Will NAND Prices Keep Rising? Enterprise SSD Demand and the Memory Cycle

NAND prices still have room to rise in the short term, but the pace of increase has already shifted from the rapid surge seen in the first half of the year to a slower high-level uptrend. The core driver of this cycle is not ordinary consumer SSD demand, but enterprise SSDs, AI data centers, and high-capacity storage demand. When you evaluate the NAND cycle, you should not look only at NAND wafer quotes. You also need to track enterprise SSD contract prices, supplier inventory, customer inventory, bit supply, capital expenditure, and gross margins at memory companies. If prices continue to rise, inventories remain healthy, and data center procurement continues, the cycle still has support. If customer inventory builds while price increases stall, risks may appear ahead of earnings.

Key Takeaways

• NAND is still in an upcycle, but price gains have shifted from sharp increases to moderate growth.
• Enterprise SSDs are the strongest demand driver behind this NAND price cycle.
• AI inference, vector databases, and caching are increasing data center SSD consumption.
• Consumer SSDs and smartphone UFS have more limited pricing tolerance.
• Cycle analysis should combine prices, inventory, bit supply, and capital expenditure.
• Memory stocks often peak before product prices do, and low valuation does not mean low risk.

Will NAND Prices Continue to Rise?

NAND prices may continue to rise in the short term, but you should not expect the rapid price increases seen in the first half of the year to continue indefinitely. According to TrendForce’s forecast for NAND Flash contract prices in the third quarter of 2026, NAND Flash contract prices are expected to rise 10%–15% quarter over quarter. This means the upcycle has not ended, but prices have entered a stage of higher base effects, weaker consumer demand, and slower growth.

To judge whether NAND is in an upcycle, you cannot rely only on spot prices for a single specification. A full upcycle usually includes four signals: NAND wafers, enterprise SSDs, client SSDs, and UFS products continue to rise in price; supplier ASPs and gross margins improve together; data center customers increase procurement or sign long-term agreements; and suppliers allocate more capacity to high-value products rather than low-margin consumer storage.

In the first half of 2026, NAND price increases were very strong. TrendForce’s view on NAND Flash contract prices in the second quarter of 2026 previously expected overall NAND Flash contract prices to rise 70%–75% quarter over quarter, mainly because AI and data center demand were strong, suppliers prioritized enterprise SSD capacity, and available supply for consumer products was squeezed. By the third quarter, the expected increase had fallen to 10%–15%, which looks more like a shift from a rapid surge to a high-level uptrend.

Slower price growth does not mean the cycle is over. A move from extreme price increases to moderate increases may simply reflect normal deceleration after a high base. The real warning signs are different: enterprise SSD prices stop rising, supplier and customer inventories both increase, or earnings forecasts for memory companies stop being revised upward. If only consumer products come under pressure first while enterprise SSDs remain tight, the cycle may be entering structural divergence rather than a full reversal.

Summary: NAND prices still have short-term upside, but the key question has shifted from “will prices rise?” to “which products are rising, how fast are they rising, and are inventories healthy?” Enterprise SSDs remain the strongest source of pricing support, while AI data centers and high-capacity storage continue to drive bit demand. Consumer SSDs, smartphone UFS, and memory cards are more constrained by end-device costs and shipment volumes. As long as enterprise SSD contract prices continue to rise, supplier inventory stays low, and data center customers do not meaningfully cut orders, the NAND upcycle is not over. If price growth continues to slow while inventories build, cycle risk will increase.

Why Enterprise SSDs Are the Core Driver of This NAND Cycle

Enterprise SSDs are the most important demand source behind this NAND price increase because AI data centers need not only GPUs, HBM, and networking equipment, but also higher-capacity, higher-performance, and lower-power persistent storage. According to TrendForce enterprise SSD contract price data, enterprise SSD contract prices continued to rise 48%–53% in the second quarter of 2026, clearly reflecting the impact of AI server and general-purpose server procurement on NAND demand.

AI training requires storage for training datasets, model checkpoints, intermediate results, and logs. AI inference requires continuous access to vector databases, retrieval-augmented generation data, user context, and cache data. As long-context models, agent workflows, and real-time retrieval use cases expand, data center SSD demand is no longer just about “cheap capacity.” It is increasingly about a combination of low latency, high concurrency, low power consumption, and high reliability.

