[Summary]

The AI market in 2026 has reached a turning point.

The coverage and discussion surrounding Anthropic's high-performance AI model, Mythos, has posed important questions to the market.

That is,

"While AI is a tool to increase productivity, it may also increase the ability to attack social infrastructure"

That is the question.

The current AI market price is

*AI model *GPU

  • Semiconductor *AI app *SaaS

Funds have been gathering to find out what things can be conveniently done with AI.

But since Mythos, the market focus has started to shift a bit.

The important thing is

"Finance, medical care, communications, electricity, and government will not be shut down even if attacked by AI"

It is.

In other words, the AI market has evolved from a mere growth theme to

“Defense infrastructure investment to maintain an AI society”

It is possible that the transition has begun.

In this article, we define this change as "defense infrastructure capitalism" and organize a group of companies that are being reevaluated in terms of Japanese stocks.

Impact of Mythos on the market

Mythos is a high-performance AI model announced by Anthropic in April 2026.

According to reports and commentary from specialized organizations, Mythos will be restricted from public release and will be used primarily for defensive cybersecurity testing.

What attracted the attention of the market was not simply the fact that a new AI model was released.

Importantly, Mythos

  • Search for unknown vulnerabilities
  • Exploit creation
  • Chaining multiple vulnerabilities
  • Advanced code analysis *Autonomous construction of attack scenarios

This shows that there is a possibility that they have abilities close to that of.

Here, the market becomes aware of a reality.

Human monitoring speed may not be able to keep up with AI attack speed.

This is the essence of "mythos shock."

AI is a useful tool.

However, it can also be used as a tool to quickly find vulnerabilities in social infrastructure and increase the reproducibility of attacks.

From this moment on, the view of the AI ​​market will change.

Not only companies that create AI,

A company that does not stop the AI society

Funds are starting to arrive.

Human SOC reaches its limits

Traditional SOC (Security Operation Center)

  • Human monitoring *Human judgment
  • Human responds

That was the structure.

However, as AI-based attacks become more sophisticated, this assumption breaks down.

AI is

*24 hours operation

  • Massive log analysis
  • Attack route search
  • Malware modification
  • Generation of defense avoidance patterns
  • Exploring privilege escalation routes

can proceed in parallel.

This is not a world where human analysts can win by working tirelessly.

In cyberspace, just like algorithmic trading in financial markets,

"Machine Speed War"

is beginning.

Even if humans are responsible for the final decision, detection, isolation, and initial response must be handled by AI.

Therefore, from now on,

  • AI anomaly detection
  • Autonomous SOC
  • Automatic isolation
  • Real-time threat analysis
  • Zero trust
  • Self-healing network
  • AI agent audit

is likely to become standardized.

Defender also consumes GPU and power

The important point here is that as AI attacks increase, defenders will have no choice but to use AI as well.

In other words, in the AI market,

“Increasing sophistication of attacks will further increase defense investment”

There is a military expansion structure.

If cyberattacks were human-centered, defenders could respond by increasing their personnel.

However, if the attacker uses AI, the defender also needs to use AI, GPUs, data centers, and electricity.

This is not just a security investment.

This is a constant defense cost to keep society from stopping.

The security market in the AI ​​era is difficult to explain solely in terms of business cycles.

This is because if finance, medical care, communications, electricity, and government stopped, the very functioning of society would be shaken even more than corporate performance.

AI is, after all, a real-world equipment industry

As the AI era progresses, physical infrastructure becomes paradoxically more important.

Ultimately, what drives AI is

  • Power
  • Communication line
  • Server
  • Cooling equipment
  • Optical fiber
  • Submarine cable
  • Data center

That's why.

In the Internet era, software was valued as a lightweight industry.

However, in the age of AI defense, the competitiveness of software is strongly constrained by physical infrastructure.

Even if you have a GPU, it won't work without power.

Even if there is a cloud, it cannot be used if communication is interrupted.

Even with AI security, if a data center goes down, it cannot be defended.

In other words, the winners of the AI society will be

Not just a "software company" but a "company that can secure physical infrastructure"

It spreads to.

Overconcentration becomes vulnerability

In the past cloud strategy, it was efficient to consolidate into a large cloud.

However, as AI cyber risks increase, a single point of failure also becomes a single point of failure.

Therefore, the market will

  • Distributed data center
  • Edge computing *CDN
  • Regional distribution
  • Local processing
  • Multiple cloud operations

may be evaluated more highly.

Until now, companies have pursued "efficiency."

