In my quest to investigate the rising RAM and storage prices, as I am looking to upgrade my workstation and build a NAS with plenty of storage for backups and media, I've been researching, ironically with AI, the market prices, the current and forecasted trends, and the news surrounding these market dynamics. I'll start with the TLDR and include the full report below in case anyone is interested in the detailed findings. I figured someone might find this useful if they're considering hardware purchases over the next 12 months.

TLDR: DDR5 RAM will rise, DDR4 stabilizing, NVMe SSDs stabilizing, SSD price relief possible in Q2 2027. Don't expect to see 2024-2025 pricing for anything ever again.

Full report:

CONSUMER RAM & STORAGE PRICING OUTLOOK — SEPTEMBER 2026

Short version:

The memory market is still in a severe upcycle, but RAM and SSDs are starting to behave differently.

My current view is:

• DDR5 is still more likely to rise than fall over the next 6–12 months.
• DDR4 is likely to remain very expensive, but prices may be closer to a plateau.
• Consumer NVMe SSDs look more likely to stabilize first.
• Meaningful SSD price relief becomes increasingly plausible during 2027, especially in the second half.
• I would not expect a return to 2024/early-2025 RAM pricing anytime soon.

My overall probability estimate for the next 6–12 months:

• Continued price increases: ~45%
• Prices remain at a high plateau: ~40%
• Meaningful price declines: ~15%

For DDR5 specifically, I think the probability of further increases is higher than that — approximately 60–65%.

EXECUTIVE SUMMARY

Retail memory pricing has increased enormously over the last year.

According to MemRadar's September 13, 2026 index:

• DDR5: +341% year-over-year
• DDR4: +154%
• NVMe SSDs: +133%
• SATA SSDs: +156%

However, the last six months look very different:

• DDR5: +15.7%
• DDR4: +1.9%
• NVMe SSDs: roughly flat
• SATA SSDs: +15.9%

So the market is no longer experiencing one uniform surge.

DDR5 is still accelerating.

DDR4 has largely stopped rising, although it remains extremely expensive.

Mainstream NVMe SSD pricing is starting to look more like a high plateau than a continuous surge.

European pricing is even more extreme.

3DCenter's German retail index puts DDR5 at 544% of its July 2025 level as of September 2026.

For comparison:

• August 2026: 486%
• September 2026: 544%

So the previously reported "486% of July 2025 pricing" figure was accurate, but it is already one month out of date.

3DCenter's September 2026 indexes are approximately:

• DDR5: 544
• DDR3/DDR4: 292
• Internal SSDs: 225
• Internal HDDs: 257

Baseline July 2025 = 100.

WHEN DID THE CURRENT PRICE SURGE BEGIN?

The major retail inflection point appears to have been around October 2025.

German DDR5 pricing increased approximately:

• October 2025: +15.8% month-over-month
• November: +49.5%
• December: +93%
• January 2026: +27.6%

After that, there was a temporary pause.

Around February through June 2026, some RAM categories flattened or even corrected slightly.

Then DDR5 started rising again:

• July: +6.1%
• August: +9.2%
• September: +11.9%

So the pattern has basically been:

1. Initial surge in late 2025
2. Huge acceleration in November–January
3. Temporary stabilization in spring 2026
4. Renewed DDR5 increases in summer 2026

SSDs are now showing a different pattern. Some mainstream 1TB SSDs have actually started falling slightly even while larger capacities remain expensive.

REPRESENTATIVE EUROPEAN RETAIL PRICES

These numbers come from 3DCenter/Geizhals and represent the cheapest available products meeting a given specification rather than one specific brand.

32GB DDR4-3200, 2x16GB:

• September 2025: €69
• December 2025: €173
• January 2026: €219
• March 2026: €200
• June 2026: €195
• September 2026: €210

Approximately +204% over 12 months.

32GB DDR5-6000, 2x16GB:

• September 2025: €88
• December 2025: €301
• January 2026: €395
• March 2026: €362
• June 2026: €367
• September 2026: €430

Approximately +389%.

64GB DDR5-6000, 2x32GB:

• September 2025: €170
• December 2025: €547
• January 2026: €699
• March 2026: €608
• June 2026: €638
• September 2026: €849

Approximately +399%.

1TB PCIe 4.0 NVMe SSD with DRAM:

• September 2025: €70
• December 2025: €106
• January 2026: €130
• March 2026: €156
• June 2026: €156
• September 2026: €156

Approximately +123%.

