Warnings that a major AI bubble has formed, or is rapidly in the process of formation, as OpenAI announces a series of deals with chip suppliers to build its computing power, are coming thick and fast.<\/p>\n
The alarms are being sounded because the deals involve a high degree of circularity whereby, in one form or another, the supplier of the chips and technology provides financial assistance to OpenAI to enable it to purchase its products.<\/p>\n
According to calculations by the Financial Times (FT), the results of which have been widely circulated, deals done by OpenAI, the maker of ChatGPT, with AMD, Nvidia, Oracle CoreWeave, and other smaller companies amount to about $1 trillion in cost. The company will need more than 20 gigawatts of computing power over the next decade, with the electricity needed to power it equivalent to the output of 20 nuclear reactors.<\/p>\n
As the FT reported: \u201cOpenAI is burning through cash on infrastructure, chips, and talent, with nowhere near the capital required to fund these grand plans.\u201d<\/p>\n
OpenAI’s CEO Sam Altman gestures while speaking at University College London as part of his world tour of speaking engagements in London, on May 24, 2023. [AP Photo\/Alastair Grant]<\/p>\n
The company has yet to make a profit, and its founder and CEO Sam Altman has said that profit-making is not really on his horizon at present. Speaking earlier this week, he said becoming profitable was \u201cnot in my top-10 concerns.\u201d<\/p>\n
\u201cObviously,\u201d he continued, \u201csomeday we have to be very profitable,\u201d and the company would get there, but \u201cright now\u201d it was in a \u201cphase of investment.\u201d In other words, it is taking a trillion-dollar gamble that the massive investments will eventually pay off.<\/p>\n
But others say a different dynamic is at work. As Gil Luria, an analyst at the investment bank and financial services firm DA Davidson told the FT: \u201cOpenAI is in no position to make any of these commitments.\u201d It was expected to make a loss of around $10 billion this year.<\/p>\n
\u201cPart of Silicon Valley\u2019s \u2018fake it until you make it\u2019 ethos is to get people to have skin in the game. Now a lot of big companies have a lot of skin in the game on OpenAI.\u201d<\/p>\n
Their connection has been established through the circular deals through which OpenAI is being financed.<\/p>\n
Under the $100 billion deal with Nvidia, the leading maker of the graphic processing units (GPUs) which form the basis for the development of AI, the money it supplies to OpenAI will be used to buy more Nvidia chips.<\/p>\n
While Altman has said the deal with Nvidia is a \u201cnew financing model,\u201d it is very much a repeat of the kind of circular deals which were made in the dot-com bubble at the start of the century before it collapsed with billions of dollars in losses.<\/p>\n
And the deal with the chipmaker AMD has added a new twist to the circularity. Under the agreement, AMD will provide OpenAI with warrants to buy up 10 percent of the company\u2019s shares for just one cent a share provided certain targets are met. As the price of AMD shares rises, OpenAI will be able to cash in and use the money to buy AMD chips.<\/p>\n
AMD chief executive Lisa Su described the deal as a \u201cpretty innovative structure, which didn\u2019t come lightly.\u201d<\/p>\n
In essence, the agreement means that the whole operation is being financed on the expectation that the surge in high-tech and AI stocks will continue indefinitely.<\/p>\n
Long-time Wall Street short seller Jim Chanos, who described financial markets as having entered \u201cthe golden age of fraud\u201d back in 2020 and who commented recently that this phenomenon had \u201cdone nothing but gallop even higher\u201d since then, pointed to one of the key contradictions in the circularity deals.<\/p>\n
In a social media post on the AMD deal, he wrote: \u201cDon\u2019t you think it\u2019s a bit odd that when the narrative is \u2018demand for compute is infinite,\u2019 the sellers keep subsidizing the buyers?\u201d<\/p>\n
There is a clear recognition that an AI bubble\u2014a situation characterized by the complete divorce of market valuations from underlying real economy\u2014is fast developing. As well-known financial journalist Andrew Ross Sorkin noted in the New York Times this week, he was asked about this issue multiple times a day at a meeting in San Francisco.<\/p>\n
His answer was that \u201cwe are\u2014we just don\u2019t know when it will end. When the dot-com bubble popped, it wasn\u2019t the end of the Internet. It just meant that the weakest companies didn\u2019t survive. And that\u2019s probably what will happen again.\u201d<\/p>\n