According to a recent MIT report, 95% of organizations are seeing no (or very limited) returns from their internal generative AI pilot programs, despite large investments in their implementation. The study has its limitations, especially given the limited sample size of professionals and executives surveyed. But it offers a stark counterargument to the optimistic narratives promoted by OpenAI, Anthropic, and other prominent genAI companies.
Skepticism around their products’ ability to help companies increase revenue and profit, even in the medium to long term, is becoming more common, and not just among AI optimists anymore. This shift has intensified after GPT-5, OpenAI’s latest LLM, failed to live up to the heightened expectations set by Sam Altman himself.
While the use case and profitability of genAI applications is still very much to be proven, the IT industry’s bet on genAI and the companies developing it is already massive. From 2013 to 2020, cloud infrastructure capital expenditure grew from $32 billion to $119 billion, driven mostly by the rise of social media platforms and video content.
Post-Covid, the curve goes wild: in 2024, spending reached $285 billion, and in 2025, the top 11 cloud providers are forecasted to invest almost $400 billion. That’s more than they’ve committed in the past two years combined and the figure mostly stems from the massive compute needs for training and inferring LLM models.
READ ALSO
A Complex Answer
The fundamental disconnect between the money companies are spending to compete in the AI race and the potential return on investment is widening as fast as their capex. A growing cadre of experts is asking the same fundamental question: with this uncertainty around the effectiveness of its real-world business application, will gen AI as a market ever be able to make money?
In a moment of profound change, the answer is complex and open to interpretation. On one side, we need to take the “inevitabilism” of Altman, Amodei, and other AI maximalists with a sizable grain of salt. It’s simply not true that a future where their particular flavor of genAI dominates the workplace and integrates into our lives at all levels is “inevitable”, despite what they’d like us to take at face value as they scour for even more funding dollars.
On the other hand, it’s undeniable that, despite the debate about its applicability, genAI represents a technological revolution. The technology itself is formidable, and its positive impact on at least personal productivity is undeniable. Yet it remains vastly unclear, as the MIT study demonstrates, whether it will ever be able to justify its technological costs.
A Bubble?
These two sides of the same coin can be true at the same time: genAI is one of the most important technologies of all time, and we’re in a bubble regarding its potential and applications. Remarkably, this admission comes from Sam Altman himself. OpenAI’s CEO, in a recent interview with US reporters, tried to deflate expectations he helped set up in the past:
“Are we in a phase where investors as a whole are overexcited about AI? My opinion is yes. Is AI the most important thing to happen in a very long time? My opinion is also yes.”
To make matters worse, all generative AI services we’re currently using personally and at work (including ChatGPT, Claude, Cursor, Microsoft Copilot, and Google Gemini) are heavily subsidized by either investors’ or companies’ money. While the combined number of their active users already exceeds one billion, both OpenAI and Anthropic will close 2025 reporting billions in revenue and even more in losses.
READ ALSO
Like Uber?
The playbook we’re seeing unfold isn’t far removed from that of other hyperscale platforms like Uber. A moment will come when prices must rise to start returning capital to the large number of investors. Uber delayed that moment for years, collecting eager investors’ money with the far-fetched promise of autonomous driving—until that didn’t work anymore.
OpenAI and Anthropic are doing the same, dangling the promise of AGI (artificial general intelligence) or ASI (artificial super intelligence) to collect billions in funding while waiting long enough for the technology to become indispensable. But while ride-hailing had immediate benefits in disrupting an established industry with lower cost solutions, the AI startups’ bet is far bolder and definitely way more expensive to maintain.
A Skeptic at Goldman Sachs
Jim Covello, the Head of Global Equity Research at Goldman Sachs, is among the genAI skeptics. He says that to earn a relevant return on investment, gen AI should be able to solve extremely complex problems that justify its immense cost.
In a Goldman Sachs report published in 2024, Covello explains:
“We estimate that the AI infrastructure buildout will cost over $1tn in the next several years alone, which includes spending on data centers, utilities, and applications. So, the crucial question is: What $1tn problem will AI solve? Replacing low- wage jobs with tremendously costly technology is basically the polar opposite of the prior technology transitions I’ve witnessed in my thirty years of closely following the tech industry. While the question of whether AI technology will ever deliver on the promise many people are excited about today is certainly debatable, the less debatable point is that AI technology is exceptionally expensive, and to justify those costs, the technology must be able to solve complex problems, which it isn’t designed to do.”
One year after Goldman Sach’s report, we are still very much hearing the same narrative, with AI companies swearing by the scaling myth, saying all they need is just a bit more data, just a bit more training, and just a bit more investor money to get all that.
A $1 Trillion Question
It’s a $1 trillion question: what happens when the financial realities can no longer be delayed, with investors and companies realizing that the chasm between costs and applications can’t be filled?
History suggests that bubbles burst when the gap between investment and practical returns becomes unsustainable. The dot-com crash of 2001 offers a sobering reminder of what occurs when investor enthusiasm dramatically outpaces actual utility, even though the fundamental technology (the Internet) was so important that it would later become ubiquitous.
If businesses begin demanding concrete returns on their AI investment and find them lacking a significant market correction could follow. Companies that have built their valuations on AI promises may face a harsh reckoning with reality, negatively affecting the global economy as a consequence.
READ ALSO