OpenAI and Anthropic now sit at the center of Big Tech’s AI cloud backlog

The AI boom now has one very ugly question hanging over it. Is the money real, or are Big Tech companies just feeding cash to AI startups and booking the same cash as cloud sales later?

That question now sits right on top of OpenAI and Anthropic, because fresh filings show both companies are tied to more than half of the almost $2 trillion in future cloud revenue sitting on the books of Microsoft (MSFT), Oracle (ORCL), Alphabet (GOOGL), and Amazon (AMZN).

It sounds too good to be true, and yes, it is wild. A tech giant invests billions in an AI firm through some financing agreement, and in that agreement, the AI firm is advised to deploy its funds on purchasing cloud infrastructure owned by the same tech giant.

And so, the AI firm receives funding, the cloud firm makes income, and Wall Street enjoys looking at some impressive figures. The money does not get very far, however. It goes out through one door and returns through another door in the guise of a new customer.

Microsoft books OpenAI cloud spending after funding the same customer

Microsoft’s OpenAI collaboration serves as an illustrative example. Microsoft spent close to $13 billion on funding OpenAI; however, this investment was not limited to cash contributions only. The majority of that investment consisted of Azure credits, which OpenAI used to develop and execute its AI models using Microsoft infrastructure.

The usage of the Microsoft servers by OpenAI generated revenues for Microsoft. As a result, Microsoft contributed financially to OpenAI’s activities, OpenAI used Microsoft resources to execute them, and Microsoft recognized that contribution as demand from its customers.

OpenAI’s cloud bill has now climbed above $60 billion a year. Its revenue is around $25 billion. That means its server costs are more than double what it brings in. For a normal company, that would look like a giant red flag. In AI land, it gets treated like growth.

Anthropic is running a similar play with Amazon. The company spent about $2.66 billion on Amazon Web Services in nine months. That was roughly the same size as its revenue at the time. So the money coming in was almost matched by the money going straight back out to AWS.

That is where the second part of the scam plays out. With more money flowing into Anthropic or OpenAI at a higher valuation, the technology giants that have invested in them can inflate the value of their stakes to make money without having sold any goods or collected any cash. A gain has been made.

Google’s parent company, Alphabet, earned $62.6 billion in the first quarter of 2026. $28.7 billion was attributed to Google’s gains in relation to its stake in Anthropic. Amazon posted $30.3 billion in earnings in the first quarter of 2026. Its Anthropic gains accounted for $16.8 billion of it.

Amazon burns real cash while AI paper gains lift reported profit

However, Amazon’s cash metrics appeared to be in a more difficult position. Free cash flow fell by 95% to $1.2 billion, and the company also invested $44.2 billion into physical data centers. This clearly demonstrates the difference between accounting profits and real cash. One sits in spreadsheets, while the latter builds real-life data centers using land, semiconductors, electricity, cooling, connections, buildings, and personnel.

This could lead to concentration risks for both companies. In particular, Microsoft has 49% of its $627 billion future backlog dependent on OpenAI. On its part, Oracle has 54% of its $553 billion future pipeline dependent on OpenAI alone.

This all looks eerily familiar to something straight out of the dot-com era. Back in 2001, when Global Crossing and Qwest Communications traded equal fiber network capacity and recorded such swaps as sales. As a result, Qwest lost $1.4 billion in fraudulent revenue. Meanwhile, Global Crossing filed for bankruptcy. The only thing that separates both cases today is the fact that such swaps by telecommunications companies were not considered legal at that time, while today’s AI cloud loop easily fits in today’s accounting rules.

According to the Kobeissi Letter, the ten largest American stocks constitute 41% of the S&P 500. Among these stocks, we find Magnificent Seven, including Apple and Tesla. This percentage is 14 points above the previous dot-com peak in 2000.

“This means about 41 cents of every dollar invested in the S&P 500 flows directly into shares of just 10 firms,” The Kobeissi Letter wrote. “Roughly 35 cents of every dollar flows specifically into the Magnificent 7 group. All while nearly 50 cents of every dollar is now going into AI-linked stocks. Mega-cap tech is all that matters right now.”

Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.