[Digital Today reporter Jinju Hong (홍진주)] An analysis says big tech companies' artificial intelligence (AI) investment and cloud revenue structures are reinforcing concerns about an AI bubble. It says a repeated pattern in which money invested in AI startups flows back into the same companies' cloud revenue is strengthening a dependence on book growth rather than real profit.

BeInCrypto, a blockchain media outlet, reported on May 25 local time that a significant portion of future cloud contracts at major big tech companies such as Microsoft, Amazon, Google and Oracle is concentrated in a few AI startups, including OpenAI and Anthropic.

Critics' main concern is a so-called “round-trip funding cycle”. Big tech companies make large investments in AI startups, and those startups then spend money using the same companies' cloud services. It means the investment money is recycled internally as revenue, rather than bringing in new cash from outside.

A leading example cited is the relationship between Microsoft and OpenAI. Microsoft has invested about $13 billion in OpenAI so far, and a significant portion is known to have been provided in the form of Azure cloud credits. OpenAI used them to train and operate AI models, and Microsoft recognised that usage as commercial revenue.

The issue is a gap between actual profitability and the cost structure. The outlet said OpenAI's annual cloud costs exceeded $60 billion, while actual revenue was presented at about $25 billion. As investments and revenue become intertwined, headline growth rates rise, but questions are being raised about whether the structure can generate sustainable cash.

The relationship between Amazon and Anthropic is also cited as a similar case. Anthropic is known to have spent about $2.66 billion on Amazon Web Services (AWS) over the past 9 months, which was analysed as a scale that nearly matches the company's total revenue. One analyst assessed that “the entire AI boom may have been built on fake revenue.”

Such a structure is also affecting corporate results. Each time an AI startup conducts a new funding round, big tech companies can revalue the worth of their stakes and reflect it in net income. Alphabet disclosed that about $28.7 billion of its $62.6 billion net profit in the first quarter this year came from revaluing its stake in Anthropic. Amazon was also reported to have linked about $16.8 billion of its $30.3 billion net profit in the same period to similar valuation gains.

Cash flow, however, appears relatively weak. Amazon's free cash flow (FCF) fell 95 percent from a year earlier to about $1.2 billion. Over the same period, it spent $44.2 billion on physical infrastructure investment such as data centre construction. That is behind criticism that despite massive AI investment, actual cash-generating power is weakening.

Reliance on the cloud business is also rising. Microsoft was analysed as having about 49 percent of its $627 billion backlog of future orders linked to OpenAI, and Oracle was also reported to be relying on a single customer for 54 percent of its total pipeline.

Another variable is that the burden of AI adoption costs is rising rapidly in the operational stage of companies. Uber is known to have provided engineers with Anthropic's Claude Code and Cursor, then used up most of its AI coding budget this year by April. Some employees' monthly API usage costs ranged from $500 to as much as $2,000. Microsoft was also mentioned as having stopped internal employees from using Claude Code, separate from its multi-billion-dollar partnership with Anthropic.

Nvidia vice president of applied deep learning Brian Catanzaro (브라이언 카탄자로) also acknowledged the burden of AI operating costs publicly, saying, “For our team, compute costs are much larger than labour costs.” The industry sees a strong possibility that total AI spending will continue to rise if usage of agent-type AI increases even if chip prices fall.

The mood is shifting from growth itself to whether AI businesses can cover their own costs. One analyst pointed out that “even the first companies to use AI at scale are showing situations where they cannot afford the costs.”

There is also analysis that the cryptocurrency market is not free from the impact. Bitcoin was found to have recorded a high correlation of about 0.75 with the Nasdaq as of January this year. That has raised concerns that if AI investment flows centred on Nvidia and OpenAI wobble, shocks could spread across the broader digital asset market.

Mainstream finance has also begun checking the possibility of an AI bubble. Fidelity recently presented five key indicators that can determine whether there is an AI bubble, and assessed that warning signals are already appearing in the quality of big tech companies' earnings and their ability to bear capital expenditure.