Over the past months there have been repeated warnings from major financial and regulatory bodies, including the International Monetary Fund, the Bank for International Settlements, the Bank of England, the Financial Stability Board and others, that the AI boom has the potential to set off a crisis.
Last week provided clear evidence that these warnings are by no means unfounded as AI-related firms experienced severe turbulence which, in one case, was likened to being hit by a tornado.
The company in question was CoreWeave, a firm which started as a crypto coin miner but then shifted its operations to become a so-called neocloud firm which buys up large amounts of computer chips from Nvidia financed by tens of billions of dollars of debt and which then rents out its compute capacity to train AI models.
It experienced a plunge of as much as 28 percent in its share price. Even more significant, the yield on its bonds rose to as high as 15 percent, well and truly into what is known as “junk bond” territory. The cost of insuring investors against it going bankrupt, via credit default swaps, rose rapidly.
According to a report in the New York Times, at one point “investors placed a more than 60 percent likelihood CoreWeave would fail to meet its obligations.” CoreWeave has at least $35 billion of debt on its balance sheet and is expected to use $80 billion in cash to finance its operations over the next three years.
As one analyst told the Wall Street Journal, the market reaction reflected growing concerns that CoreWeave had a higher default risk and that its “enormous borrowing requirements could overwhelm its ability to generate cash and refinance debt at reasonable rates.”
But CoreWeave’s travails are only a very sharp expression of developments in the market as a whole.
A report in the Financial Times (FT), based on an analysis by Morgan Stanley Research, noted that “more than $9 billion worth of CDS [credit default swaps] linked to bonds issued by Oracle, SpaceX, Alphabet, Amazon, Meta, Broadcom and Nvidia was traded on average each week in September, up from less than $2 billion in July.”
It is significant that the rise in CDS trading is now involving major AI firms.
The FT has commented that investor confidence in AI companies has taken a significant hit because markets are “spooked” by concerns that revenues from the technology will not match expectations.
In last week’s upheaval the giant chip-making company Nvidia, the largest corporation in the world by market value, had $350 billion wiped off the value of its shares. The Philadelphia Semiconductor Sector fell 11 percent in the week, and the tech-heavy Nasdaq 100 index was down 5.5 percent.
There has been something of an upturn this week amid the issuing of reassuring statements that the boom will continue. But this is not an indication of any return to stability. In fact, as the head of Broadcom has commented, AI investors have “no clue what they are investing in” and the market had become a “casino.”
CoreWeave is not the only firm with debt problems. Oracle has $160 billion in long-term debt and is now the fifth largest borrower in the US corporate bond market. Its credit rating was downgraded by S&P in July to just one notch above junk bond status amid indications that further downgrades could be made.
The Japanese SoftBank Group is another with massive debts—some $164 billion as of the middle of this year driven by major investments in AI including a commitment of $64.6 billion to back OpenAI. Its bonds are already rated as junk and the cost of insuring against default is rising rapidly.
In a report last week, the New York Times commented: “The broader question is whether companies like CoreWeave, Oracle and SoftBank are the industry’s weak links or are harbingers of deeper problems in the AI buildout.”
There are clear indications that the latter is the case. Summing up the experience of the week to the Times, David Munves, a credit strategy researcher, said there was a “collective realization” that the risks of the AI surge were greater than had been anticipated.
“The market’s waking up to the fact that there is potentially a mismatch between the money they have to commit today and borrow today and the ability to generate revenues down the road.”
A recent report by Bain and Company, a global management consulting firm, is one of a number which have been providing some hard numbers on this gap.
According to the Bain analysis, to sustain the present level of investment would require “an AI market approaching $6 trillion annually.”
Some of that revenue, it said, was coming into focus. Consumer AI products, subscriptions and advertising could generate up to $400 billion by 2031. The adoption of AI by companies could bring in as much as $1.4 trillion.
But this would leave “about $4.2 trillion of new revenue to reach the $6 trillion market that we estimate will be needed to finance the buildout.”
Commenting on the scale of the market for AI needed to finance the buildout, Katie Martin, a columnist with the FT, commented: “Really? An industry worth one-and-a-half times the UK’s entire annual economic output every year. This is struggling to pass the smell test.”
Apart from the all-important question of where the revenue for this technological development is to come from, the AI buildout, which has become increasingly significant for US and global economic growth and stock markets around the world starting with Wall Street, is running into the more “normal” problems of capitalist development.
The rise in interest rates is one of these. The yield on the 10-year US Treasury bond, the benchmark for global markets, is now at more than 5 percent. This is not a particularly high rate by historical standards but for a period of 15 years following the global financial crisis of 2008 financial markets became increasingly addicted to ultra-low rates close to zero.
Another constriction is the delay in securing approvals for the building of AI data centers amid rising opposition from communities to their construction on various grounds including the increased cost of electricity which they will entail.
It has been estimated that this year at least 47 large projects have been delayed, withdrawn or blocked, that only 35 percent of planned US data centers are credible and that more than 1,000 communities have tried to introduce moratoria or restrictions on AI infrastructure, according to research carried out by the FT.
Far from the development of AI being some kind of smooth transition as depicted in the various prospectuses issued by its corporate proponents, it is becoming increasingly clear that the vast development of the productive forces it embodies is running headlong into all the irrationalities and contradictions of the capitalist system.
In its report on AI issued at the end of last month, the Bank of England Financial Policy Committee warned of what it called the “risk of a sharper correction” in financial markets, despite their resilience so far, and the potential for a crisis going beyond the AI sector.
If AI-driven productivity gains failed to increase productivity as much as had been forecast this would “affect not only AI-related asset valuation but also sovereign debt markets.”
The turmoil of last week points in that direction.