What if AI doesn’t generate profits?

Jorge Costa Oliveira

In a recent article, we saw how excessive debt, when not matched by an adequate equity ratio or by income generated within the timeframe of incurred financial obligations, can lead to corporate insolvency even in sectors with a promising future – as happened with the American railroad industry in the second half of the 19th century. A similar situation now appears to be emerging among some U.S. technology companies, particularly OpenAI (and, to a less extent, Oracle).

The technology sector is currently the locomotive of the U.S. financial market. In 2025, the so-called “Magnificent Seven” accounted for roughly 35% of the total value of the S&P 500 and were responsible for about 80% of U.S. stock market gains. It is also the main pillar supporting the U.S. economy itself. Reports from S&P Global show that investment in technology and data centers accounted for around 80% of U.S. economic growth in the first half of 2025; without this sector, American GDP growth would have been approximately 0.1%.

The massive investments committed to the development of artificial intelligence (AI) in the United States – amounting to hundreds of billions of dollars – assume a return on investment over an undefined time horizon, even as company executives promise short-term returns. But is this claim realistic?

The strategy adopted by most technology companies has been to make direct investments across multiple segments of the value chain. While it is understandable that tech firms want to avoid being blamed for sharp increases in electricity and water prices in the locations where they install their data centers (in some cases, the complains of local folks are that electricity costs raised 200% and more), it is highly questionable whether investing directly in power-generation facilities to supply said data centers is the smartest move from a financial standpoint. It certainly is not when it significantly increases a company’s financial exposure and is mostly based on debt.

Meanwhile, a recent MIT report reveals that despite the wave of enthusiasm around generative AI – more than 80% of surveyed organizations have experimented with tools such as OpenAI’s ChatGPT and Microsoft Copilot, and nearly 40% report full deployment – and despite these tools helping individual employees’ productivity, 95% of the organizations surveyed say that AI tools have not achieved a significant organizational impact nor generated a measurable return on investment. Only about 5% of integrated AI pilot projects are “extracting millions in value,” the report found.

Most troubling of all is the choice made by U.S. companies to prioritize the development of artificial general intelligence (AGI) rather than focusing on domain-specific AIs. In the recent past, initiatives presented by tech firms as revolutionary have ended in fiasco – such as Meta’s bet on the metaverse, which consumed $77 billion without delivering returns. How much longer will investors continue to believe that fabulous AGIs are just around the corner, that these will enjoy massive market demand, and that they will translate into substantial sales for the companies that create them?

linkedin.com/in/jorgecostaoliveira