In the world of artificial intelligence, the financial operations of large AI laboratories are becoming increasingly clear. Current discussions surrounding the finances of these companies highlight the crucial role of computational power economics. It is increasingly evident how a significant portion of these companies' rapidly growing revenues is absorbed by costs associated with the constantly rising demands for computing resources.
For example, one of the leading AI companies recently published data indicating that margins related to computational power are actually deteriorating. Although the company experienced robust revenue growth in the billions of dollars, its operating loss simultaneously deepened, primarily due to escalating infrastructure costs. In the recent quarter, computational power expenses consumed 58% of revenues, an increase from 48% in the previous period. At a time when investors might expect improvement, the trend is moving in the opposite direction.

Got an idea for a website or app?
Let's talk it through — non-binding, with no pressure, and a concrete next step.
The reason for this development is primarily enormous demand, which forces AI laboratories to secure computational power at higher prices. Instead of long-term contracts with large cloud service providers, they must turn to smaller, more flexible suppliers who offer immediate solutions but at significantly higher rates. This strategy, driven by rapid growth and the urgent need for scaling, leads to a substantial increase in operating costs. These expenditures are projected to be the largest item in companies' operating costs going forward.
Although some AI companies are actively working on building their own computational infrastructure in an effort to reduce dependence on external suppliers, the pace of demand growth is so high that these internal projects cannot yet keep up. Therefore, collaboration with external, and often more expensive, suppliers remains a necessary reality. This underscores how closely the success of AI laboratories is tied to the availability and cost of computing resources.
From a certain perspective, these are so-called "problems of prosperity" – as long as revenues are growing by tens of percentages per quarter, most investors will not be overly concerned by the high costs of computational power. However, it is a stark reminder of AI laboratories' deep dependence on the massive infrastructural construction of data centers and their vulnerability to any complications in this process. A potential slowdown in data center construction, for example due to economic pressures or local opposition, could significantly complicate the already challenging situation surrounding computational power. The future of leading AI laboratories thus remains largely influenced by the prices and availability of computational power.