However, in the AI era,

"Never stop"

is the premise of corporate value.

Redundancy is insurance, not waste.

In the case of critical infrastructure, we may enter an era in which society as a whole will pay the insurance premiums.

Japan's biggest weakness is its legacy system

A major weakness in the Japanese market is its old core system.

In particular,

  • COBOL
  • Mainframe
  • Old UNIX
  • Black box business infrastructure
  • Individualized operation

can become a structural risk in the AI era.

Traditionally, the problem with legacy systems is

*High maintenance costs

  • Reduced human resources
  • Renovations are slow
  • Difficult to create new services

It has been talked about in this context.

But in the future, it will become even more serious.

If AI speeds up the analysis of large amounts of code, it could also accelerate the discovery of vulnerabilities in older systems.

In other words, legacy renewal is not just about efficiency.

“Defense investment to rebuild the system to be unstoppable even if it is infiltrated”

Become.

“The cliff of 2025” becomes the “cliff of 2026”

Until now, DX has been

*Cost reduction

  • Dealing with labor shortages
  • Improve operational efficiency
  • Data utilization

It has been talked about as

However, in the AI ​​era, the meaning of DX will change.

The important thing is that

"If it remains old, it will become a target of attack, and if it is infiltrated, social functions will stop"

That is the point.

As a result,

  • Financial DX
  • Medical DX
  • Administrative DX
  • Infrastructure DX
  • Manufacturing industry DX

may begin to be treated as defense spending rather than growth investment.

Until now, the reason why companies have postponed DX has been that it is difficult to see the return on investment.

But from now on,

“Procrastination itself is a risk”

Become.

This is a tailwind for Japan's systems integrators, IT consultants, and cloud providers.

However, it is difficult to evaluate it as a mere extension of business.

The market values ​​companies that can modernize large systems without stopping them and reconfigure security and operations based on AI.

Defense AI also consumes power

AI defense requires enormous computational resources.

Attack detection, log analysis, malware analysis, intrusion route estimation, abnormal communication identification.

To do this in real time, cloud and data center processing power are essential.

In other words, the more widespread AI defense becomes, the more

*GPU demand

  • Server demand
  • Data center demand
  • Cooling demand
  • Electricity demand

increases.

Until now, the shortage of GPUs has been talked about as the biggest constraint in the AI ​​market.

However, in the medium to long term,

"Out of power before GPU"

There is a possibility that such a situation will come.

In fact, with the expansion of AI data center investment, investment themes are expanding to include electric power companies, power transmission and distribution, transformers, air conditioning, liquid cooling, storage batteries, nuclear power, and renewable energy.

AI is no longer a light digital industry.

It is a national-level equipment industry.

1. SIer/IT service

Japan has huge existing systems in government offices, finance, medical care, transportation, electricity, etc.

Therefore,

*NTT data

  • Nomura Research Institute (NRI)
  • Fujitsu
  • Hitachi, Ltd. *SCSK
  • IT service company affiliated with ITOCHU Techno Solutions

is not just a commissioned development company,

“National infrastructure defense implementation company”

may be reevaluated as

What is important is not whether we can create a new system.

The question is whether finance, government, medical care, and transportation can be revamped without stopping.

This ability is rare in the AI ​​era.

2. Cybersecurity

As AI attacks become more sophisticated, cybersecurity companies will most directly benefit.

The areas of focus are:

*EDR *XDR

  • SOC operation
  • Zero trust
  • ID management
  • Vulnerability management
  • Cloud security
  • AI usage audit

It is.

In Japan, there is a possibility that not only specialized security companies but also major system integrators, telecommunications companies, and cloud providers will take up this field.

Security is no longer an optional feature.

This will be the basic fee in the AI ​​era.

3. Sovereign Cloud

In the age of AI, ``where to store data'' itself becomes a security issue.

Especially in Japan,

“Protect domestic data domestically”

This theme tends to be reinforced.

Therefore,

  • Sakura Internet
  • Internet Initiative (IIJ) *NTT Group *KDDI
  • Softbank

Domestic cloud and communications infrastructure such as Japan may be reevaluated from the perspective of national data sovereignty.

However, we need to be careful about expectations here.

Sovereign cloud is an important theme, but the capital investment burden is also heavy.

Investors are

  • Occupancy rate
  • Power procurement *Customer unit price
  • Continuity of government projects
  • Depreciation burden
  • Differentiation from foreign clouds

We need to look at it calmly.

