NANDフラッシュ市場 2026 AIデータセンター向けeSSDが利益の中心へ Samsung 首位・全方位供給 SK + Solidigm QLC eSSDが焦点 Kioxia / SanDisk 日米JVとBiCS Micron G9 NANDで技術先行 見るべきは数量ではなく、容量・ASP・投資規律

First, the conclusion

The leading role in the NAND flash market is shifting from the number of smartphones and PCs to the capacity of AI data centers.

This is the core of our investment theme. It's important to talk about 512GB or 1TB of storage on a single smartphone, but around AI servers, the SSD capacity on a single device will go from tens of TB to over 100TB. Even if the growth in volume is slow, if the installed capacity jumps, demand for NAND bits will increase. Furthermore, eSSDs for data centers have higher unit prices and higher quality requirements than those for consumer use.

TrendForce reports that in the first quarter of 2026, the total sales of the world's top five NAND companies increased by 83.7% from the previous quarter to over $38.9 billion. Samsung has 31.6%, SK hynix Group has 17.6%, and Kioxia, Micron, and SanDisk each have 13.9%. These five companies account for approximately 90.9%.

The market is already buying a lot of this structure. That is why from now on, simply saying "there is strong demand for AI" is not enough. How much will ASP rise? How much will each company hold back on increasing production? Will they be able to achieve yields in a generation with more than 300 layers? When it comes to NAND stocks, the more positive factors there are, the more difficult it is to determine peak profits.

what is happening

In the NAND market in 2026, AI server investment and tight storage will be at the same time.

GPUs and HBM seem to be the focus of AI investment, but in actual data centers, the load of storing and reading increases, including training data, inference logs, RAG data, KV cache, model updates, and backups. Coupled with the HDD shortage, some storage demand is flowing to QLC enterprise SSDs.

As a result, the profit structure of NAND manufacturers has changed considerably. Not only will bit shipments increase, but the mix of high-capacity eSSDs will also rise. If the mix improves, ASP will rise and utilization rates will improve, making it easier for memory manufacturers with fixed costs to see their profits grow faster than their sales.

However, this also includes risks typical of memory stocks. If companies see high ASP and increase capital investment, supply will increase with a delay of several quarters to several years. The more there is a current shortage in the market, the easier it is for the seeds of the next oversupply to grow.

structural change

The NAND industry is an oligopoly. Based on sales in the first quarter of 2026, the top five companies - Samsung, SK hynix Group, Kioxia, Micron, and SanDisk - accounted for around 90% of sales. There are high barriers to manufacturing equipment, yield rates, customer certification, and capital investment, so it is not a market that will collapse all at once with the entry of new players.

On the other hand, the positions among the top five companies are different.

Company/GroupPositioning for 1Q 2026What investors look at
Samsung ElectronicsNo. 1 with a sales share of 31.6%DRAM/HBM investment allocation, eSSD recovery, production increase discipline
SK hynix Group / Solidigm2nd place. Solidigm's high capacity QLC eSSD contributesContinued orders for large capacity SSDs for AI servers, capital allocation for NAND and HBM
Kioxia3rd place. 1Q26 sales increased 80% compared to the previous quarterRatio for AI infrastructure, BiCS generation transition, market sensitivity due to NAND specialty
Micron Technology4th place with SanDiskG9 NAND's technological superiority expands its adoption in data center SSDs
SanDiskAfter becoming independent from WD, continued JV with KioxiaCapital efficiency as an independent company, co-investment with Kioxia, product mix

China YMTC is here. YMTC is a NAND manufacturer with strong national policies in China, and cannot be ignored in the long term. However, U.S. export controls restrict access to advanced semiconductor manufacturing equipment and software, and uncertainty remains about the speed of mass production of advanced generations. From the perspective of top Western manufacturers, YMTC's restrictions can also be a factor in alleviating oversupply.

Beneficiary area

The most immediate beneficiary area is high-capacity eSSDs for data centers.

Solidigm is based on the former Intel NAND division, and has a large presence in the SK hynix group for high-capacity QLC eSSDs. TrendForce also believes that Solidigm's orders for high-capacity QLC enterprise SSDs boosted sales for the SK hynix group in the first quarter of 2026.

KIOXIA and SanDisk have great significance in joint production between Japan and the United States, centered around Yokkaichi and Kitakami. In July 2026, both companies announced the start of production of 10th generation 3D Flash at Kitakami Factory Fab2. Kioxia alone has begun shipping samples of its 10th generation BiCS FLASH, showing technical indicators such as 332 layers, 59% higher bit density, and 4.8Gb/s interface speed.

