year ago to $300–480; by Gamers Nexus’ measure, NVMe is up about 115% and SATA about 75%.

2.2 Upstream Contract Prices and Full-Year Expectations#

Upstream contract prices moved earlier and harder than retail:

NAND’s quarterly increase reached as high as 70%–75%, pulling the full-year expectation into the 138%–250% range. This table also shows that retail increases are not the end of the story - upstream wafer price pressure is still being released.

2.3 How Long Will the Shortage Last?#

This is the question players care about most, and the key one for the “wait-and-see” crowd. Forecasts differ in methodology but agree in direction:

All three basically point to H2 2027 as the window for supply-demand rebalancing, while the SK hynix CEO’s warning is more aggressive - the shortage could last until 2030. This means storage costs are unlikely to fall meaningfully for some time.

3. GPUs: RTX 5090 at a 135% Premium, 5080 SUPER Delayed#

GPUs are the other climax of this price surge, and also the category with the most “delay” drama.

3.1 RTX 5090 Retail Premium#

The RTX 5090’s August 2026 Newegg median price was about $4,699.99, roughly 135% above the $1,999 MSRP; some listings reach $4,929–5,499. The China market is just as extreme: spot prices start around RMB 30,000, with some blower models at RMB 35,000–37,000. The RTX 5080 has also risen from about RMB 7,300 to about RMB 12,000.

One main driver of GPU price increases is rising GDDR7 memory costs. NVIDIA notified board partners in May that the 5090/5090D V2 SKUs would be raised by about another $300.

3.2 RTX 50 SUPER Series Delayed#

The RTX 50 SUPER series was originally due in 2026 but is now paused or postponed. These cards carry about 50% more VRAM than current models (5080 Super at 24GB, 5070 Ti Super at 24GB, 5070 Super at 18GB). Note that sources describe the launch window differently: some say “launch is paused with no set date”, others say “postponed to CES 2027 (January 2027)”, and more aggressive reports claim the whole RTX 50 Super refresh (including the completed 24GB 5080 Super design) has been shelved and will not be produced in 2026; NVIDIA has not confirmed any launch date. So it should not be treated as a single fixed milestone.

3.3 Reasons for the Delay and Production Cuts#

The core reason for the delay is memory supply. These SUPER cards rely on denser 3GB GDDR7 modules costing about 3x the current 2GB modules, and the same 3GB modules are used in NVIDIA’s AI cards (such as the RTX PRO 6000 Blackwell and Rubin CPX), with NVIDIA prioritizing the higher-margin AI business. In addition, industry reports say NVIDIA cut current RTX 50 series output by about 15%–20% (at least through Q3 2026), and the RTX 5060 was pulled from retail for at least 6 months - though none of this has official statements and should be treated as industry reporting, not established fact.

4. AI Siphoning: The Real Driver of This Price Surge#

Put the memory, storage, and GPU tables side by side and it becomes clear they are not three separate stories but different expressions of one root cause. I call that root cause “AI siphoning”.

4.1 How AI Infrastructure Pulls Away Memory and Storage Capacity#

IDC expects AI data centers to consume about 70% of global memory capacity in 2026, versus only 20%–30% in 2022. A single AI server carries about 8–10x the DRAM of a traditional server. High-margin HBM is drawing away advanced capacity in volume, and consumer DRAM and NAND capacity are inevitably squeezed.

4.2 Why This Isn’t “Scalper Hype”#

This price surge is different from past ones. It is not a bubble pumped up by end-market speculators hoarding goods, but a reallocation of upstream capacity:

• The first-order cause of the price increases is capacity being preferentially allocated to the higher-margin AI business, not channel withholding;
• So the window for prices to fall doesn’t depend on “when scalpers unload inventory” but on when capacity supply rebalances - which is exactly why Gartner, SK hynix, and Bernstein all point to H2 2027;
• The same logic shows up in GPUs: GDDR7 memory being grabbed by AI cards is what delayed the SUPER series.

Once you understand the “AI siphoning” throughline, you won’t be dragged around by simple “prices up/prices down” sentiment, and you can judge the long-tail nature of this cycle.

4.3 The Increases Have Passed Through to Complete Systems#

Storage price increases don’t just affect people buying components individually. Data shows memory’s share of laptop system costs has surged from about 15% to about 35% - meaning this round of increases has passed through from upstream wafers to system makers, and ordinary consumers buying prebuilts or laptops face the same cost pressure.

5. Build and Upgrade Decisions by Budget#

With the cost structure understood, the question is how to act. I’ll break it down by budget tier. One caveat up front: what follows is decision guidance based on verified prices and forecasts, not purchase advice; whether the “wait-and-see” crowd holds or buys depends on whether you’re willing to factor in the H2 2027 rebalancing window.

5.1 Budget-Sensitive (Complete System Under About 5,000–7,000 yuan)#

5.2 Balanced Value (Complete System 8,000–12,000 yuan)#

5.3 High-End/Flagship Players (Complete System Over 15,000 yuan)#

• Players at this tier are relatively insensitive to per-component costs, but if the target is an RTX 5090, know that the retail premium is already at 135%, with some listings breaking $5,000.
• If you’re chasing a flagship, always verify the gap between current prices and MSRP, and avoid paying an excessive premium driven by “shortage anxiety”.

5.4 A “Value Window” Worth Watching#

Against a backdrop of universal price increases, one direction is relatively special: old-platform flagship inventory. Old flagships from the RTX 4000 and RX 7000 series are seen by some channels as one of the few remaining “real value windows” - though that window will close quickly, and the longer you wait the harder it gets. If you can accept last-generation platform performance, this tier is worth watching, but again, verify actual inventory and prices.

6. Zixing’s Wait-or-Buy Decision Framework#

Finally, as a hardware editor, let me distill this round’s core judgments into a few actionable principles:

1. Separate “needs” from “can wait”: memory and SSDs are the fastest-rising components this round, but most firms point to H2 2027 as their rebalancing window. If your current build works, don’t rush to buy at the top; if you truly need them, accept the reality of short-term cost increases.

2. Don’t apply “wait-and-see” thinking to this cycle: in the past, waiting brought price cuts, but this time it’s a long-cycle capacity reallocation driven by AI infrastructure, and a sharp drop is unlikely soon. Whether to wait depends on whether you can accept H2 2027 as your time anchor.

3. Use “AI siphoning” as your baseline: as long as HBM keeps attracting advanced capacity, consumer DRAM and NAND supply will stay tight. Rather than watching price news, watch capacity allocation and institutional views on the cycle window.

4. Prioritize the components with the tightest supply: on balance, memory, SSDs, and VRAM are the three directions with the tightest supply and the most persistent upward pressure, and the ones most worth weighing rationally.

5. Be cautious with “unverified” reports: the exact RTX 5080 SUPER launch date, the unit cost of 3GB GDDR7, power figures, and the like are mostly rumors or industry reports; NVIDIA has not confirmed them, so don’t build decisions on them.

I hope this accounting helps you see the real logic behind 2026’s rising build costs - not panic, but rational trade-offs. Buy with clarity, and wait with clarity too.

— Zixing, NTGame Hardware & Technology Editor