The longer the AI-fuelled stock market boom continues, the more new and risky schemes are developed to try to sustain it. The AI boom has already seen the growth of circular financing in which companies provide money to others to enable them to buy their products.
Now a new mechanism is being developed. Earlier this week, the leading chipmaking firm Nvidia, the biggest company in the world by market value at more than $5 trillion, announced that it had reached an understanding with some of Wall Street’s biggest names to provide financing for purchase of its AI chips.
The agreement, which involves a consortium of Apollo Global, Blackstone, BlackRock, Brookfield Asset Management, Goldman Sachs and KKR, provides that they will partner with Nvidia to provide $500 billion to invest in the AI buildout.
Under the deal, the final details of which have still to be determined, the firms will provide capital to finance purchases of Nvidia chips “at attractive rates for Nvidia customers.”
The aim of the new system is to develop an alternative to the previous method of circular financial whereby Nvidia directly provided money to firms wanting to buy its products.
Announcing the new financial arrangement, Nvidia founder and chief executive Jensen Huang said: “We began by building chips; today we are creating a new class of productive, investable infrastructure: AI factories.”
Voicing the sentiments of the Wall Street firms involved in the agreement, Jon Gray, president and chief operating officer of Blackstone told the Financial Times (FT): “We continue to be enormous investors globally across the Nvidia ecosystem, and this announcement underscores our confidence in their platform and the future of AI infrastructure.”
Pointing to the most essential feature of the agreement, Huang said it was the “first time that technology chips have become an investable asset class,” functioning like productive “infrastructure.”
Wall Street has developed a complex system of credit and debt arrangements based on assets such as cars, housing, buildings and income streams. But this system has assumed that the underlying asset will retain at least some of its value over time. But the value of computer chips is notoriously volatile—a chip can be top of the range at one point and then rapidly become obsolescent at another as a superior or cheaper product is developed.
Reporting on the agreement under the headline “Wall Street giants bet Nvidia’s AI chips will defy the laws of finance,” the FT noted that it was “betting that AI chips can defy one of finance’s basic rules: that fast-moving technology quickly loses its value.”
The issue was raised even more pointedly by Jack Albin, a founding partner at the $260 billion family office firm Cresset, an investor in Nvidia, to the Wall Street Journal (WSJ).
He said the proposed deal was “great for Nvidia, which needs its customers to have access to capital. But if you’re a debt investor relying on compute power as collateral? I mean, historically, that’s an asset that’s had the shelf life of a lettuce.”
Seeking to boost the deal and counter such observations based on historical experience, Huang said Nvidia still had customers using old generation chips and this was a sign they retained their value for a long time after their release and that the extreme demand for Nvidia’s chips would mean they would make excellent collateral unlike the past. In other words, in the often-used phrase “this time it’s different.”
“Nvidia’s AI factory platform is really an investable asset, an infrastructure asset,” he said. “The reason for that is because it’s productive, it’s revenue generating.” That latter comment covers over the fact while AI chips are certainly returning a profit in some areas, there is yet to be income and profit generated by the vast data centres now under construction.
And there is the prospect that the technology on which they are based could be rapidly superseded or undercut by a cheaper chip developed in China.
Speaking to the business channel CNBC Ben Emon, a portfolio manager at the financial firm Pimco, said “depreciation is the one key risk here” with one of the biggest threats to Nvidia’s financing model coming from Chinese competition.
Summarising his remarks, CNBC said: “If Chinese production pushes hardware prices into freefall, the collateral backing hundreds of billions in private loans could erode faster than the debt itself, leaving investors exposed to losses.”
BlackRock chief executive Larry Fink made a perhaps inadvertent reference to the financial dangers in the proposed system. In comments to the WSJ, he likened the financing of AI computing power to what took place at the start of his career as banks and investments firms pioneered mortgage-backed securities in the 1970s.
The development of that set of operations led to the sub-prime crisis which was the trigger for the global financial crash of 2008.
History, of course, does not repeat itself. But there are parallels and they lie in the fact that firms which take up the new methods for raising funds will be among those who are the financially weakest among their AI cohorts and not able receive backing from other sources.
The WSJ began its report by noting that Huang was “running into a problem: Many of his customers can’t afford to buy his company’s coveted AI-powering chips.”
Describing the proposed agreement as “exotic money pipeline for the AI boom,” it said that according to its critics it was a “system that will cover up weaknesses in some corners of the market.”
The issue was not the heavyweights, such as Meta, Microsoft and Google (although it should be noted there are concerns about their level of debt) but was “about the smaller AI labs, cloud companies and enterprises that have ravenous demand for Nvidia chips but face high interest rates if they want to finance purchases of them.”
Such companies, however, do play a part in the overall boom. “The AI boom, and Nvidia’s momentum can’t keep up at this blistering pace unless those kinds of companies get the hardware they need,” the article noted.
To organise the operation, the big Wall Street funds will seek money from pension funds, insurance companies and sovereign wealth funds, tying them into any possible failure.
The proposed agreement is illustrative of the dynamic of the AI-based financial bubble. Nvidia’s motivations are clear: it is a way of lessening the circularity of its present financial arrangements, though it will still be on the hook for 25 percent of any losses.
But what of the involvement of major Wall Street firms and their boosting of the scheme? It is an expression of the need to preserve the value of their existing investments—running into hundreds of billions of dollars—by ensuring that the boom continues.
In other words, not to back the development of ever more exotic and risky financial mechanisms risks bursting the entire bubble. But such a process—sometimes likened to riding a bicycle where forward momentum must continue to stop it falling over—cannot continue indefinitely.
As investor Michael Burry, who became famous and very wealthy by shorting mortgage-backed securities prior to 2008, noted: “Structuring credit is a natural part of the system. Structuring unnatural credits late in the bull phase is where the worry comes in.”