2TB PCIe 4.0 NVMe SSD with DRAM:

• September 2025: €117
• December 2025: €157
• January 2026: €199
• March 2026: €210
• June 2026: €245
• September 2026: €299

Approximately +156%.

The important thing here is that 1TB mainstream NVMe pricing has already stopped rising in this dataset, while 2TB storage is still under more pressure.

U.S. RETAIL PRICING

Tom's Hardware's September 2026 tracker shows roughly:

DDR5:

• 32GB DDR5-6000: around $400–$420 at the low end
• 64GB DDR5-6000: around $889

For comparison, many 32GB DDR5 kits were around $100–200 in October 2025.

DDR4:

• 32GB DDR4-3200: approximately $190

That compares with roughly $60–90 before the shortage accelerated.

So DDR4 is no longer a cheap escape route from expensive DDR5.

For SSDs, mainstream U.S. PCIe 4.0 pricing is roughly:

• 1TB: around $149–240 depending on model
• 2TB: around $248–389

Availability still exists. The problem is mainly price rather than shelves being completely empty.

WHAT IS CAUSING THIS?

The biggest driver is AI/data-center demand.

The important distinction is that HBM isn't just consuming memory bits. It is disproportionately consuming wafer capacity.

TrendForce estimates HBM will account for approximately:

• 18% of top-three DRAM manufacturers' wafer input at the end of 2025
• 22% at the end of 2026
• 30% at the end of 2027

But HBM only represents approximately:

• 8% of DRAM bit output in 2025
• 9% in 2026
• 13% in 2027

That mismatch matters.

HBM consumes much more manufacturing capacity per bit than ordinary PC DRAM.

TrendForce also expects HBM plus server RDIMM memory to consume around 51% of global DRAM bit supply in 2026.

That means consumer desktops and laptops are competing for what is increasingly becoming the lower-priority part of the DRAM market.

WHAT ABOUT THE CLAIM THAT AI DATA CENTERS CONSUME 70% OF MEMORY PRODUCTION?

There is some basis for it, but I would not quote "70%" without qualification.

CSIS has cited an estimate of roughly 70% of worldwide memory output being consumed by data centers.

However, different reports define "memory output" differently.

It may include:

• HBM
• Server DRAM
• Enterprise NAND
• Other data-center memory

TrendForce's figures are easier to interpret:

• Around 51% of DRAM bits going toward HBM plus server RDIMMs
• Around 22% of DRAM wafer input going into HBM alone by the end of 2026

So a safer statement is:

"Data centers are now absorbing roughly half or more of relevant DRAM production, with some broader estimates approaching 70% depending on how memory output is defined."

WHY CAN'T MANUFACTURERS JUST MAKE MORE RAM?

They are expanding capacity, but semiconductor fabs take years to build and ramp.

This is the biggest reason prices may remain elevated.

TrendForce says several DRAM manufacturers have more capacity planned for 2027, but meaningful additional output is not expected until the second half of 2027, with larger contributions coming in 2028.

Micron has announced more than $250 billion in U.S. investment through 2035.

Kioxia and Sandisk are also expanding NAND capacity in Japan.

But these are multi-year projects.

They do not suddenly produce millions of cheap consumer DIMMs next quarter.

WHY ARE MANUFACTURERS PRIORITIZING AI PRODUCTS?

Because they are much more profitable.

Samsung, SK Hynix and Micron are all emphasizing:

• HBM
• Server DRAM
• Enterprise SSDs
• Long-term hyperscaler contracts

Consumer RAM is simply less attractive economically.

Sandisk provides a good example on the NAND side.

Its fiscal Q4 2026 results showed data-center revenue increasing dramatically while consumer revenue declined sequentially.

That gives suppliers every incentive to allocate their best capacity toward enterprise customers rather than cheap consumer SSDs.

IS CONSUMER DEMAND STARTING TO BREAK?

Yes.

There are now clear signs of demand destruction.

IDC expects global PC shipments to decline around 11.3% in 2026.

At the same time, average PC selling prices are expected to rise around 17–18%.

That means consumers are responding to higher prices by buying fewer PCs.

Counterpoint has also reported manufacturers:

• Reducing RAM/storage specifications
• Concentrating limited memory supply in premium products
• Cutting exposure to low-margin entry-level models

This is important because demand destruction can stop prices from rising even before supply becomes plentiful.

That appears to be what is beginning to happen in consumer SSDs.