Micron's strength is that it was able to mass-produce G9 NAND quickly. The company announced the mass production shipment of 9th generation TLC NAND in July 2024, and is promoting its performance for AI and data-intensive applications. As of 2026, the focus will be on how to convert the G9 to data center SSDs.

It's worth watching for peripheral players as well. Memory module/SSD brands like Kingston, SSD controller companies like Phison, and specialty NAND/NOR players like Macronix and Winbond have revenue opportunities outside of the cutting-edge 3D NAND oligopoly. However, they will not directly benefit from the price increases for AI eSSDs; they will be affected by procurement prices, inventory evaluation, and customer mix.

headwind area

The first headwind is smartphones and PCs.

TrendForce believes that while server orders will be strong in the second quarter of 2026, rising memory costs and device prices will dampen demand for smartphones and PCs. In other words, not all NANDs are equally strong. There is a polarization situation where there is a shortage of products for data centers, and there is a limit to the ability to pass on prices for consumer products.

Another headwind is the expected value of stock prices. Memory stocks are bought all at once when they start making profits. But it's also memory stocks that look the cheapest when profits are at their highest. Even if a company looks cheap when looking only at the P/E ratio, the market may start pricing in the next ASP decline.

The same is true when looking at Kioxia (285A). For the full fiscal year ending March 2026, sales revenue recovered to 2,337.6 billion yen and operating profit to 870.4 billion yen, a result of the simultaneous improvement in NAND market conditions and demand for data centers and enterprises including AI applications. Even if the financial results are good, the reaction will be slow if the stock price moves ahead of the market. From now on, we will be looking at the sustainability of profit margins rather than profit levels.

KPIs that investors should look at

When looking at the NAND market, shipment volume alone is not enough.

KPIwhat to seeWhat it means in the stock market
NAND ASPPrice per GB, contract price, spot priceThe most direct driver of profit margins
Bit GrowthGrowth in total supply capacityIf demand exceeds the next price drop factor
eSSD sales ratioData center/enterprise ratioConfirmation of high value-added mix
Adoption of QLC high capacity SSDOrders for 60TB, 120TB and aboveSee the quality of AI storage demand
Utilization rate and inventoryFactory operation, customer inventory, distribution inventoryEarly signs of ASP peak out
Capital investment amountAddition of new capacity and investment in generation transitionOversupply risk and technological competitiveness
Yield of over 300 layer generationsBiCS10, G9/G10, next generation V-NAND, etc.Cost competitiveness and adoption speed

Personally, what I want to see most in the second half of 2026 is "supply discipline." The market already knows that demand is strong. The question is to what extent Samsung, SK hynix, Kioxia, SanDisk, and Micron can suppress the addition of new capacity and protect ASP while lowering costs through generation migration.

risk scenario

In a bullish scenario, eSSD demand will continue for a long time due to the increase in AI inference, while HDD shortages and data center power saving needs will also provide tailwinds for NAND. If each company suppresses the addition of new capacity and moves from over 200 layers to over 300 layers with good yield, ASP and profit margins will likely remain high.

In a bearish scenario, demand for PCs and smartphones will further slow down due to rising memory prices, and demand for servers alone will not be able to absorb the entire demand. This, combined with increased production by various companies, buildup of customer inventories, and a temporary suspension of investment in AI data centers, will cause NAND prices to reverse sooner than expected.

Geopolitics cannot be ignored either. U.S. export restrictions to China will restrain Chinese companies from becoming more advanced, but will also increase the cost of disrupting supply chains. If regulations become stronger, it will be a headwind for YMTC, but it will also create uncertainty in the procurement of equipment, materials, and customers. When demand for semiconductor memory is strong, political risk is more likely to be a factor in stock prices.

Related pages

summary

In 2026, the NAND market will see a change in the way it generates profits due to eSSDs for AI data centers. Due to the combination of oligopoly by top manufacturers, HDD shortages, penetration of QLC eSSDs, and transition to the 300+ generation, the current market conditions are quite strong.

However, NAND is cyclical until the end. As soon as we see strong demand, rising ASP, and high profit margins, stock prices tend to move ahead. What investors should be looking at is not the demand for AI itself, but whether companies can turn that demand into high-value products, long-term contracts, yields, and investment discipline.

Companies with a strong NAND focus, such as Kioxia and SanDisk, are highly sensitive to profits if the theme is right. On the other hand, if market conditions reverse, the pain will be quicker. The NAND investment theme for the second half of 2026 is likely to be one in which ASP, Bit Growth, eSSD ratio, and competitive CAPEX are checked every quarter, rather than being completely bullish.

source