The Arrival of 'AI CAPEX Fatigue': Companies Whose Massive Investments Are Valued vs. Those That Are Sold Off
The era when the AI boom was a 'tailwind for stock prices' is coming to an end. The discussion on Bloomberg Tech shows that AI investment (CAPEX) does not automatically lead to a higher valuation; rather, the reality is that the moment the gap between the 'scale of investment' and the 'tangibility of monetization' is exposed, the market judges mercilessly. The NASDAQ 100 has crumbled significantly, and while Meta is welcomed despite also increasing AI investment, Microsoft is being sold off heavily. Herein lies the 'tectonic shift' in investor sentiment for 2026.
1. 'AI CAPEX Fatigue'—The 'Double Standard' for Investment the Market Has Started to See
The program describes this shift in atmosphere as a 'vibe shift.' The point is simple: for AI capital expenditure (CAPEX) to be justified, a three-part set of ① growth, ② profit margins, and ③ a future recovery story is required.
A typical example that occurred on the same day was the 'contrast' between Meta and Microsoft. While Meta declared massive investments, it showed strong growth in its advertising business, giving the market a sense that 'even if they invest, they can recover it.' On the other hand, Microsoft's balance between Azure growth and CAPEX was questioned, and the concern that 'investment is running too far ahead' prevailed.
What investors care about is not the amount itself, but the efficiency of the investment (i.e., when and how much it will turn into profit).
2. Microsoft: CAPEX might be 'correct.' But the market cannot wait
The focus of the discussion was the CAPEX expansion shown by Microsoft and the outlook for Azure growth. The program succinctly expresses the anxiety held by investors:
'It looks like the growth in CAPEX is outpacing the growth in Azure'—this 'perception' is what shakes the stock price.
In response, the analyst's explanation was framed as 'short-term pain and long-term fruit.' Microsoft is said to have two strategic priorities.
The first is Copilot (integration into first-party apps)
—the logic being, 'The more Copilot adoption increases, the more investment allocation to support it will also increase.' What is important here is that Copilot is not just a feature addition, but can become a 'multi-year profit margin improvement cycle.' The program also mentions that 'in the long term, Copilot will boost gross margins.'The second is internal R&D (long-term product cycles such as medical diagnostics)
—in other words, investments that cannot be measured by short-term cloud growth 'alone' are increasing.
However, the market dislikes 'uncertainty' on a quarterly basis more than long-term rationality. The impact of the sharp drop mentioned in the program ('the biggest decline since March 2020') is likely close to the consensus of investors who 'understand, but cannot buy right now.'
2-1. The Gaze Toward OpenAI Dependency and 'Circular Transactions'
As a deeper point of discussion, the relationship between Microsoft and OpenAI is addressed. The program pointed to the fact that profits were boosted by valuation gains from OpenAI investments, and the scrutiny toward a circular structure such as 'providing funds → recovering as cloud usage fees.'
In response to this, moves such as 'Anthropic coming to Azure' are discussed as 'risk mitigation' to diversify exposure. While investors want to believe in the winners of AI, they fear that the path to victory will become a single line (i.e., a single specific partner).
3. Meta: The reason massive investment was permitted is because 'the existing business is strong'
Meta saw its stock price rise even while a large number was presented as a CAPEX ceiling. The reason is that the explanation showed not only the 'future of AI investment' but also simultaneously demonstrated that the 'cash machine' of current advertising is strong.
The following perspective is shared within the program:
'Even if growth slows down in the future, there is a monetization lever that can be operated sustainably.'
AI will advance ad display optimization, improving impressions and ROI. Furthermore, in the medium term, there is a possibility that monetization outside of advertising will emerge. For investors, the bridge from 'investment to recovery' is visible.
3-1. The 'next LLM' doesn't have to win—being in the competitive circle is enough
What is interesting is how expectations are placed on Meta's next model (referred to as *'Avocado'* in the program).
'It doesn't need to take the top 3 spots. It is enough if it can show it is competitive.'
In other words, what the market wants is not 'the best in the world,' but model quality at a level that supports the company's own monetization. Meta was valued because it did not overly hype this and was able to talk about it in connection with its strong foundation of advertising.
4. Tesla: CAPEX is a down payment for 'Robotaxi'—though the story tends to be exaggerated
Tesla's CAPEX plan ('$20 billion this year') became a topic of conversation as a level significantly higher than a normal year. The comments from the ARK side clearly place the focus of the investment.
'Robotaxis will dominate for the next 5 years. They could account for over 90% of the enterprise value.'
What investors should look at is whether the investment in factories and AI infrastructure is a 'dream' or 'preparation for mass production.'
On the other hand, there is a calm sense of distance in the program regarding the 'massive chip factory (logic, memory, and package integration)' concept mentioned by Mr. Musk. Given the scale of fixed costs and the difficulties in the industrial structure, it is reasonable to read this as an expression of a sense of crisis regarding supply constraints rather than 'feasibility.'
In short, Tesla's CAPEX is a down payment for 'future dominance,' but at the same time, it is prone to 'exaggeration.' Investors are in a phase where they seek verifiable milestones of progress (number of units, operating regions, safety requirements, unit economics) rather than numbers.
5. ServiceNow and Apple: In the AI Era, 'Existing Strengths' Can Become a Hindrance
The CEO of ServiceNow uses strong language to correct market misconceptions.
'AI will not rewrite our programs. AI makes us better, and we make AI better.'
Furthermore, citing years of workflow track records (massive transaction volumes), he argues for a 'moat that cannot be replaced overnight.' This makes sense, but the stock price is a different matter. The market looks at where growth rates will re-accelerate rather than 'correctness'.
Regarding Apple, more biting expressions appear in the program.
'They are so successful that in the long run, it becomes a pain.'
The larger the sales and profits, the harder it is for an organization to change significantly. The difficulty of moving from a *'legacy platform'* to an AI/agent-centric experience is discussed as investor anxiety. Here, too, the question is not 'Are you investing in AI?' but rather,can you transition to the future without breaking your existing business?is the question.
6. Amazon's AI Training Data Issue: Technical Competition Hits 'Governance'
One of the most important points of discussion in the latter half of the program is the case whereAmazondetected and reported a large amount of content suspected to be child sexual abuse material (CSAM) within its AI training data (*Note: The program also mentions that the material in question was removed before training). The focus here is not on sensationalism, but on the point that 'being unable to trace the provenance of data' becomes a social risk.
In reports, it is said that while the reporting aggregation agency requested 'details such as source and location information' to assist in investigations, the company explained that it was 'external source material without detailed origins.' The more intense the AI development race becomes, the more data companies collect. However, the 'responsibility for the collected data' also becomes heavier at the same time.
This case symbolizes that we have entered an era where the success or failure of AI is not determined solely by performance, but where data procurement, auditing, and accountability (governance) influence corporate value.
Conclusion: The AI Market Moves from 'Investment Amount' to 'Evidence of Recovery'
If I were to summarize the day's discussion in one word, it is that AI is no longer a 'dream growth story,' but has entered a test of capital efficiency. Meta built a bridge with the strength of its existing business, while Microsoft spoke of long-term rationality but was sold off due to short-term uncertainty. While Tesla's scale of narrative could justify its investment, verification points are becoming important. And the Amazon case showed that the AI race will ultimately collide with issues of social systems and ethics.
What investors need is not a perspective of 'betting on AI,' but a perspective to discern which companies can accumulate 'evidence.' AI will not be decided in the next quarter. That is precisely why the quality of a company's explanation and the progress backed by numbers will determine future stock prices.
