Earlier this year, ChatGPT disproved an 80-year-old mathematical conjecture, formulated by Hungarian mathematician Paul Erdős and known as the “planar unit distance problem,” demonstrating how large language models (LLMs) are making their entry into fields considered off-limits for machines.
There are two schools of thought on this topic. On one side, we have the AI “boomers” who believe the technology will be absolutely revolutionary and will benefit a few dominant hyperscalers in the AI sector, and on the other, there are those who think the business model of large language models (LLMs) is based on over-enthusiasm and that issues like ‘hallucinations’ have not been resolved. These “doomers” do not foresee a disaster for the labor market, but for the investments of AI supporters.
The market swings between these two views, depending on which perspective prevails in investors’ minds. Although the “boomers” have now regained ground, at the beginning of this year it was the “doomers” who had the upper hand. These dominant views often ignore the most likely outcome: a middle ground characterized by a more gradual implementation of artificial intelligence as a commercially useful technology, from which many companies operating in various sectors will benefit, rather than seeing all the benefits concentrated in the hands of a few mega-caps.
WHY BULLS MIGHT BE WRONG
We have heard several bullish theses in the past, including that Amazon would replace the entire retail sector, PayPal and fintech would take over banks, or open-source databases would supplant Oracle. None of these predictions came true, and we believe there are lessons to be learned.
History suggests that technology adoption rarely converges on a single platform or a few dominant platforms. Rather than seeing all benefits concentrated in the hands of a few AI giants, we believe there is ample room for large companies to invest in tailored solutions based on proprietary data, private models, and open-source alternatives.
For everyday activities, cutting-edge technology is not necessary. A “good enough” AI model will likely prove adequate for many business needs. Consequently, the sector will be open to a variety of providers who can integrate AI solutions into existing workflows and products, avoiding the risk that everyone ends up dependent on a few dominant companies.
There is also the powerful force called inertia. Historically, large software companies have never absorbed all smaller providers. Even in well-established software sectors, no single provider has come close to complete market concentration, and we would expect AI to follow the same pattern.
One reason is that creating and integrating new systems is a costly and slow process that depends on the trust relationship between client and provider. Taking this into account, smaller software providers will not be put out of business overnight and will have time to adapt to the new world of artificial intelligence, while benefiting from the relationships already established with their customers.
There are other obstacles to the bullish AI thesis. The high energy demand of large-scale data centers, rising memory costs, and the shortage of specialized personnel continue to pose challenges to its growth. This year, Uber demonstrated how easily costs can become a barrier after the company exhausted its entire AI budget in the first quarter of 2026 and is now limiting spending per user.
Concerns about data privacy, liquidity limits of AI providers, and the complexity of implementing these tools at the enterprise level could also slow adoption. Finally, regulation must also be considered. The arguments of the “boomers” tend to be based on the assumption that government regulatory response (at least in the United States) will be limited.
However, if the scenario of strong expansion begins to materialize — with increased unemployment due to worker replacement by AI and a contraction in consumer spending — the risk of social and political consequences would rise. Today, about 70% of the US GDP is made up of consumer spending. If significant labor market disruption is followed by regulatory reforms, these could further limit AI growth and expansion.
Taking the incorrect forecasts about Amazon and Oracle as examples, it is noticeable that established companies have often proven more resilient than expected.
In the past, venture capital funds even established a rule: a product must be ten times better than the existing solution to have a chance of supplanting the market leader. This suggests that the pace of AI solution adoption could be more moderate than boom predictions imply.
WHAT THE MARKET IS TELLING US
One factor behind these conflicting narratives is the current market trend. Because the pace of innovation in artificial intelligence makes the future hard to predict, investors are focusing mainly on short-term cash flows, such as memory chips and short-term earnings, rather than long-term value.
The “subscription” software model (SaaS), which once guaranteed high valuations based on the assumption of unlimited recurring revenues, has undergone a sharp downsizing. The market has shifted toward momentum and short-term cash flows, a dynamic amplified by retail investors and options trading platforms.
When investors have a clearer understanding of the economics of artificial intelligence, the market should return to rewarding that high-quality earnings growth that has historically driven solid and lasting returns.
THE MIDDLE WAY
Artificial intelligence is undeniably a revolutionary technology, and companies that adapt to it should gain significant advantages. What we wonder is whether it will lead to massive market concentration or the destruction of legacy players, as the boom narrative sometimes predicts.
Instead, we foresee significant differences emerging at the corporate level in the adoption and integration of artificial intelligence. In our view, established players who quickly find ways to leverage the new technology to improve efficiency and customer experience will likely have a competitive advantage, while those who fail to do so may fall behind.
This implies a bottom-up investment approach, rather than betting everything on AI hyperscalers or other segments of the AI supply chain. For long-term investors, we believe the more interesting question is not who is developing AI, but which quality growth companies are best positioned to benefit from it.
