Apollo economic chief Torsten Slok was on the warpath last week, issuing a series of warnings about the AI boom as Wall Street retreated from the sector.
On Tuesday, he highlighted the recent slowdown in token spending growth, saying the compute demand curve is flat and highly sensitive to price changes.
At the same time, most companies investing in technology have yet to see significant productivity gains, Slok added, with AI now looking more like a cost center than a growth engine.
“The main economic flaw in the AI case is that the marginal cost of producing computing exceeds the marginal benefit it brings to users, raising questions about claims of unlimited demand for data centers,” he explained. “The amount of computing and number of data centers needed depends on the returns that businesses are able to generate with that computing. So far, the benefits of AI for businesses have been slow to materialize, implying a growing reluctance to pay for AI services.”
The main determining factor in data center construction models is not physical capacity or power, he said. Rather, it is the actual price that users are willing to pay for computing.
But companies will only pay a premium price if it results in higher revenue growth, bigger profits or substantial savings — and there’s no sign that exists on a large scale, Slok wrote.
“Maybe it will come in the future. Maybe not,” he warned.
On Thursday, he sounded the alarm on the “circularity” of symbolic spending. AI companies have already faced scrutiny for financing deals in which they provide money or guarantee debts to other companies that are also big customers, raising doubts about the sustainability of their growth.
Slok suggested that a similar dynamic exists between AI startups and the venture capital firms that invest in them, saying that 80% of the top 1% of AI spending investors are backed by venture capital. He says this means startups are paying their compute bills with venture capital rather than organic profits.
And on Saturday, he said AI had “brought no financial gain” to U.S. companies outside of large hyperscalers. While profit margins and revenue per worker are increasing for the Magnificent Seven, the rest of the S&P 500 has not seen such improvement.
Echoing previous notes from the week, Slok said Corporate America overall doesn’t need the latest AI model and doesn’t want to pay for it. But even as AI becomes cheaper and commoditized, large companies may struggle to use the technology to improve their bottom lines.
“AI may be a great calculator, but in business, intelligence is all about judgment,” he writes. “Judgment depends on context, and the answer varies depending on the human or model you ask. Perhaps most importantly, AI does not understand accountability, trust, culture, unwritten rules, ethical compromises, or fairness, all of which shape reputation risk and ultimately lead to revenue growth and profit margins.”
His series of notes came as AI-related stocks fell, particularly chip names, with traders doubting how long the orgy of capital spending could last. A handful of hyperscalers plan to spend around $800 billion this year, with investments expected to exceed $1 trillion next year.
But AI companies may find that raising all that capital won’t be as easy. Meta has already issued a lot of debt and this week announced the issuance of a new class of stock, in part to help CEO Mark Zuckerberg pay a possible California wealth tax.
With the social media giant planning to spend between $135 billion and $145 billion this year, the cost of insuring Meta’s debt against default hit a record on Thursday. Five-year credit default swaps widened 7.6 basis points to an all-time high of 114.79 basis points.
At the same time, smaller AI companies like neoclouds are under pressure. For example, CoreWeave stock has crashed more than 30% this month and its five-year credit default swaps have soared as much as 1,145 basis points.
The worries don’t end there. Mark Habib, S&P’s global head of digital infrastructure, pointed to “interconnected financing structures” between hyperscalers and data center projects or smaller AI companies.
“It’s safe to assume that there is more (debt) than we see,” he said at a roundtable in London, according to Reuters.
“If it’s not reported, it’s not part of the disclosure by hyperscalers. Additionally, if we don’t evaluate transactions where this liability may accrue, then the fact is we may not have visibility.”
Gn bussni
