Debate Intensifies Over Scale of AI Boom

Recent data highlight both the reach and the limits of current artificial intelligence deployment. OpenAI reports that more than 800 million people now use its large language models each week, a little over three years after ChatGPT sparked the latest AI boom.

Workplace usage has also expanded. As of November 2025, 41% of American workers said they used AI in their jobs, up 10 percentage points from a year earlier, according to research from the Federal Reserve Bank of St. Louis.

At the company level, adoption is particularly strong in the information sector, where 37.5% of firms actively use AI, research by The Motley Fool shows. However, the intensity of use is lower than headline numbers suggest.

Only 13% of workers report using AI every day, and on average Americans spend 5.7% of their working hours with AI. The articles characterize this as meaningful but not yet matching the sweeping change implied by some market valuations.

Supporters of the AI investment boom point to accelerating spending and robust corporate fundamentals. Meta (META), Microsoft (MSFT), Alphabet (GOOGL) and Amazon (AMZN) have collectively signaled more than $500 billion in AI-related capital expenditures for this year.

The companies driving today’s boom are described as enormously profitable businesses with large cash flows, in contrast with loss-making firms that dominated earlier technology bubbles. This distinction is presented as a key difference from the dot-com era.

Technological progress is another pillar of the bullish view. The articles note that so-called agentic AI systems, which can autonomously execute multistep tasks, are nearing prime time.

If such systems progress to reliably handle complex workflows now requiring humans, the pieces argue the economic implications could be enormous. While this is acknowledged as uncertain, the trajectory of improvement is described as real.

On the cautionary side, the pieces highlight stretched market pricing. The cyclically adjusted price-to-earnings ratio for the overall stock market is said to be far above normal levels, exceeded only at the peak of the dot-com bubble and briefly during the Covid-19 earnings collapse.

AI-linked revenues are also described as vulnerable to circularity. A significant portion of current AI income is reported to come from companies selling to other companies within the ecosystem, rather than from end users outside it.

The articles state that organic, external revenue needs to expand dramatically for prevailing valuations to be justified. They suggest that this reliance on intra-industry demand is a structural weakness of the present boom.

Investor patience is portrayed as thinning, with concerns that the return on massive AI investments from outside the ecosystem has yet to be clearly demonstrated.

Leverage in AI infrastructure is identified as a key, and often overlooked, risk. The build-out of data centers that power AI is described as heavily debt-financed.

Companies such as CoreWeave are cited as having taken on high levels of leverage to fund capacity, effectively betting that AI demand will grow fast enough to service their obligations.

Historical parallels are drawn, noting that cheap and abundant capital supported previous bubbles until interest rates rose or lending tightened, at which point highly leveraged players were the first to face stress.

The pieces list potential macro triggers, including sticky inflation, a sluggish labor market, geopolitical tensions and escalating conflicts, as factors that could test the resilience of indebted AI infrastructure.

Despite rising adoption and investment, measurable productivity gains from AI are described as limited so far. A Bank of England survey cited in the articles found that nine in 10 senior managers said their firms’ AI initiatives had produced no observable impact on labor productivity.

This reported disconnect between large market valuations and the absence of clear productivity improvement is presented as a central tension in assessing where the current AI cycle stands.