Not just semiconductor stocks | How far has profit-taking selling in the AI ​​market spread?

First, the conclusion

Today's decline is a bit rough to describe as simply a ``lower semiconductor price''.

Looking at the lineup of products sold, it has expanded to include memory, silicon wafers, front-end processing equipment, EUV inspection, back-end processing equipment, electronic components, and science and semiconductor-related equipment. In other words, there was a movement to lock in profits in a group of stocks where funds were concentrated around AI investment.

This is more like an adjustment to expectations than an end to the theme. AI server investment, NVIDIA related, HBM, NAND, advanced packaging, semiconductor materials. All themes remain strong in the medium to long term. However, the earlier a stock moves in price, the more likely it is to be sold due to a slight rise in interest rates, a correction in U.S. semiconductor stocks, changes in exchange rates, or waiting for order confirmation.

What we should be looking at in today's market is not ``Are semiconductor stocks over?''

What we want to see is which stocks were sold solely based on expectations, which stocks are actually facing increased market risk, and which stocks have further evidence to confirm. These three.

What's happening?

On the market screen on July 7th, the following declines were noticeable.

In this paper, we treat this value as the in-situ observation value for July 7th. It is not a closing price or official statistics, but a material for reading market sentiment on the day.

The key point is that the stocks sold are not confined to a single industry. Kioxia has memory, SUMCO has wafers, Tokyo Electron and Lasertec have manufacturing equipment, DISCO has back-end processes, MARUWA has electronic components, and JEOL has science and measurement equipment and semiconductor-related equipment.

In other words, investors did not just sell ``semiconductor manufacturers.'' The entire value chain that was expected to benefit from AI infrastructure investment was once lightened.

Semiconductor manufacturers were not the only ones sold today.

When looking at the semiconductor market, if you only look at KIOXIA and SUMCO, it seems like you are talking about market stocks. Looking at Tokyo Electron and Lasertec, it seems like they are talking about equipment stocks. If you include DISCO, you'll be talking about AI/HBM and advanced packaging. If MARUWA and JEOL are included, the scope will expand to include electronic components and measurement/inspection areas.

How it spreads is important.

If the semiconductor industry is really only deteriorating, it would be easy to rely on one negative factor, such as memory prices, capital investment, or wafer demand. When there is widespread selling like this time, there is a high possibility that the position adjustment of the overall theme is taking place before individual fundamentals.

For AI-related stocks, stock prices tend to react first, even before business performance improves. Therefore, in a rising phase, ``expectations that have not yet turned into profits'' are also bought. Conversely, during an adjustment phase, performance that has not yet deteriorated is the first to be questioned.

Today's decline appears to be exactly that type.

Why was it sold?

The first is to sell to take profits in high P/E stocks.

It is quite easy to factor in the assumption that AI investment will continue for semiconductor equipment companies such as Lasertec, Disco, and Tokyo Electron. The stronger the expectations for growth, the more likely the company will be sold at the moment it appears that the next round of growth will slow, even if orders and financial results are not bad.

The second is the link between US semiconductor stocks and the SOX index.

Japanese semiconductor stocks are not driven solely by domestic materials. The risk tolerance of overseas investors changes depending on the combination of NVIDIA, memory stocks, SOX index, US interest rates, and dollar/yen. If profits are taken in AI-related stocks in the U.S., selling will likely spread to Japan's semiconductor stocks as well.

The third reason is doubts about the memory market.

While Kioxia will benefit from AI storage demand, it will be strongly affected by NAND prices. Performance for the fiscal year ending March 2026 has improved significantly, but when the numbers look good, it is suspected that memory stocks will peak out. Even if the P/E looks low, it may be a sign that the market doesn't yet have confidence in the sustainability of earnings.

The fourth factor is the impact on materials and components stocks.

MARUWA and SUMCO are in a different position than equipment stocks. However, stocks that have been bought based on expectations for demand for AI servers, semiconductors, and electronic components are sold together when theme funds are withdrawn. Rather than a decline in performance, the effect was that they were all placed in the same basket in the portfolio.

AI expectations have not changed

Don't get me wrong here. This does not mean that demand for AI has disappeared.

Data center investment, GPU, HBM, advanced packaging, NAND storage, semiconductor manufacturing equipment. The direction of this demand is still uncertain. However, stock prices are not driven by demand itself, but by the speed at which demand turns into profits and how much expectations are factored into that.

What the market sold today was not so much the AI ​​theme itself, but the positions that were driven higher by the AI ​​theme.

The numbers are good. The theme is also strong. The problem is the inclusion.

