[Summary]
Since spring 2026, there has been a noticeable correction in semiconductor-related stocks that drove the AI market from 2024 to 2025, such as Lasertec, Advantest, Tokyo Electron, and Socionext.
However, this does not mean that the demand for AI will simply disappear.
What is happening in the market is not the end of the AI market, but rather the transition to the second round of the AI market.
The first round focused on GPUs, semiconductor manufacturing equipment, and cutting-edge processes. In the second round, the evaluation axis has begun to expand to include power, cooling, communications, power transmission and distribution, and data center equipment that makes AI work in reality.
In other words, funds are shifting from companies that create AI to companies that can run AI.
Market Note | The Nikkei average also reflects the “AI-selected market”
As of May 15, 2026 14:33 JST, the Nikkei Stock Average (INDEXNIKKEI: NI225) is
| Index | Value | Change from previous day | Change rate |
|---|---|---|---|
| Nikkei Stock Average | 61,356.41 | -1,297.64 | -2.07% |
and has fallen significantly.
What is important is that even though the index as a whole is declining, the stocks that are being sold and the stocks that are likely to be bought are starting to change.
While profit-taking sells in semiconductor manufacturing equipment and high PER growth, the market is focusing on the next investment theme.
- Power equipment
- Data center
- Cooling/liquid cooling
- Optical communication
- Power transmission and distribution infrastructure
I'm starting to see it.
In other words, the decline in the Nikkei average is not just a risk-off;
Signs that the place of funds has begun to change in the AI market
It needs to be read as.
Nikkei Stock Average Index Overview
Why are AI-related stocks sold so much?
AI-related stocks continue to fall.
AI favorites that led the market from 2024 to 2025, such as Lasertec, Advantest, Tokyo Electron, and Socionext, have been making adjustments all at once since spring 2026.
Among investors,
- Is the AI bubble over?
- Has the NVIDIA myth collapsed?
- Will the semiconductor market completely end?
There is widespread anxiety.
However, it would be remiss to view what is currently happening only as the collapse of AI.
The essence is
Price adjustment based on AI expectations rather than disappearance of AI demand
It is.
AI investment itself continues. On the other hand, the stock market is beginning to reevaluate how high its P/E ratio can be tolerated.
It is not the demand for AI that is collapsing.
The first thing to clarify is that the current sharp decline in AI-related stocks is due to
- Suspension of AI use
- Disappearance of data center demand
- GPU surplus
- AI boom ends
This does not mean that
AI investment centered on GAFAM continues to be at a high level. Enterprise AI, AI agents, inference demand, and data center investment are also expanding.
The problem is
Price expectations went too far
There is a particular thing.
In the AI market from 2024 to 2025,
*Anything related to AI will be posted.
- Justice when it comes to NVIDIA
- There is no problem with a high PER as it will grow in the future
This led to the formation of a market with strong expectations.
However, in 2026, the market is gradually starting to see reality.
The biggest factor is interest rates.
Rising interest rates will compress the dreams of AI stocks
The reason why AI-related stocks are particularly weak is that many stocks are evaluated based on ``distant future profits.''
In an era of low interest rates,
- Profit after 10 years
- Future market share
- Unrealized growth expectations
It was easy to give high value to
However, when interest rates are high, the market starts to think like this.
Is it okay to value future profits so highly?
A rise in interest rates reduces the present value of future profits.
In other words, the discount rate for the AI dream is increasing.
This is not a denial of the AI theme, but rather a normalization of valuations.
The days when a PER of 50x was allowed are coming to an end.
When interest rates are low, stock prices tend to rise based on growth potential and story alone.
However, when interest rates are high,
Can you actually make a profit now?
is asked.
The market is currently
- Companies that talk about AI
- Companies that claim to be related to AI
- A company that has only high expectations in the AI theme
I'm starting to look at it carefully.
On the contrary, it is easy to evaluate
- Orders are actually increasing *Profit margin is improving
- There is cash flow
- The benefits of capital investment are reflected in the numbers
It is a company.
The AI market is moving from a market that dreams about to a market that confirms numbers.
The real bottleneck in the AI market is electricity
Currently, the theme of the market is starting to change from GPU performance itself.
The biggest bottleneck now is electricity.
AI data centers consume far more power than traditional servers. Additionally, GPUs generate a lot of heat.
Therefore, GPUs are not the only thing that will be needed in the future.
What really matters is
- Power generation capacity
- Power transmission and distribution
- Transformation equipment
- High efficiency power supply *Liquid cooling system *Air conditioning
- Optical communication
- Data center real estate
It is.
