The AI Paradox: Why Aren't the Promised Productivity and Climate Gains Showing Up?
Originally published in Turkish. Read the original →
Big Tech's promises of revolutionary productivity and climate solutions from AI don't match the hard data we have today. We look at the disappointment AI has caused in the business world and the ecological cost of ever-growing data center infrastructure.
For years, AI has been marketed as an unprecedented revolution in the business world, an economic leap forward and a savior against the climate crisis. Today it faces a serious “reality check.” Recently published research and comprehensive datasets show a widening gap between the billion-dollar promises of the tech giants and the concrete results on the ground. Across a broad spectrum, from corporate productivity reports to climate impact, AI has not yet delivered the expected miracle.
The Solow Paradox of Modern Times
Technology failing to show up in output statistics is nothing new in economic history. In 1987, Nobel laureate economist Robert Solow pointed to this strange situation with the remark, “You can see the computer age everywhere but in the productivity statistics.” Indeed, productivity growth, which had hovered around 2.9% between 1948 and 1973, fell to 1.1% after 1973, just as computers were beginning to integrate rapidly into business culture. That decline only recovered after the mid-1990s.
Today, some economic analysts are proposing a similar “J-curve” theory for AI. According to this view, in the first phase of technological adoption internal company dynamics are poorly aligned and productivity drops, while a dramatic leap follows once the right workflows are in place. The real question, however, is when this process will be complete and whether the gain in output will justify billions of dollars of AI investment.
According to Executives, Employment and Output Aren't Changing
A large-scale survey by the National Bureau of Economic Research of 6,000 senior executives in the US, the UK, Germany and Australia produced some striking findings. About 90% of these companies say that AI's impact on overall productivity or employment over the past three years has been negligible.
Although two-thirds of the executives report that an AI tool is used within their company, when weekly AI usage per person is measured, the figure averages no more than 1.5 hours. Worse still, a quarter of the executives say they use no AI functionality at all in their own work.
Yet despite all this, market forecasts keep painting a rosy picture. Analysts expect overall productivity momentum to grow by 1.4% and company output by 0.8% over the next three years. On layoffs, companies expect a 0.7% contraction, while employee-focused surveys expect a small 0.5% increase. Meanwhile, even though the word “AI” was used liberally in corporate meetings throughout 2024–2025, that enthusiasm has barely shown up on balance sheets.
Nobel laureate Daron Acemoğlu and 2024 research reports from MIT also point out that the cumulative macro-productivity contribution to be expected from the AI revolution may stay at only around 0.5% over the next 10 years, suggesting that expectations are heavily inflated.
The Climate Crisis: Solution or Source of the Problem?
Beyond productivity, another major marketing argument for AI was its potential to radically cut carbon emissions. Yet according to a report by energy researcher Ketan Joshi, who put the tech giants under scrutiny, only a quarter of the 154 different “AI is climate-friendly” claims presented to the public by tech and energy companies rest on actual academic papers. The remaining claims consist either of entirely unsupported statements or commercial expectations.
Google's claim that “thanks to AI, we can cut global emissions by 5% to 10% by 2030” is at the very center of the debate. Such a huge claim (a cut equivalent to the entire annual carbon footprint of the EU countries) rests not on academic research but on PR material that Google produced with a consulting firm based on “user scenarios.” The irony is that in the sustainability report Google published the same year, its own data showed that relentless data center investment had pushed the company's emissions up compared with the previous year.
Today, the insatiable energy demand of large language models (LLMs) can lead to newly planned fossil fuel power plants and even the return of coal plants that should have been shut down. While traditional, lightweight machine learning applications used for years have made grid distribution more efficient, the “large, generative AI” models that arrived on the scene in 2023 have created an enormous energy black hole rather than protecting the ecology.
It seems that AI needs far more time, effort and structural transformation than it claims in order to revolutionize both economic output and the well-being of the planet.
