AI bubble warning: AI investments are a negative return for most tech giants

Title: The Impossible Math of the AI Boom
Author: Joachim Klement
Translation: Peggy, BlockBeats

Author: BlockBeats

Source:

Reprint: Mars Finance

Editor's note: The AI boom is moving from a technological narrative into a phase of financial validation.

Over the past year, market discussions about AI have focused more on model capabilities, the computing power gap, and application prospects, but this article reminds us that what truly needs to be calculated is the capital return behind this boom. Microsoft, Alphabet, Amazon, Meta, Oracle, and other large-scale cloud service providers are investing hundreds of billions of dollars into AI data centers, and according to current analyst expectations for revenue and capital expenditure, most companies—except Amazon—may have implied negative returns on investment.

This means that the similarity between the AI bubble and the previous internet bubble is not just market sentiment, but the high binding of capital expenditure to macro growth and stock price expectations. The author points out that 93% of US GDP growth over the past four quarters can be explained by technology investments; if cloud providers cut back on data center, chip, and infrastructure investments, not only will industry chain companies like Nvidia, TSMC, and ASML be impacted, but the US economy itself could also quickly come under pressure.

More concerning is that if AI companies like OpenAI and Anthropic push for IPOs at market sentiment peaks, it may not just be a financing event but a risk transfer: early capital and existing shareholders will shift the uncertainties of the AI narrative onto retail investors, pension funds, and other investors willing to continue buying into growth stories.

The core question of this article is not whether AI has a future, but who will pay for this costly infrastructure race once the marketing hype subsides.

Below is the original text:

In December 1996, Federal Reserve Chairman Alan Greenspan described the prosperity of technology, media, and telecom stocks at the time as showing signs of "irrational exuberance." Nearly 30 years later, we can make the same judgment about today’s AI boom.

However, although this round of tech enthusiasm shares similarities with the previous internet bubble, there is an important difference: part of today’s boom has already far exceeded the scale of the TMT bubble back then. By 2025, US corporate investments in IT equipment and software will approach $1.5 trillion. At the peak of the TMT bubble, this figure was $466 billion, which, even adjusted for inflation, is only about $829 billion.

In fact, the US economy is currently almost entirely driven by the tech boom. Based on my calculations, 93% of US GDP growth over the past four quarters can be explained by technology investments. Even at the height of the TMT bubble, this ratio barely reached 60%.

Developers of large language models like OpenAI and Anthropic are preparing for major IPOs later this year to capitalize on investor optimism about their growth prospects. Meanwhile, Microsoft, Alphabet, Amazon, Meta, and Oracle plan to invest hundreds of billions of dollars over the next five years into data centers to support these models.

The problem lies here: the math behind the AI boom is becoming increasingly complex. I have collected analyst consensus expectations for these large cloud providers’ capital expenditures and revenues from 2025 to 2030.

Over these five years, their capital investments are expected to grow at an unprecedented rate of 20% annually. Meanwhile, revenues are projected to grow 15% per year. If we make an extremely bold assumption—that these companies have no costs—then the additional revenue can be viewed as their profit from the new investments in AI data centers. However, even under this extreme optimistic assumption, the implied return on investment for all companies except Amazon is highly negative.

These figures suggest that if these large cloud providers continue on their current trajectory, the AI boom could turn into one of the largest shareholder value destruction events in history. But they still have two options.

The first is that the rise of AI will generate far more revenue than currently expected for these companies. But this also presents a mathematical challenge: if these large cloud providers aim for a 10% return on investment, they would need to find an additional $2 trillion to $5 trillion in annual revenue. For companies with current annual revenues totaling only $1.5 trillion, this is nearly impossible.

The second option is that investments planned for data centers, chips, and other areas will ultimately not materialize. This could be due to investors becoming more cautious about the industry or because debt financing for data centers becomes more difficult.

So, what happens if these companies announce cuts to some of their investment plans?

From Nvidia to ASML, Samsung, and TSMC, the stock prices of the largest companies are built on these investment plans and the demand expectations they generate.

Remember, US GDP growth is currently entirely driven by rising tech expenditures. If these expenditures begin to decline, the US economy could quickly slip into recession. Even a modest decline of 4% to 6% in tech investments could have this effect. Historically, after smaller-scale tech booms in the 1960s and during the 2009 recession, similar investment declines occurred.

Such a moderate correction in investment spending could still push US, UK, and European stock markets into new bear markets. The replay of the early 2000s tech stock crash, where markets fell 50% or more in the first year, is a real risk.

The next question is: when might we see these large cloud providers announce investment cuts?

I believe this is unlikely to happen before 2026. Companies like OpenAI and Anthropic will still strive to maintain market enthusiasm at least until their IPOs are completed, which could support this boom in the short term. But afterward? The "impossible math" faced by these large cloud providers will not change, and the marketing hype will eventually fade. Reality will arrive.

Perhaps not in 2026, but possibly in 2027 or 2028. After all, Greenspan spoke of "irrational exuberance" in December 1996, and the bubble burst three years later in 2000.

From this perspective, the IPOs of these AI companies are likely just a large-scale transfer of investment risk: risk is shifted from current owners to retail investors, pension funds, and other investors willing to buy into the narrative.