The AI revolution is spreading from the semiconductor revolution to the infrastructure revolution.
From GPU competition to physical infrastructure supremacy
Up until now, the market has been focused on "companies that make GPUs."
But things will be different in the future.
The market gradually
A company that can make AI work in reality
funds are beginning to be transferred to.
AI cannot be achieved by software alone.
Power, cooling, communications, semiconductors, servers, buildings, and even construction capabilities are required to operate large-scale AI.
In other words, in the second round of the AI market,
- Power equipment
- Transformer
- Electric wire
- Cooling
- Optical communication
- Data center
- Industrial infrastructure
is likely to become the new leading role.
The essence of Lasertech's sharp decline
Lasertec's sharp decline cannot be explained solely by concerns about business performance.
The essence is to balance supply and demand.
In the current AI market,
- You can win by buying the best deals
- Increased in conjunction with NVIDIA
- Semiconductors are in perpetual growth
The success experience has been accumulated.
As a result,
- Credit buying balance
- Leverage
- High PER allowed
- Short-term fund concentration
swelled.
However, when the market starts to collapse, a reversal occurs.
When you divide the milestones,
- Selling to avoid margin calls
- Argo selling
- Credit consolidation
- Loss cut chain
is likely to occur.
In this phase, stock prices are driven by supply and demand and position restructuring rather than PER, PBR, and theoretical stock prices.
In other words, Lasertec is currently being sold as an overheated position liquidation stock rather than a stock whose performance is collapsing.
Winner of AI market “2nd round”
The first round from 2024 to 2025 was a market where those with GPUs won.
In the second round after 2026,
Companies that put AI into reality
easily attracts attention.
What the market will see next
- Power that powers AI
- Equipment to cool AI
- Communication that connects AI
- Data center housing AI
It is.
In other words, AI-related infrastructure is starting to become a favorite.
Areas that are likely to attract attention in the Japanese market
Power/Transmission/Distribution
In the AI era, companies that control electricity will become stronger.
What tends to attract attention is
- Transformer
- Power transmission and distribution equipment
- Ultra high voltage cable
- Power receiving and substation related
- Power control
It is.
As long as data center expansion continues, it will be easy to make this a long-term theme.
Cooling/liquid cooling
AI servers generate a lot of heat.
Therefore,
*Liquid cooling *Air conditioning
- Heat control
- Cooling material
- High efficiency fan
Relationships tend to attract attention.
As GPU competition intensifies, cooling demand also increases.
Optical communication/high-speed wiring
In AI servers, communication speed between GPUs is important.
Therefore,
- Optical fiber
- Optical transceiver *High speed connector
- Fast wiring
- Low latency network
is likely to become an important theme in the medium to long term.
Conversely, dangerous AI stocks
The companies that the market will look at harshly in the future are companies that rose solely due to expectations.
Particularly dangerous are the
- Loss-making AI venture *Companies whose actual profits are unclear
- SaaS with only AI concept
- PER dependent growth
- Companies whose customer retention rate is invisible
It is.
AI themes alone cannot survive.
What the market will demand from now on is
*Actual profit *Cash flow
- Power efficiency *Customer retention rate
- backlog *Actual demand
It is.
In other words, only ``companies that can make money with AI'' will be selected, not ``AI-like companies''.
Indicators you should really watch in the future
When looking at AI-related stocks, it is not enough to track just semiconductor stocks.
What really matters is the following.
| Indicators | Meaning of viewing |
|---|---|
| US 10-year bond yield | AI valuation compression pressure |
| Crude Oil Price | Data Center Operating Cost |
| Natural gas prices | Leading indicators for electricity prices |
| Copper price | AI infrastructure construction demand |
| Power stock index | AI money transfer destination |
| Data center REIT | AI actual demand confirmation |
| NVIDIA capital investment | AI investment continuity |
| US utility stocks | Infrastructure shift confirmed |
The AI market is not limited to semiconductors alone.
It is necessary to look at interest rates, electricity, energy, copper, data centers, and power transmission and distribution.
Summary
The current AI market is not collapsing.
In fact, AI is starting to turn from a dream theme to a real and huge industry.
In the first round, you could win by buying a GPU.
In the second round, the companies that can make AI work in reality will win.
In other words, in the future, not only semiconductors,
- Power
- Cooling
- Communication
- Interest rate
- Energy
- Infrastructure
This increases the advantage of investors who can see up to
The AI market price from 2024 to 2025 will rise no matter what you buy.
But things will be different after 2026.
The market is now
Who can really profit from AI?
We are beginning to begin a full-fledged selection process.