Research also reflects this direction. The Tutti paper on SSD-backed KV Cache discusses extending KV Cache to NVMe SSDs in long-context LLM serving. This shows that SSDs are moving from traditional cold storage toward becoming performance components in AI inference pipelines. Whether this type of architecture becomes widely commercialized still depends on system design and software ecosystems, but the direction already shows that NAND storage is becoming more valuable in AI inference.

High-capacity SSDs are a key variable in this cycle. Micron has started shipping the 245TB Micron 6600 ION SSD, a product aimed at AI, cloud, enterprise, and hyperscale data center workloads. Its value proposition is not simply low price per drive, but rack density, power efficiency, and capacity per unit of space. Samsung has also begun mass production of the PM1763 enterprise SSD, designed for AI and HPC servers with PCIe 6.0, 9th-generation V-NAND, and a 4nm controller, emphasizing performance and energy efficiency.

Enterprise SSD demand is already showing up in company financial results. SanDisk’s fiscal second-quarter 2026 results showed that data center revenue grew 64% quarter over quarter, driven by adoption among AI infrastructure builders and large-scale AI deployment customers. By fiscal third-quarter 2026, SanDisk’s data center revenue grew 233% quarter over quarter, while revenue growth was also supported by a higher-value customer mix and higher prices.

Summary: Enterprise SSDs are at the center of this NAND cycle not because SSDs are simply replacing HDDs, but because AI data centers are changing the value of storage. Training data, model checkpoints, vector databases, retrieval-augmented generation, KV Cache, and high-concurrency inference all make high-capacity, high-performance SSDs closer to a core part of compute infrastructure. Enterprise SSD demand strength, contract pricing, and customer long-term agreements are key indicators for whether the NAND cycle can continue. Still, you need to distinguish real deployment from early procurement: if servers, networking, power, and rack construction cannot keep pace, some SSD purchases may first become customer inventory.

Why NAND Supply Cannot Quickly Meet Demand

NAND supply cannot increase immediately just because prices rise. Advanced 3D NAND requires complex manufacturing processes, higher layer counts, yield ramp-up, controller adaptation, and customer qualification. Suppliers also tend to prioritize limited capacity for enterprise SSDs, high-capacity QLC, and data center customers rather than rapidly expanding low-margin consumer NAND. As a result, even when prices rise, short-term supply may not be released quickly.

NAND bit supply mainly comes from five variables: wafer starts, 3D NAND layer count, bits per cell such as TLC or QLC, yield, and capacity utilization. Higher layer counts can improve storage density, but they also bring more complex etching, stacking, defect control, and reliability challenges. New NAND generations must also go through SSD controller adaptation, firmware tuning, thermal design, enterprise customer qualification, and system compatibility verification before reaching scale shipments.

SK hynix has begun mass production of 321-layer QLC NAND Flash, with initial use in PC SSDs before expansion into data center enterprise SSDs and smartphone UFS. QLC can improve capacity density, but enterprise deployment cannot be judged by capacity alone. Endurance, sustained write performance, error correction, controllers, and firmware optimization all matter. In other words, higher layer counts and QLC improve long-term supply efficiency, but they do not create unlimited enterprise SSD supply in the short term.

Kioxia’s 2026 Investor Day materials expected NAND supply and demand to remain tight in 2027, citing strong AI data center demand and constrained NAND supply. The company also planned average annual capital expenditure of about 470 billion yen over the next three years. Capital expenditure has a dual meaning: in the short term, it shows that companies see demand and customer commitments; in the medium to long term, it may create new supply that pressures future prices.

Micron’s fiscal third-quarter 2026 results also showed that the AI era is increasing the strategic value of memory and storage, with the company becoming more active in capital expenditure, products, and customer agreements. Its earnings presentation materials further stated that data center SSD revenue exceeded USD 5 billion and more than doubled sequentially, while DRAM and NAND industry demand significantly exceeded supply and tightness could extend beyond 2027. This supports the case for tight short-term supply, but it also reminds you to watch the timing of future expansion.

