On September 30, a new semiconductor-related ETF was listed on the Tokyo Stock Exchange.
It is the ‘NEXT FUNDS Solactive Global Memory Index Linked Exchange Traded Fund’ (639A) from Nomura Asset Management.
Its feature is that it narrows its investment target to memory-related companies within the semiconductor sector.
There are already many semiconductor ETFs. There are products that invest in Japanese semiconductor-related companies, as well as products that invest in overseas semiconductor companies such as those in the United States.
Despite this, an ETF specifically focused on memory has been launched.
What kind of companies does it hold, and how is it different from general semiconductor ETFs?
Looking at the actual index composition, a slightly different characteristic emerges from the impression one gets from the explanation of ‘diversified investment in up to 20 stocks’.
The role of memory has changed with the spread of AI
When people think of semiconductors, many likely think of GPUs (Graphics Processing Units) handled by NVIDIA.
GPUs are semiconductors that process large amounts of calculations at high speed. They are indispensable for AI learning and inference.
However, AI does not work efficiently with high computing power alone.
A mechanism to supply the data necessary for calculations at high speed is also required.
This is where the presence of semiconductor memory is increasing.
There are several types of memory.
HBM (High Bandwidth Memory) is a high-performance memory for exchanging large amounts of data at high speed. It is used in combination with AI-oriented GPUs and the like.
DRAM is memory for a computer to temporarily store data being processed. HBM is also based on DRAM technology.
NAND flash is a memory semiconductor that can retain data even when the power is turned off. It is used in SSDs and the like, and is related to storage demand in data centers.
As investment in AI expands, attention is gathering not only on semiconductors that handle calculations, but also on memory that supplies and stores data.
However, an increase in memory demand does not necessarily mean that the profits of related companies will increase.
In the past, the memory market has also repeated cycles of demand expansion, production capacity enhancement, oversupply, and price declines.
AI demand is a new growth factor, but it does not mean that the industry-specific business cycle has disappeared.
Even with a maximum of 20 stocks, the top 3 companies account for about 70%.
639A aims to track the ‘Solactive Global Memory Index’.
It selects the top 20 companies by market capitalization from those that meet criteria such as business content, targeting companies related to global memory devices and storage systems.
The index is based on composition ratios according to market capitalization, and in principle, a 25% cap is set for individual stocks.
The main constituents of the Solactive Global Memory Index as of the end of June 2026 are as follows.
Source: Nomura Asset Management. Index composition ratios as of the end of June 2026. This does not indicate the current actual inclusion ratio of the ETF.
The top 3 companies alone account for 67.7%.
It can be seen that the concentration on major companies in the memory market—Micron, SK Hynix, and Samsung Electronics—is extremely high.
By the way, there is one thing that concerns me here.
Although the index has a 25% cap for individual stocks, Micron’s ratio is 28.8%.
This is not a numerical error.
The 25% cap is primarily applied during periodic rebalancing. Subsequent stock price fluctuations may cause the ratio to exceed this limit.
In other words, it is not a mechanism that adjusts daily to ensure each company remains at or below 25% at all times.
Also, Samsung Electronics is not a company that only handles memory. It also has businesses in smartphones, home appliances, and other semiconductor sectors.
This ETF is not a product that only gathers companies specializing in memory, but rather tracks an index that selects companies with significant involvement in memory-related businesses.
Even if you can invest in up to 20 stocks, investment risk is not evenly distributed across 20 companies.
How is it different from regular semiconductor ETFs?
When considering 639A, a comparison with existing semiconductor ETFs is essential.
For example, Nomura Asset Management’s ‘NEXT FUNDS Nikkei Semiconductor Stock Index Linked Exchange Traded Fund’ (200A) is a product that aims to track an index covering 30 Japanese semiconductor-related companies.
It invests not only in semiconductor design and manufacturing but also in related Japanese companies such as those in semiconductor manufacturing equipment and materials.
In contrast, 639A targets global memory-related companies.
The main differences between the two ETFs can be summarized as follows.
639A: Global Memory ETF
-
Investment target: Global memory-related companies
-
Number of index constituents: Up to 20 companies
-
Features: Concentrated in memory and storage-related fields
-
Trust fee: 0.605% per year (tax included)
-
Currency risk: Subject to the influence of foreign stocks and exchange rates
200A: Nikkei Semiconductor Stock ETF
-
Investment target: Japanese semiconductor-related companies
-
Number of index constituents: 30 companies
-
Features: Broadly targets the Japanese semiconductor industry
-
Management fee: 0.165% per year (including tax)
-
Currency risk: Primarily yen-denominated Japanese stocks
Source: Official materials from the management companies of both ETFs. Confirmed in October 2026.