If we don't differentiate this, we will see all declines as bearish news. On the other hand, it is dangerous to assume that it is a push because it has been lowered. Is it an adjustment in expectations, or have orders and prices actually started to deteriorate? This determination will be the next focus.

You can see the differences by industry

If we lump them together with "semiconductor stocks," the meaning of today's decline becomes vague.

Even though they are related to AI, the way they generate profits is completely different.

Tokyo Electron will wait to see if customers continue to make capital investments. Lasertec sees a recovery in orders for EUV-related equipment. DISCO will see whether demand for AI/HBM post-processing shows up in monthly sales. KIOXIA View NAND Price and Inventory. SUMCO sees a recovery in wafer supply and demand.

Today's decline gives us an opportunity to look at this separately.

What's happening in the market?

Rather than rejecting the AI ​​theme, investors seem to be cutting back on once-winning positions.

In a rising market, stocks that are in the middle of a theme are bought. Semiconductor devices, memory, materials, post-processing, electronic components. When capital is strong, it spreads throughout the value chain. However, during the adjustment phase, reverse rotation occurs. First, the items that were going up will be sold first.

Now comes the difficult part.

If US semiconductor stocks recover, the SOX index does not collapse, and major AI stocks such as NVIDIA are bought again, today's decline may remain within the profit-taking range. On the other hand, if there is weakness in memory prices, capital investment plans, and order forecasts, this will not just be a sign of weakness, but will lead to a reassessment of profit expectations.

The market has not yet drawn a conclusion. That's why the next financial results and order data are important.

KPIs that investors should look at

The key points that need to be looked at in the future are that the index alone is not enough.

In the US, see comments on SOX index, NVIDIA, memory stocks, and capital investment related to AI servers. Domestically, we will check orders and sales forecasts for Tokyo Electron, Lasertec, and DISCO, NAND prices and eSSD demand for Kioxia, wafer market conditions for SUMCO, and electronic component demand for MARUWA.

Exchange rates cannot be ignored either. A weak yen tends to be a tailwind for companies with a high export/overseas sales ratio, but stock prices are not supported by a weak yen alone. Right now, the priority is to confirm whether AI demand is actually turning into profits rather than the exchange rate.

Interest rates also work. When domestic and foreign interest rates rise, growth stocks that have higher expectations for future growth tend to have their valuation multiples lower. The reason why stocks with high P/E ratios are easily sold is not just because of poor business performance.

Risk scenario

On the bulls' view, today's decline is a healthy position adjustment after the rally. As long as AI capital investment continues and orders for semiconductor manufacturing equipment, monthly sales of back-end equipment, NAND prices, and demand for electronic components do not collapse, the theme will remain.

On the bearish side, today's decline marks the beginning of a reassessment of earnings expectations. In particular, memory prices will slow down, equipment orders will not grow, US semiconductor stocks will continue to fall, and valuation multiples of high PER stocks will not recover. In this combination, it is difficult to see the decline as simply taking profits.

Personally, I think it is too early to call the end of the AI ​​market at this stage. However, it is possible that the phase in which "anything is bought with AI" may be over for now. Now it's time to choose.

Summary

It would be more natural to view today's decline as a result of profit-taking concentrated in high-growth stocks related to AI and semiconductors, rather than a deterioration in the semiconductor industry as a whole.

Kioxia's NAND price, SUMCO's wafer market conditions, Tokyo Electron's capital investment, Lasertec's EUV orders, DISCO's back-end process demand, MARUWA's electronic components, and JEOL's measurement and semiconductor-related equipment. Each has different KPIs to look at.

There is still demand for AI. However, stock prices were already looking far ahead. Today's market was a day that dampened those expectations once and for all.

Related pages

• [Why Lasertec and semiconductor stocks are weak | 3 points to consider in the July market] (/news/market/2026/07/07/lasertec-semiconductor-selloff.html)
• [Why Kioxia stock is rough | Looking at AI storage expectations and NAND market risk separately] (/news/market/2026/07/07/kioxia-nand-selloff.html)
• [Reading the polarization of the semiconductor sector | The reason why the price movements of 8035, 6920, and 285A diverged] (/strategy/2026/07/07/semiconductor-sector-divergence-8035-6920-285a.html)
• Why only semiconductor stocks were sold and Nintendo was bought | Reasons for the diverging price movements of large stocks
• Ranking of 20 Japanese semiconductor stocks: Compare AI, HBM, equipment, and materials

source

• Tokyo Electron "Investor Information"
• Lasertec “Shareholder/Investor Information”
• KIOXIA Holdings “Shareholder/Investor Information”
• Disco “Shareholder/Investor Information”
• MARUWA "IR Information"
• JEOL “IR Information”
• SUMCO “Investor Relations”