DDR4 IS NOT REALLY A BUDGET REFUGE ANYMORE

DDR4 is currently cheaper than DDR5 in absolute terms, but it is still massively more expensive than it was.

The interesting thing is that DDR4's upward momentum has largely stopped.

MemRadar has DDR4 slightly down over the last three months.

German retail data also shows relatively flat DDR4 pricing since early 2026.

So DDR4 may be closer to a plateau.

However, long-term DDR4 production is being reduced, which means there is no guarantee that prices return to old lows.

WHAT IS HAPPENING WITH SSDs?

This is where the outlook becomes more optimistic.

TrendForce's newest client-SSD research says OEM customers have reached the limit of what they are willing to pay.

Buyers are increasingly purchasing only the minimum quantities they need.

TrendForce therefore expects Q4 2026 client SSD contract prices to plateau at high levels.

More importantly, TrendForce expects NAND supply growth to begin exceeding demand during the second half of 2027.

That makes NAND/SSDs the strongest candidate for eventual price relief.

WHAT ABOUT THE PHISON CEO SAYING SSD PAIN COULD LAST UNTIL 2030?

That warning should be interpreted carefully.

Phison CEO Pua Khein-Seng reportedly said it can take roughly four years for meaningful new NAND capacity to come online.

That is believable from a manufacturing perspective.

But it does NOT necessarily mean:

"Consumer SSD prices will rise every year until 2030."

Those are different statements.

Prices can stabilize or fall because:

• Consumer demand collapses
• OEMs reduce orders
• Existing fabs improve yields
• Higher-density NAND increases bit output
• Inventory builds
• Enterprise demand slows

TrendForce currently expects NAND tightness to start easing in the second half of 2027.

So there is genuine disagreement over how long the shortage will remain severe.

WHAT DO ANALYSTS CURRENTLY EXPECT?

TrendForce:

DRAM remains structurally tight into 2027.

Meaningful new DRAM supply arrives mainly in late 2027 and 2028.

NAND conditions should begin easing earlier, potentially in the second half of 2027.

Gartner:

Expects memory pricing to remain very strong.

It does not expect meaningful broad relief until around late 2027.

IDC:

Expects the memory shortage to continue through 2026 and well into 2027.

Meaningful relief for the PC market may not arrive until the end of 2027.

Counterpoint:

Recently revised its Q3 DRAM forecast upward.

Its expected increase went from roughly 5–10% to approximately 10–20% QoQ.

That revision happened within roughly one month.

J.P. Morgan:

Expects the memory shortage to take years to fully unwind.

SK Hynix CEO:

Has warned that 2027 could potentially become the industry's worst-ever memory shortage year.

UBS and Bank of America:

According to Reuters, both have models showing undersupply potentially extending into 2028.

So the broad analyst consensus is still:

DRAM stays tight longer than NAND.

A WARNING ABOUT FORECAST ACCURACY

This market has repeatedly made analysts look foolish.

TrendForce's late-2024 outlook actually predicted weaker DRAM pricing during 2025 because of:

• Weak consumer demand
• Inventory
• Chinese DRAM capacity

That forecast was invalidated as the market shifted toward AI/server allocation.

By September 2025, TrendForce was forecasting renewed price increases.

Counterpoint has also revised 2026 forecasts upward very rapidly.

So I would not trust any exact forecast such as:

"DDR5 will fall 17% in Q2 2027."

The useful part of these forecasts is the direction and underlying supply/demand logic, not the exact percentage.

CURRENT LEADING INDICATORS

Several indicators still look bullish for DRAM.

1. Supplier inventories remain extremely low.

TrendForce says DRAM supplier inventories are historically low.

Low inventories make a sudden price collapse less likely.

2. New supply is going mainly to servers.

Incremental DRAM production is disproportionately being allocated to higher-margin enterprise customers.

3. Suppliers are signing long-term agreements.

SK Hynix and others are locking in multi-year contracts with large AI and cloud customers.

That means consumer buyers cannot simply outbid them for spare capacity.

4. AI infrastructure spending remains extremely strong.

TrendForce still expects very large hyperscaler capex growth in 2026 and 2027.

5. NAND capacity expansion is becoming more visible.

This is one reason SSD pricing may normalize sooner than DRAM.

6. OEM pushback is real.

PC shipments are falling while prices are increasing.

That will eventually impose a ceiling on what memory suppliers can charge consumer markets.

SCENARIO 1: PRICES KEEP RISING

Probability: approximately 45% overall.