Summary: NAND supply is constrained by wafer starts, 3D NAND layer count, TLC/QLC conversion, yield, customer qualification, and capital expenditure. Higher layer counts and QLC technology can increase long-term bit output, but they cannot immediately solve enterprise SSD shortages because high-capacity products still need controller, firmware, thermal, and customer validation. In the short term, suppliers are prioritizing high-value data center products, so supply and demand may remain tight. In the medium to long term, if capital expenditure and new technology ramps are released at the same time, bit supply may again exceed demand and create cycle reversal risk.

Which Core Indicators Should You Track for the NAND Cycle?

To judge whether the NAND cycle can continue, you should track at least six categories of indicators: product prices, inventory, bit shipments, enterprise SSD demand, capital expenditure, and supplier profitability. Looking only at NAND wafer prices can cause you to miss changes in the customer structure of finished SSDs. Looking only at corporate profits can cause you to miss leading signals that the cycle is maturing. A more reliable method is to put prices, inventory, supply, and earnings in one framework.

Price indicators should be layered. NAND wafer prices reflect upstream supply and demand. Enterprise SSD contract prices reflect data center customer procurement strength. Client SSD and UFS prices reflect how much cost pressure PC and smartphone makers can absorb. If enterprise SSDs continue rising while client SSDs and UFS weaken first, the cycle still has structural support. If enterprise SSD prices also stop rising, the risk level increases.

Inventory indicators should distinguish between supplier inventory and customer inventory. Falling supplier inventory usually means stronger supplier pricing power. Rising customer inventory can mean two different things: cloud and server customers may be locking in supply ahead of long-term deployments, or they may be overstocking because they fear further price increases. If customer inventory grows faster than server go-live schedules, GPU arrivals, power connections, and network deployment, later orders may slow.

In earnings reports, you should focus on 10 data points: quarter-over-quarter NAND ASP changes, NAND bit shipments, data center SSD revenue, enterprise SSD customer qualification progress, gross margin, unit costs, inventory turnover, NAND capital expenditure, next-quarter supply-demand guidance, and long-term supply agreement coverage. Revenue growth also needs to be broken down: is it coming from higher prices, bit shipment growth, or a higher mix of premium products? These three sources have different implications for stock valuation.

If you follow related US-listed stocks and semiconductor ETFs, you also need to check trading costs in addition to cycle indicators. US stock trading costs may include not only commissions, but also platform fees, external agency fees, trading activity fees, and FX costs. Under Biya’s US stock trading fee structure, US stock trading commission is USD 0, while platform fees, external agency fees, and other charges are subject to the fee center and order page. Service availability depends on the user’s location, identity verification result, platform rules, and applicable laws and regulations.

Summary: The most reliable framework for judging the NAND cycle is to place prices, inventory, bit supply, enterprise SSD shipments, and company profitability in the same table. Prices tell you current pricing power. Inventory helps determine whether demand has been overdrawn. Bit supply determines whether future oversupply may appear. Data center SSD revenue verifies whether AI demand is entering earnings. Capital expenditure determines the next round of supply pressure. Only when both enterprise and consumer prices weaken, inventory continues to rise, and earnings forecasts are revised downward does the cycle move closer to a full reversal.

When Might the NAND Cycle Peak, and How Should You View Related Memory Companies?

A NAND cycle peak usually does not show up first in current-period profits. It appears earlier in slowing price increases, rising customer inventory, higher capital expenditure, and weaker stock-price reactions to good news. Long-term demand for enterprise SSDs may still grow, but NAND remains a capital-intensive and price-volatile cyclical industry. When comparing memory companies, you need to separate “structural enterprise SSD growth” from “short-term profit leverage from NAND price increases.”

When the cycle approaches maturity, several signals usually appear:

1. NAND contract price growth narrows for two consecutive quarters.
2. Enterprise SSD prices still rise, but client SSD and UFS prices turn down.
3. Cloud inventory grows faster than server deployment.
4. Suppliers restart more idle capacity or accelerate expansion.
5. New process generations drive rapid bit supply growth.
6. Corporate gross margins keep rising, but earnings forecasts stop moving higher.
7. Earnings beat expectations, but share-price reactions become weaker.

