Big Tech are already dedicating up to 70% of their ebitda to AI. The same figures prior to the outbreak of the points

The most powerful technology companies are immersed in A frantic career for mastering artificial intelligence. Therefore, the investment in infrastructure and technology that allows them to advance in that race is crucial. From its operational benefits a very high percentage is destined in the construction of data centers, buy chips, expand your computational capacity and the like. For some experts and analysts in the sector, this expense level is already reaching Typical levels of past bubbles.

Echoes of the Puntocom. Big Tech are already allocating between 50% and 70% of their EBITDA (Operating Benefit before Tax) to fixed asset investments, mostly from AI and cloud infrastructure, according to the GQG Partners analysis. This resembles AT&T behavior in the Puntocom bubble (a company that before the debacle allocated 72% of its benefits) or Exxon in the energy bubble 2014 (allocating 65%).

Percentage of operational benefit that Big Tech are spending against the level of AT&T and Exxon before bubbles. Image: GQG Partners

Issues. For GQG Partners and as Share Also Tobias Carlisle, founder of Acquirer’s Funds, this high proportion is usually a sign of structural risk, with many expensive investments, assets that can be obsolete, and delay in seeing real returns. In the image we can see the percentage of total operational benefit that Microsoft, Amazon, Alphabet, Meta and Oracle invest in fixed assets (CAPEX)

In detail. There is a huge investor appetite, and the figures reflect it. Alphabet has raised its forecasts and Plan to spend more than 85,000 million dollars This year, almost everything destined to reinforce your cloud and your services. Goal, meanwhile, provides A range of between 66,000 and 72,000 million In 2025, promising to invest a total of 600,000 million dollars By 2028. On the other hand, Microsoft has also announced tens of billions in new facilities to accelerate the training of AI models. Together, Financial Times esteem That the annual capex added of the large technological will exceed 300,000 million dollars, an unpublished figure in the sector.

That level of investment worries analysts. Goldman Sachs He pointed out Recently, Hyperscalers (AWS, Microsoft Azure and Google Cloud) have already committed hundreds of thousands of millions in Capex and R&D, which implies that, to justify it, they must generate much higher income in the coming years. On the other hand, Bank of America warns That the costs of depreciation and amortization of these infrastructure can grow even faster than the income that companies manage to extract from them, which could put their operational margins at risk.

Morningstar, meanwhile, remember That the semiconductor industry, a centerpiece of this race, remains prone to boom and fall cycles, at risk that the current euphoria due to AI derive in a cooling in 2025-2026. Even Sam Altman, CEO of OpenAi, He has admitted That there is a “bubble” around AI, although it clarifies that technology itself will have an immense long -term value.

Between the lines. Investing a lot in infrastructure does not guarantee on its own that future income will compensate for such high expense. If the euphoria between investors is maintained, perhaps yes, but if subsequently penetrates doubt and uncertainty, things can change. He GQG Partners analysis It also talks about “hidden” costs, accelerated depreciation, technological obsolescence, Huge maintenance of data centersenergy, and other factors that could erode much faster than real benefits are believed.

Cover image | Generated by AI with Freepik

In Xataka | “Psychosis by AI” does not exist, it is only another invented diagnostic label: one that has come to stay