Of course, 200A, which targets Japanese stocks, is not the only general semiconductor ETF. There are also ETFs that target semiconductor companies in the U.S. and around the world.
Even so, the feature of 639A lies not only in the target region, but also in the fact that it narrows the scope of the industry it invests in to memory-related sectors.
With general semiconductor ETFs, you can invest in companies with a wide range of businesses, such as GPUs, CPUs, semiconductor manufacturing equipment, and semiconductors for communications.
With 639A, the performance and stock price trends of memory-related companies are more strongly reflected in the investment results than in investments in the entire semiconductor industry.
Therefore, for investors who believe that memory demand will grow faster than the semiconductor industry as a whole, it becomes a targeted option.
Conversely, if supply and demand in the memory market deteriorate, it is possible that the price movement will be more severe than that of broader semiconductor ETFs.
The important thing is not whether you can invest in the growth theme of memory, but the difference in how much you want to concentrate your investment on that theme.
Comparing it with buying individual stocks
Another point of comparison is individual stocks such as Micron or SK Hynix.
If you are considering investing in the memory market, there is also the method of holding these companies directly.
With individual stocks, you can choose the companies you evaluate and freely decide the investment ratio.
On the other hand, with 639A, you can invest in multiple related companies according to the rules of the index. There is no need to judge the growth potential and competitiveness of each company one by one.
However, considering the feature that the top three companies in the index account for about 70%, it would not be appropriate to hold it with the same mindset as an ETF that diversifies investments across a broad stock market.
Also, not all companies in the memory industry benefit in the same way.
For example, the way profits grow differs between companies that have an advantage in HBM technology development and supply capacity, and companies whose performance is easily swayed by price fluctuations of general-purpose products.
You can use an ETF to leave company selection to the index, but that does not guarantee that it will always be the optimal company selection.
The past strong performance of an index is not the same as the actual performance of an ETF.
The investment management company’s materials also introduce the past performance of the Solactive Global Memory Index.
Caution is required here.
The calculation start date for this index is June 19, 2026.
Index data prior to that date includes a history calculated by applying the index rules retroactively.
Therefore, the movement of the index from 2020 cannot be considered as the return that would have been obtained by actually holding 639A during that period.
Additionally, it is necessary to consider the fact that products targeting this sector are being launched after AI-related stocks and memory stocks have already risen significantly.
A high past growth rate is not a basis for expecting it to continue rising at the same pace in the future.
Even if the memory market expands in the future, if those expectations are already priced into the stock, investment results will not match the market’s growth rate.
Things to check before buying 639A
Here is a summary of the product overview for 639A.
-
Ticker code: 639A
-
Listing date: September 30, 2026
-
Target index: Solactive Global Memory Index
-
Trust fee: 0.605% per year (tax included)
-
Trading unit: 10 shares
-
NISA: Eligible for the growth investment quota
-
Dividend record date: September 10 every year
-
Currency hedge: None in principle
The closing price on October 9, 2026, was 198.7 yen, and the transaction amount per 10 shares was 1,987 yen. The total net assets as of that date were 1.46 billion yen.
While you can invest with a relatively small amount, a small investment amount is not the same as low investment risk.
Also, for ETFs that have just been listed, you should pay attention to trading volume and bid-ask spreads.
Since ETFs are traded on an exchange, the trading price does not always match the net asset value. When placing an order, you need to check the quotes and trading conditions.
Regarding trust fees, other expenses may also be incurred during the holding period.
How to think about concentrated investment in memory
With the arrival of 639A, Japanese investors can now hold global memory-related companies in a single ETF.
However, investing in memory-related companies themselves has been possible until now through semiconductor ETFs and individual stocks.
What is new is that it has become easier to choose investment in memory-related companies as an independent product from within the broad semiconductor industry.
Looking at its contents, while it invests in up to 20 stocks, the composition ratio of the top three companies accounted for about 70% based on index data as of the end of June 2026.
It is easier to understand its characteristics if you view it as a product that focuses on major companies in the memory industry rather than a product that invests broadly in growth sectors.
The fact that the spread of AI is bringing new growth opportunities to the memory market is not the same as memory-related stocks generating high returns in the future.
With that difference in mind, I think it is a new product worth comparing with existing semiconductor ETFs and individual stocks.
Main reference materials
*Composition ratios use index data as of the end of June 2026, and prices and total net assets use data as of October 9, 2026. This does not recommend the purchase or sale of any specific product.