Many valuations are extremely high and the AI fantasy is fueling virtually everything that has to do with it. Energy, data centers, chips, ...
Now the question arises as to what is really behind the hype. Is there an end in sight or will the AI boom continue for years to come and does it make perfect sense to invest now?
How sustainable is the boom? What do experts, commentators and opponents say? Are the valuations justified?
On the latter: In addition to the companies that are mentioned here from time to time, which are still not very profitable, this time there are also many companies that are highly profitable and really have the cash compared to previous hype phases.
However, these too must first fulfill their very high expectations over the next few years. Then we will see whether the current valuation is justified and whether share prices can continue to rise.
I see the saturation of the market as a problem for the near future (1-2 years). By this I mean that demand will fall, as fewer chips will be needed and sales will therefore grow more slowly. This would be a problem for the very cyclical semiconductor sector in particular!
I found an interesting source on this from Deloitte (https://www.deloitte.com/us/en/insights/industry/technology/technology-media-telecom-outlooks/semiconductor-industry-outlook.html).
Critical points that are summarized in the article:
- Only 0.2% of chips sold (expensive AI GPUs) account for 50% of global semiconductor sales. If demand falls (even slightly) here. The entire sector will collapse. And the greatest demand is dependent on just a few large companies such as Microsoft, Meta, Alphabet, etc.
- A large part of the growth does not come from more units sold, but from exploding prices. Which has always ended badly in the past. For certain memory chips (HBM), price jumps of 50% are expected by mid-2026. If demand declines slightly, prices will fall and if new chips are produced that are significantly cheaper and of comparable quality, this could result in strong headwinds in terms of sales and margins.
- Data centers: Strong AI development is currently expected. If monetization takes longer or is lower than expected, data centre projects could be cancelled or postponed, which would have a negative impact on chip sales. In addition, the construction of data centers could be canceled for regulatory reasons and sharply higher energy prices for consumers.
However, the order books for 2026 are still full to the brim and there could be a slump in 2027/2028 at the earliest. Ultimately, this depends on how the energy supply is managed, how well companies can use AI and what real added value is generated (keyword ROI).
This brings us to my second source from PWC: https://www.pwc.com/gx/en/issues/technology/ai-performance/want-ai-roi-go-for-growth.html
I draw out the following points for myself here:
- So far, only 20% of the companies they surveyed manage to generate real added value from their AI investments. However, PWC sees a problem in the use of AI.
- Simply saving costs with AI does not make sense in the long term and probably only has short-term effects. The aim should be to achieve growth with AI.
- However, the question is whether all companies can and will implement this. If 80% of companies realize that AI is "nice" but does not improve their balance sheet, they will cut their budgets for AI and chips. Especially as prices continue to rise, the cost/benefit question will become increasingly important.
PWC tends to see a flawed use of AI by companies (focus on cost-cutting measures) and believes that it is possible to generate real growth with AI.
However, other sources also report that it is not so easy to use AI correctly and thereby achieve more turnover or higher margins.
In the following analysis by Rand Corporation (2025)
and by Gartner (2026) (https://mybusinessfuture.com/en/80-ai-failure-rate-2026-how-rand-and-gartner-expose-the-ai/), it is reported that only 19.7% of all AI projects are successful. Compared to "normal" software projects, this is twice as high a failure rate. This means that a lot of money is still being burned through AI investments.
Breakdown of when/why a project was canceled:
- 33.8 percent of projects are canceled before they reach the production phase.
- 28.4 percent make it to the production phase, but do not deliver the expected benefits.
- 18.1 percent run but never cover their costs.
The analysis is the result of 65 AI initiatives over a period of three years.
What is also interesting is that the failure is mainly due to 3 factors:
- Data quality: The models are fed with bad data, which leads to hallucinations or unusable results. (Everyone knows this from Chati & Co.)
- Organization: A pilot project often works in a small department. However, the company-wide rollout then fails due to internal resistance, a lack of responsibility or bureaucratic hurdles
- Use case drift: Projects start with a clear goal (e.g. "Reduce hotline costs"), but after three months get lost in technical gimmicks (e.g. "Which prompt engineering framework do we use?") without solving the actual problem.
My overall conclusion for the current AI development:
YES, AI will have an increasing influence on companies and our lives in the future. However, I am of the opinion that a very rapid AI development is currently being priced in (as in other phases of exaggeration on the stock market), which is unlikely to happen so quickly. So far, we have also seen very few AI successes that increase margins in the long term or cause sales to rise disproportionately.
I expect a crash when it becomes clear that things will not happen so quickly and some companies reduce their investments in AI. Then the semiconductor sector (shovel sellers) will lose first and then the rest.
I do not assume that there will be a shock and collapse like that of hydrogen stocks in 2021 or the dotcom bubble in 2000.
However, I do assume that there will be a crash in the next 1-2 years. Which will then mainly throw out the unprofitable companies and still drag down the companies that earn cash. And a longer bear market for the indices, as these are heavily driven by AI.
How do you see the whole thing? Do you expect the boom to continue over the next few years or will the market become saturated? What do you think would be the consequences?
I hope you enjoyed my first real post. Feel free to leave feedback ;)!
VG and green courses
Tim
PS: Post is self-written. Gemini assisted with source search and summaries.
$NVDA (+1,94%)
$IREN (-3,34%)
$AMD (-2,03%)
$TSM (-0,22%)
$MSFT (-0,99%)