4. Semiconductor manufacturing equipment

The more AI attacks there are, the more defenders will need AI.

As a result, not only AI development but also AI defense will be connected to semiconductor demand.

In Japanese stocks,

  • Tokyo Electron
  • Advantest
  • Lasertech
  • SCREEN Holdings
  • Disco

etc. will continue to be important.

However, there are some caveats here as well.

Semiconductor equipment stocks are strongly influenced by business cycles and valuations.

It is necessary to separate the long-term theme of ``increasing demand due to AI defense'' from the short-term order adjustments, inventory circulation, and stock price levels.

Even if AI is real, stock prices may not always be accurate.

5. Power/data center area

Expansion of AI defense infrastructure will also lead to electricity and capital investment themes.

What attracts attention is

  • Power company
  • Power transmission and distribution equipment
  • Transformer
  • Electric wire *Air conditioning *Liquid cooling *Storage battery
  • Data center real estate

It is.

The AI ​​market is not limited to semiconductors alone.

Rather, from now on,

"Peripheral infrastructure to keep AI running"

There is a possibility that the search will spread.

What investors should be looking at here is not just thematic nature.

The question is whether it actually leads to orders, utilization rates, rate revisions, and capital investment recovery.

In an era where AI writes code, monitors, and defends, humans ultimately matter.

However, it is not simple work that increases value.

The important thing is

  • Advanced security personnel
  • System architect
  • National infrastructure designer
  • AI control human resources
  • Risk Management Officer
  • Human resources who can cross audit, legal affairs, and security

It is.

In the AI ​​era, work will be automated.

But responsibility is not automated.

Who is responsible for system design?

Who decides the scope of AI use?

Who is responsible for explaining accidents?

In this area, the value of humans actually increases.

As Japan continues to have a shortage of highly skilled IT personnel, companies with excellent security personnel and architects may be able to increase their price negotiation power.

While rising labor costs are a cost, they can also be a factor in improving profit margins for companies that can pass them on to customers.

Investors should check the following five points when looking at the market after the "Mythos shock."

1. Is AI defense connected to sales?

Simply talking about AI and security is not enough.

What is important is that in fact

*Security agreement

  • Operational monitoring contract *Cloud usage fee
  • Data center utilization rate
  • System renewal project

It is connected to.

2. Is defense infrastructure a recurring revenue?

AI defense does not end once it is introduced.

Continuous monitoring, updates, log analysis, and vulnerability response are required.

Therefore, rather than a one-off project,

Continuous billing type/operation type/managed service type

companies are more likely to be evaluated.

3. Can you withstand the capital investment burden?

Data centers, cloud computing, power, and semiconductor equipment require huge investments.

Even if the theme is strong, shareholder returns will deteriorate if it takes time to recover the investment.

What you should see is

*ROIC

  • Operating cash flow
  • Depreciation burden
  • Occupancy rate
  • Power cost
  • Is there a long-term contract?

It is.

4. Dependence on national projects

As defense infrastructure progresses, the number of government projects may increase.

While this is a tailwind, it also makes it more susceptible to budget, policy, and bidding systems.

The national security theme is strong.

However, the timing of monetization depends on politics.

5. Is valuation overheated?

This is the most important point.

The theme of AI defense infrastructure is strong.

However, strong themes are always prone to being overbought.

Investors are

"Companies necessary to society" and "Companies worth buying at current stock prices"

need to be considered separately.

AI is real.

Defense infrastructure demand is also real.

Still, if stock prices move too far ahead, returns will be poor.

The essence of Mythos Shock is not just improving the performance of AI models.

The really big thing is

"Defense costs to maintain an AI society will become huge"

This is a structural change.

In the future, AI

  • Power
  • Communication
  • Finance
  • Medical
  • Defense *Administration *Logistics

It becomes deeply embedded.

As a result, the companies evaluated in the market will change.

Until now,

“A company that creates AI”

was the main character.

But from now on,

"A company that protects, never stops, and continues to update the AI society"

may increase in value.

This is not just a thematic investment.

This is a multi-decade capital cycle that will rebuild social infrastructure in the AI ​​era.

What investors need is not to jump at the word AI.

The key is to find out which company's costs to support an AI society will translate into sales and profits.

Reference information

This article is for educational and informational purposes only, based on public information. It is not a recommendation or solicitation to buy or sell any specific security or financial product. Although care is taken with accuracy, the content and future investment outcomes are not guaranteed. Final investment decisions should be made at your own judgment and responsibility.