More like 60–65% for DDR5 specifically.

Conditions:

• AI spending remains extremely strong
• HBM production keeps consuming more wafer capacity
• New fabs ramp slowly
• Server customers continue locking up supply
• Manufacturers maintain production discipline

Possible 6–12 month retail changes:

DDR5:
+20% to +45%

DDR4:
+5% to +25%

Consumer NVMe:
+5% to +20%

This is still the most likely individual scenario for DDR5.

SCENARIO 2: HIGH-PRICE PLATEAU

Probability: approximately 40%.

Conditions:

• AI demand remains strong
• Consumer demand weakens further
• OEMs refuse additional increases
• Existing inventory improves somewhat
• Supply remains tight enough to prevent a major crash

Possible changes:

DDR5:
-5% to +10%

DDR4:
-10% to +5%

Consumer NVMe:
-10% to +5%

This is my base case for consumer SSDs over the next several months.

SCENARIO 3: MEANINGFUL PRICE DECLINE

Probability: approximately 15%.

Conditions:

• AI capex growth slows significantly
• New capacity ramps faster than expected
• PC/smartphone demand weakens substantially
• Supplier inventories rebuild
• Enterprise buyers stop hoarding or signing aggressive long-term contracts

Possible price declines:

DDR5:
-15% to -30%

DDR4:
-20% to -35%

Consumer NVMe:
-20% to -35%

I think this becomes considerably more plausible during Q2–Q3 2027 than during late 2026.

MY PRODUCT-SPECIFIC OUTLOOK

DDR5:

Most vulnerable to further increases.

My approximate probabilities:

• Higher in 6–12 months: 60–65%
• Roughly flat: 25–30%
• Meaningfully cheaper: 10–15%

DDR4:

Most likely to plateau.

Production cuts prevent a major supply glut, but current pricing has already destroyed much of the demand.

NVMe SSDs:

Best candidate for price relief.

I would expect stabilization before I expect a dramatic crash.

A more meaningful decline becomes increasingly plausible during the second half of 2027.

SATA SSDs:

Poor value in many cases.

Low-volume legacy production can make SATA SSDs surprisingly expensive compared with NVMe.

HDDs:

Still useful for large bulk storage, but prices have also risen dramatically.

They are not immune to the broader data-center storage boom.

WHAT SHOULD A CONSUMER DO?

If building in the next 0–3 months:

Buy RAM when you see an acceptable price.

I would not postpone a real build specifically because you are expecting DDR5 to suddenly return to 2025 prices.

32GB is currently much more rational than overbuying 64GB if you do not actually need it.

High-capacity DDR5 has recently risen faster than mainstream 32GB kits.

If upgrading an existing DDR4 machine:

Buy opportunistically.

There is less urgency than with DDR5 because DDR4 has been relatively flat recently.

But I would not assume DDR4 will become cheap again simply because it is an older technology.

If buying SSD storage:

I would be more willing to wait.

Consumer SSDs are showing the clearest signs of reaching a pricing ceiling.

1TB mainstream NVMe appears to be one of the first categories where selective declines are already appearing.

1–2TB PCIe 4.0 consumer drives are probably the most likely segment to offer improved deals first.

If your build is 6–12 months away:

I would NOT buy every component today just because memory is expensive.

My preferred strategy would be:

• Buy DDR5 early if you find a genuinely good deal.
• Be more patient with SSDs.
• Do not overbuy RAM capacity.
• Do not assume SATA SSDs are cheaper than NVMe.
• Buy CPUs and GPUs according to their own market cycles rather than treating the whole PC as one commodity.

CURRENT "SWEET SPOTS"

There aren't many true bargains, but there are a few relatively better-value areas.

1TB mainstream PCIe 4.0 NVMe:

This is probably the category closest to stabilizing.

Some German 1TB SSD categories have already declined month-over-month.

2TB consumer NVMe:

Still expensive, but likely one of the first higher-capacity markets to improve if NAND conditions loosen.

32GB RAM rather than 64GB:

For users who genuinely only need 32GB, avoiding unnecessary capacity is probably the single easiest way to reduce exposure to the current RAM bubble.

PCIe 5.0 SSDs:

Interestingly, some PCIe 5.0 drives now cost only modestly more than comparable PCIe 4.0 products.

For example, in the September German data:

• 1TB PCIe 4.0 with DRAM: roughly €156
• 1TB PCIe 5.0 with DRAM: roughly €173

That is only around an 11% difference.

