November 20, 2025
On November 17 Bloomberg reported that Peter Thiel's hedge fund sold its 537,742 shares of Nvidia during the third quarter. This month SoftBank sold its stake in Nvidia reportedly worth $5.8 billion. Bloomberg sees the sales as a retreat from investments in Nvidia, the leading provider of artificial intelligence chips. However, SoftBank says it sold its Nvidia shares in order to increase its investments in OpenAI. These are large numbers, but small compared to Nvidia's $4.6 trillion market capitalization. It is the size of the market capitalization and the large depreciation that might be the problem.
Michael Burry's expressed uneasiness about problems resulting from very rapid developments in AI, especially as they relate to Nvidia and Palantir (Nvidia makes AI chips and Painter develops AI software) has focused some attention on the depreciation problem resulting from rapid change that could bring down share prices in the AI arena, including Google and Microsoft.
Michael Burry was the only person who saw in advance the collapse of the mortgage derivative market. No one in the financial sector saw it coming, and neither did the Secretary of the Treasury, the Chairman of the Federal Reserve, and the Chairman of the Securities and Exchange Commission. These three stooges went to Congress and stopped Brooksley Born, Chairman of the Commodity Futures Trading Commission, from regulating derivatives.
We can conclude from this that the people assumed to be smart are not. The only smart people are the few who can get outside of the box and look at reality objectively.
Larry Sparano (On Target) recently interviewed me about the situation.
A Possible AI Collapse
The basic problem seems to be that AI investments rapidly depreciate because of the rapid rate of change, with the result that the rapidly rising depreciation requires ever more new investments into the industry.
Some of the financing seems to be circular. Here is a possible scenario. The company OpenAI contracts to spend more money than they have. The money is to go to Oracle. On the basis of expected incoming revenue from OpenAI, Oracle announces it is going to buy massive amounts of chips from Nvidia. Nvidia then makes a $100 billion investment in OpenAI. It other words it is a credit circle.
The massive data centers being built by Google and Microsoft are not a one-time investment, they are a huge recurring expense. The data centers depreciate rapidly because of rapid developments such as the displacement of CPU processors by the faster GPU processors. If the companies underestimate the speed at which their investments are depreciating, they could be faced with a depreciation tsunami with depreciation write downs exceeding new investment inflows and cash flow. It is possible to have profits and no cash flow, because reported profit doesn't include the capital spending necessary to keep up with rapid technological development. For now it seems that new investments into the companies are taking the place of profits, and it is new investments into AI that are creating the bubble. However, expenses cannot forever exceed profits without stock prices and high P/E ratios collapsing.
One way to hide the problem is to extend the expected life of the investments from, say, three years to six years. This gives a longer time to spread out the depreciation and thereby a less of a hit on cash flow. But if the life of the assets turns out to be three years, the company is in trouble.
Yet another problem is that rapidly depreciating computer units are being used as collateral to finance loans, and new businesses called neoclouds borrow money to use to offer computing units for rent. As the business is rapidly growing, there quickly will be hundreds of billions of dollars in loans based on rapidly expiring collateral.
If I have managed to understand the situation, the AI boom is dependent on new money entering the industry to provide the money to renew rapidly depreciating assets. If the inflow of new money reduces or halts, the bubble pops.
As my readers perhaps know, I am more interested in the societal affects of AI than in a market crash from an AI bubble. The societal effects are more deadly than a market crash.