Wigglesworth says, "Beware of AI debt!"

Investment Analysis Clubs Macro Economic Trends and Risks
1

Sorry, I know it’s immature but I just can’t help thinking the name “Wigglesworth” belongs in a children’s book instead of a sober NYT article. :wink:

https://www.nytimes.com/2026/08/26/opinion/ai-debt-economy-hyperscalers.html

This Is How the A.I. Debt Binge Sinks the Economy, by Robin Wigglesworth, Aug. 26, 2026

Despite all the hoopla about Nvidia becoming the first company valued at over $5 trillion, the listing of Elon Musk’s SpaceX and the looming initial public offerings of Anthropic and OpenAI, artificial intelligence is no longer primarily a stock market story. It is a story of debt, and that should worry us. Unfortunately, manias fueled mainly by credit almost inevitably end badly…

But we do have a lot of experience with debt-fueled investment sprees, such as the railway euphoria of the 19th century, the telecommunications revolution of the 20th century and more recently the 21st century housing boom. They all resulted in nasty economic downturns, even if the underlying technology proved transformative…

Morgan Stanley… expects the overall debt splurge to hit nearly $600 billion for 2026. That is more than the combined 2026 budgets for the Departments of Justice, Transportation and Education.

And that is just the start, judging by what the hyperscalers have told investors. Their future capital expenditure plans amount to roughly 3 percent of U.S. G.D.P. a year, in 2027 to 2029, according to calculations by Torsten Slok, the chief economist of Apollo, an investment firm. Back-of-the-envelope calculations suggest that’s almost $1 trillion a year. It also would, in relative terms, represent a larger and faster relative ramp-up than even that of U.S. residential housing in the years preceding the 2008 crash…

Goldman Sachs analysts recently tallied $1.5 trillion in aggregate lease commitments by the A.I. hyperscalers… Then there is another estimated $1.5 trillion of “purchase commitments” — promises to buy chips and electricity. Those promises don’t appear on balance sheets either…[end quote]

Stock prices mostly overlook the fact that the hyperscalers’ profits mostly come from within the AI ecosystem and not from outside customers. It also overlooks the transformational competition from open-source and distilled AI that makes expensive frontier AI a rare luxury rather than a foundation of businesses.

Bond holders are interested in being paid their coupons.

The scale is astronomical. The failure of one could pull down the others. Even a whiff of a problem could cause all interest rates to rise.

And the potential for a massive crash is building.

According to Gemini (I’m summarizing this)…

All these bonds need $180 billion per year in net profit—just to keep the debt from compounding, without paying back a single dollar of principal.

Software and cloud providers operate at high gross margins, but after accounting for power, infrastructure depreciation, and corporate overhead, current non-tech enterprise revenue (that is, non-AI customers) total $40B to $60B in net earnings.

We are seeing a massive shift to spend efficiency, driven heavily by Small Language Models (SLMs) and distillation. CIOs are using the expensive LLMs to train the SLMs and then use the SLMs to do 90% of the work.

Without a dramatic surge in novel, high-margin enterprise use cases that strictly require ultra-massive frontier models, the revenue generated by non-tech enterprise adoption will fall well short of covering the interest on the current wave of AI bonds.

Summary Timeline of the AI Squeeze

Timeframe Vulnerability Focus Catalyst for Failure Primary Impact
2026–2027 High startup burn & cash consumption Stalled funding rounds & revenue shortfalls Down-rounds & bankruptcy among pure-play AI labs
2027–2028 Private credit & off-balance-sheet SPVs Debt refinancing wall + hardware depreciation Defaults on data-center real estate & compute leases
2028–2030 Corporate bond market & hyperscaler CapEx ROIC collapse leading to sharp CapEx retrenchment Widespread credit spread widening & tech asset write-downs

This isn’t a sudden, simple crash. It happens in stages. But the amounts are so huge that even unrelated bond yields will be impacted.

Wendy

6 Likes
2

Looks like I’ve got a year or two to sell my NVDA and VRT stock.

DB2

3 Likes
3

Ridiculous TAMs are also fueling this crazy environment. Anthropic is forecasting a TAM of $30 trillion dollars. This will no doubt inflate their upcoming IPO. SpaceX projected a TAM of $28.5 trillion when they filed. Nobody knows exactly how they’re calculating these astronomical numbers. My guess is that they’re pulling them straight outta their wazoo. Maybe they’re just overly optimistic.

"Anthropic is not the first company to propose a larger-than-life TAM. SpaceX recently estimated its TAM at $28.5 trillion. Back in 2019, Uber famously cited a total addressable market (TAM) of $6 trillion by calculating the total mileage value of all personal cars and public transport worldwide.

At the time, some financial analysts and valuation experts criticized this $6 trillion figure as aggressive marketing rather than serious math. Many are equally skeptical of Anthropic’s figure."

“Another way of looking at absurdity of the $30 trillion addressable market claim: annual U.S. GDP is currently $32.5 trillion,” Fred Hickey, tech analyst and editor of The High-Tech Strategist, an investment newsletter, wrote on X. “And yet this nonsense (wild proclamations and predictions) is allowed to continue so that Wall St. & Silly-con-Valley can extract as much money from unwitting ‘investors’ as possible, before the inevitable stock market bubble collapses.”

“Sophisticated investors are going to build their own cash‑flow models,” he said. They will look at Anthropic’s current markets, its contracts and near‑term product roadmap, and then forecast revenue over the next five or so years on that basis. Near‑term revenue targets—such as Anthropic’s reported ambition to reach close to $200 billion in annual sales by the end of the decade—are what serious investors will pay closer attention to, Brunicki said."

I fear there are lots of unsophisticated investors pumping up this jam.

5 Likes