The debate about artificial intelligence has gained scale, but still stumbles on misreading of the market.
Many treat the issue as if there were a single bubble forming. It’s not the case. There are three simultaneous bubbles, interdependent, but of different nature. Understanding this difference is the first step to transforming hype into opportunity.
Three Bubbles in AI: Pricing, Capacity, and Discourse
The first is the price bubble. Companies linked to AI are evaluated as if a relevant part of the future is already guaranteed. Multipliers rise quickly and, when the correction comes, they punish those who bought expectations as if it were delivery.
Continues after advertising
This movement is financial, not technological. Prices are volatile and tend to converge with fundamentals. Adoption is following a different pace. For the investor, the risk is confusing narrative with cash generation.
The second is the capacity bubble. Investment in data centers, power, cooling and GPUs has grown far beyond current demand. Part of this infrastructure will be idle in the short term. But history shows that today’s excess often prepares the ground for tomorrow. Costs fall, elasticity increases and new applications become viable.
This is the pattern described by Carlota Pérez in other technological outbreaks. For the investor, the question is to identify who can get through the period of idleness without compromising the balance sheet and who will be positioned to capture the drop in costs when the cycle turns.
Continues after advertising
The third is the speech bubble. Pilots are launched without a defined problem, without a baseline and without integration. Hype replaces method. The result is frustration. An MIT study showed that 95% of corporate AI projects had no impact on P&L.
The problem, most of the time, is not the technology, but the design and governance. The attentive investor knows that vanity indicators do not translate into ROI. Value appears when there is clear pain, objective metrics and real integration into the process.
These three bubbles coexist, but do not cancel each other out. The drop in multiples does not erase already visible gains in automation, reduced rework and better decision-making. Idle capacity does not eliminate savings in back officeservice and analysis. And rhetorical excess does not prevent disciplined companies from transforming well-structured pilots into operation. For the investor, the point is to distinguish noise from value.
Continues after advertising
How to turn hype into opportunity
The central issue is capital allocation. When the market supply is not good enough, transaction costs increase and companies need to develop their own skills, with a direct impact on CAPEX.
When the supply reaches an adequate standard, supplier economies of scale and cash preservation prevail. This choice is dynamic and depends on the price cycle, capacity elasticity and learning curve. The mistake, from the investor’s point of view, is to support companies that try to anticipate steps without a mature operational base, increasing the risk of lock-in and volatility in results.
Another important point is to get out of the restricted view of cost. AI should not be seen just as a tool to reduce expenses. By making tasks cheaper and faster, it allows you to increase frequency, scale and quality. This opens up space for new markets and revenues that were previously unviable.
Continues after advertising
Investors who only look at cost synergies may miss most of the growth story. The modularity of the tools, for example, facilitates combinations between foundational models, compact models and orchestration layers. This arrangement expands the set of viable projects and improves the expected return.
From narrative to cash: where the value lies
The balance is positive. The utility is already visible in high-volume, low-differentiation activities, where AI reduces errors and frees up time for higher-value tasks. The gradual drop in costs expands the set of economically sustainable applications.
Companies that learn to govern data, mitigate risks and measure results make their projects more predictable and closer to business priorities. This tends to be reflected in more resilient margins, lower volatility and more sustainable valuation.
Continues after advertising
For the investor, the reading is clear. There are three bubbles in AI: pricing, capacity, and discourse. They can occur together, but they do not mean the same thing and do not nullify the potential of the technology. Separating the categories helps to understand where the market is just pricing expectations, where there is excess capital that can become a future advantage and where the discourse does not deliver results.
The financial cycle corrects exaggerations, the infrastructure remains and the utility curve continues to advance. The moment calls for sobriety, not euphoria. But it is exactly sobriety that opens up space for consistent investment opportunities.