Different companies also have different cycle exposures. Samsung Electronics has broad NAND scale and a complete product line, but its memory, smartphone, display, and other businesses reduce its pure NAND sensitivity. SK hynix and Solidigm have a clearer enterprise SSD, QLC, and AI data center strategy, while their HBM and DRAM businesses must also be considered. Micron offers convenient US-market access and transparent disclosures, while being exposed to DRAM, HBM, and data center SSDs at the same time. Kioxia has higher NAND business purity, making it more sensitive to price cycles and capital expenditure. SanDisk has greater exposure to NAND and end storage products, with fast-growing data center revenue, but you still need to watch ASPs, bit shipments, customer long-term agreements, and valuation.

Static P/E ratios also cannot be used alone. When NAND manufacturers are near peak profits, P/E ratios often look low because the market has already begun pricing in future ASP and gross margin declines. A better valuation method is to combine mid-cycle profit, free cash flow, inventory, capital expenditure, and price trends. If a stock has already priced in tight supply and high profits for 2026–2027, it may still pull back if the pace of price increases slows, even if earnings remain strong.

Summary: A NAND cycle peak is usually a gradual process rather than a sudden reversal in one quarter. The earliest signals come from slowing price increases, weakness in consumer products, rising customer inventory, and weaker stock-price reactions to good news. Stronger signals include enterprise SSD prices stopping their rise, supplier ASPs missing expectations, rapid bit supply release, and earnings forecast cuts. When comparing Micron, SanDisk, Kioxia, Samsung, and SK hynix, you should first clarify whether your thesis is based on NAND price increases, AI enterprise SSD demand, or broader memory competitiveness, then evaluate valuation and capital expenditure risk.

If you want to turn the NAND cycle into an actionable investment watchlist, you can put NAND contract prices, enterprise SSD shipments, supplier inventory, capital expenditure, cloud AI investment, and memory stock valuations into one table. Through US stock information search, you can first check basic information for Micron, SanDisk, semiconductor ETFs, or related US-listed stocks. If the relevant service conditions are met, you can also use Biya to follow multi-asset markets including US stocks, Hong Kong stocks, and digital assets. Before trading, you should confirm order types, fee structures, FX impact, and your own risk tolerance. If you need to manage market information on mobile, you can use the Biya App. Public market information, trading rules, and fee structures are for reference only and do not constitute investment advice.

FAQ

Why Are NAND Wafer Prices Different from Enterprise SSD Prices?

NAND wafer prices and enterprise SSD prices do not move in perfect sync. NAND wafer prices mainly reflect upstream supply and demand, while enterprise SSD prices also include controllers, firmware, endurance, thermal design, customer qualification, and long-term agreements. Even when using the same NAND, different SSD products may show different price trends because of performance and customer mix.

Why Does AI Inference Depend More on Enterprise SSDs?

AI inference depends more on enterprise SSDs because inference workloads need continuous access to vector databases, user context, retrieval data, and model cache. The number of inference calls is usually much higher than a single training run. As long-context models and retrieval-augmented generation expand, high-performance SSDs take on more capacity storage and real-time read tasks, though actual demand still depends on application scale and system architecture.

Will Wider Adoption of QLC Enterprise SSDs Push NAND Prices Down?

Wider adoption of QLC enterprise SSDs will not necessarily push overall NAND prices down directly. QLC can reduce cost per unit of capacity and increase capacity per drive, but lower costs may also encourage data centers to deploy more capacity. The price direction still depends on supplier yields, controller capability, bit supply growth, and real data center demand.

Can Enterprise SSDs Fully Replace Data Center HDDs?

Enterprise SSDs are unlikely to fully replace data center HDDs in the short term. SSDs have advantages in latency, performance, power consumption, and rack density, while HDDs still have cost advantages in cold data, backup, and large-scale archival storage. Most data centers use tiered storage based on data access frequency rather than relying on only one storage medium.

How Often Should Retail Investors Update NAND Cycle Data?

Retail investors can review NAND spot prices and channel inventory monthly, while updating contract prices, enterprise SSD shipments, ASPs, gross margins, inventory, and capital expenditure quarterly. When Micron, SanDisk, Samsung, Kioxia, or SK hynix report earnings, or when cloud providers adjust AI capital expenditure, investors should reassess their original view.

What Trading Costs Matter When Investing in NAND Memory Stocks?

When investing in NAND memory stocks, you should look beyond share prices and consider commissions, platform fees, external agency fees, trading activity fees, FX conversion costs, and ADR-related fees. Fee structures vary by market and platform, so you should check the fee schedule, order details, product documents, and local regulatory requirements before trading.