The AI boom has created a circular financing system in which a relatively small group of companies invest in one another, finance their customers, buy one another's products and then use the resulting contracts as evidence of future growth.
Summary of a strongly argued and openly bearish critique of the current AI investment boom, centred on Oracle, Larry Ellison and OpenAI.SummaryOracle is presented as the clearest example. Its enormous $300 billion OpenAI computing contract helped propel expectations for Oracle's future cloud revenues and contributed to the extraordinary rise in Larry Ellison's wealth. But OpenAI remains loss-making, meaning that Oracle's expected revenues ultimately depend on OpenAI being able to finance and monetise enormous amounts of computing capacity.
The article argues that this phenomenon extends across the AI ecosystem. Nvidia invested in OpenAI while OpenAI would use much of that capital to acquire access to Nvidia hardware. Nvidia is simultaneously an investor in CoreWeave, whose business depends heavily on Nvidia GPUs, and has agreed to purchase some CoreWeave capacity that cannot otherwise be sold. Microsoft is simultaneously an OpenAI shareholder, computing supplier and competitor. The author sees these relationships as creating financial circularity rather than completely independent sources of demand.
The most important development: the bond market is becoming sceptical
According to the article, credit markets are beginning to impose financial discipline that equity markets initially did not. S&P reportedly downgraded Oracle to BBB-, only one notch above junk, while Oracle faces very large capital expenditures, negative projected free cash flow and approximately $160 billion of debt.
The author's central distinction is therefore between the equity market, which rewarded announcements of enormous future AI revenues, and the credit market, which increasingly asks a simpler question:
Where will the cash ultimately come from?
The article describes Oracle's dependence on OpenAI as particularly significant because OpenAI reportedly represents roughly half of Oracle's enormous contracted backlog.
The second risk: AI prices are falling
Even if AI demand continues to grow enormously, competition could destroy the pricing assumptions underlying today's infrastructure investments.
It points particularly to Chinese AI companies such as DeepSeek and Moonshot, which are offering increasingly capable models at dramatically lower prices than leading American providers.
This produces what the author considers a fundamental problem:
Massive infrastructure investment requires enormous future revenues, but increasing competition may make AI itself progressively cheaper.In other words, AI usage could explode while returns on AI infrastructure disappoint.
OpenAI as the central concentration risk
The article claims that contracts connected with OpenAI represent more than $330 billion of approximately $880 billion in future contracted revenue across Microsoft, Oracle and CoreWeave.
It also contrasts OpenAI's roughly $20 billion annualised revenue with approximately $1.4 trillion of infrastructure commitments over eight years. The author's point is not that all $1.4 trillion is immediately payable, but that the difference between present revenues and future commitments is enormous
ORIGINAL ARTICLE : The House of Ellison is on the Brink Author Mike Brock