At that kind of premium, PCIe 5.0 can make sense if your motherboard supports it.

SATA SSDs:

Generally not attractive right now unless you specifically need SATA.

Some 1TB SATA drives with onboard DRAM cost substantially more than equivalent NVMe products.

BOTTOM LINE

My current recommendation is:

RAM:
Buy selectively now if your build or upgrade is happening within about six months.

DDR5:
Do not count on meaningful short-term price relief.

DDR4:
Deal-shop rather than panic-buy.

NVMe SSD:
Wait for promotions if you can. This is the category I expect to improve first.

SATA SSD:
Only buy for compatibility or a specific use case.

HDD:
Still useful for bulk storage, but compare price per TB carefully because HDD pricing has also risen sharply.

If I were building a PC today, I would prioritize securing the amount of DDR5 I actually need and would be more willing to postpone additional SSD capacity until later.

CONFIDENCE LEVEL

Overall confidence in the next six months:
Medium-high.

Confidence that DDR5 will be cheaper six months from now:
Low.

Confidence that SSD pricing will provide better buying opportunities than DDR5 before September 2027:
Moderately high.

The biggest risk to this forecast is a substantial slowdown in AI infrastructure spending.

The biggest upside risk for prices is that HBM/server demand remains stronger than expected while new DRAM capacity takes longer to ramp.

SOURCES

All accessed September 15, 2026.

MemRadar Memory Price Index:
https://memradar.com/price-index/

3DCenter September 2026 memory-price index:
https://www.3dcenter.org/artikel/speicherkrise-preisindex-september-2026

3DCenter December 2025 memory-price index:
https://www.3dcenter.org/artikel/speicherkrise-preisindex-dezember-2025

Tom's Hardware RAM price tracker:
https://www.tomshardware.com/pc-components/ram/ram-price-index-2026-lowest-price-on-ddr5-and-ddr4-memory-of-all-capacities

Tom's Hardware SSD price tracker:
https://www.tomshardware.com/pc-components/ssds/ssd-price-tracking-2026-lowest-price-on-every-m-2-ssd

TrendForce Q3 2026 memory-price outlook:
https://www.trendforce.com/presscenter/news/20260703-13134.html

TrendForce 2027 DRAM/NAND outlook:
https://www.trendforce.com/presscenter/news/20260730-13158.html

TrendForce client SSD outlook:
https://www.trendforce.com/research/download/RP260805YV

TrendForce NAND supply outlook:
https://www.trendforce.com/presscenter/news/20260721-13148.html

TrendForce HBM wafer-capacity analysis:
https://www.trendforce.com/presscenter/news/20260602-13074.html

TrendForce hyperscaler capex / HBM + RDIMM analysis:
https://www.trendforce.com/presscenter/news/20260825-13198.html

TrendForce September 2026 DRAM update:
https://www.trendforce.com/presscenter/news/20260907-13219.html

Gartner semiconductor/memory outlook:
https://www.gartner.com/en/newsroom/press-releases/2026-04-08-gartner-forecasts-worldwide-semiconductor-revenue-to-exceed-us-dollars-one-point-3-trillion-in-2026

IDC PC/memory shortage outlook:
https://www.idc.com/resource-center/blog/pc-market-enters-volatile-territory-as-memory-shortage-persists-through-2027/

Counterpoint Memory Price Tracker:
https://counterpointresearch.com/en/reports/memory-price-tracker-and-forecast-july-2026

J.P. Morgan memory-shortage analysis:
https://www.jpmorgan.com/insights/global-research/artificial-intelligence/dram-memory-shortage-from-ai

Samsung Q2 2026 results:
https://news.samsung.com/global/samsung-electronics-announces-second-quarter-2026-results

SK Hynix Q2 2026 results:
https://news.skhynix.com/en/q2-2026-business-results/

Micron investor results:
https://investors.micron.com/financials/quarterly-results/

Sandisk fiscal Q4 2026 results:
https://investor.sandisk.com/news-releases/news-release-details/sandisk-reports-fiscal-fourth-quarter-2026-financial-results

Kioxia/Sandisk Kitakami production update:
https://apac.kioxia.com/en-apac/about/news/2026/20260703-2.html

CSIS — "AI Has a Memory Problem":
https://www.csis.org/analysis/ai-has-memory-problem

This version should survive copy/paste into most forums much better because it doesn't rely on Markdown tables.