During periods of accelerated technological transformation, the market tends to oscillate between two extremes: unrestrained euphoria and absolute pessimism. The rise of artificial intelligence seems to have revived both simultaneously.
On one hand, projections emerge suggesting a future of virtually unlimited abundance. On the other, predictions of mass unemployment, the destruction of established business models, and the inevitable bursting of a new tech bubble proliferate.
Historical experience suggests caution when faced with these narratives.
Markets have a long history of overestimating the short-term effects of major innovations and underestimating their long-term impacts. This was the case with electrification, the internet, and mobile devices. All indications suggest it will be no different with artificial intelligence.
The first recurring misconception is the automatic comparison between the current AI race and the dot-com bubble of the late 1990s. While there are evident excesses in certain assets, there is a fundamental difference between the two periods: the demand that sustains current investments is concrete.
The explosion in spending on computing infrastructure does not stem from abstract expectations about a distant future. It responds to an immediate and growing need for processing, storing, and training artificial intelligence models. Revenues from major technology platforms continue to grow at an impressive rate, driven primarily by businesses related to cloud computing and AI.
This is not, therefore, an expansion financed by empty promises or business models lacking apparent economic viability. The relevant debate is not whether there is demand. There is. The central question is who will be able to capture the economic value generated by this transformation.
Up to this point, the biggest beneficiaries have been the suppliers of the critical infrastructure needed to fuel the AI revolution. Semiconductor, component, and equipment manufacturers operate in an environment of scarcity, with strong pricing power and high profitability.
The major buyers of this infrastructure, however, are experiencing a different situation. Companies like Microsoft, Amazon, Google, and Meta are registering robust revenue and profit growth, but at the cost of unprecedented capital investments. The market, at least for now, has rewarded the sellers of "shovels and pickaxes" more than the explorers of the new gold rush.
Perhaps no discussion, however, has garnered as much attention as the supposed mass replacement of workers by artificial intelligence.
The hypothesis stems from a seemingly simple logic: if machines are capable of performing intellectual tasks previously reserved for humans, millions of jobs will disappear. The problem is that this analysis presupposes a static economy, in which there is a fixed amount of work to be distributed.
Economic history suggests exactly the opposite.
When a technology drastically reduces the cost of a productive activity, the effects are rarely limited to the replacement of workers. In general, new products, new markets, new services emerge and, consequently, new demands for capital and labor.
Jevons' Paradox offers a good example. Efficiency gains often increase the total consumption of a given resource instead of reducing it. The same phenomenon can occur with artificial intelligence. By making certain activities cheaper and more accessible, the technology tends to expand its use and create economic opportunities that we cannot even anticipate today.
The case of software development is illustrative. Just a few years ago, programmers were seen as the first victims of automation driven by generative AI. However, the most recent data indicates a recovery in demand for software engineers, even in the face of the rapid evolution of tools capable of writing code.
The most plausible interpretation is that we are observing a process of increased productivity, not large-scale replacement.
A similar argument can be applied to the so-called "SaaSpocalypse," an expression coined to describe a supposed collapse of traditional software companies in the face of the rise of AI.
The narrative became popular because it seems intuitive: if artificial intelligence reduces development costs, why would companies continue paying for enterprise software?
The answer lies in the complexity of the business environment.
Enterprise systems are not just lines of code. They involve integration, support, security, regulatory compliance, training, and business continuity. Completely replacing these platforms imposes high costs and risks for large organizations.
Therefore, it is premature to conclude that AI will eliminate current industry leaders. In many cases, the technology could act as a growth accelerator, allowing established companies to incorporate new capabilities into their offerings and increase their relevance to customers.
Naturally, not all businesses will emerge victorious from this process. Some software categories will face increased competitive pressure. Certain service models may suffer structural erosion. Others, such as cybersecurity and critical infrastructure, tend to benefit from increasing technological complexity.
The problem is that the outcome will hardly be binary.
Artificial intelligence represents an economic paradigm shift comparable to the great technological revolutions of recent decades. Like any transformation of this magnitude, it will produce winners and losers, create new opportunities, and render certain activities obsolete.
But reality is often more complex than both enthusiasts and doomsayers suggest.
The global economy does not appear to be heading towards a scenario of mass unemployment or the widespread extinction of software companies. Nor is there sufficient evidence to claim that we are facing a mechanical repetition of the dot-com bubble.
What we are observing is something more plausible — and perhaps more interesting: a profound redistribution of value within the economy, driven by a technology whose ultimate impact is still far from being fully understood.
As is often the case in major transitions, opportunities will likely be seized not by those who make the most dramatic predictions, but by those who can distinguish noise from structural transformation.
José Medeiros is a partner, head of the San Francisco office, and portfolio manager at São Pedro Capital.
