Listen to the audio version of this article (generated by AI).
Wall Street punishes Meta… the question for the rest of your portfolio… three ways a boom ends… “three booms wearing the same trench coat”… the signposts that would make it worse
Last night, Meta Platforms (META) reported solid third-quarter earnings – and got punished for it.
Revenue jumped 26% to $64.6 billion, beating estimates. Excluding a one-time $9.9 billion legal charge, earnings came in at $7.40 per share, well above the $6.54 consensus.
Wall Street didn’t care.
As I write on Thursday, Meta is down more than 6%, the Nasdaq 100 is off 2.5%, and many AI darlings are down double digits. To be fair, Meta isn’t the only issue. Adding to the gloom are Tesla (TSLA), which also reported last night and cut its spending outlook, as well as Lam Research (LRCX), which issued disappointing guidance.
Returning to Meta, the problem isn’t what it’s earning – it’s how it’s paying for AI.
In Q3, Meta’s capex ate up more than all the cash its business generated, for the first time since 2012. Free cash flow turned negative, to –$2.1 billion. And Meta now carries more debt ($83.7 billion) than cash and securities ($82.4 billion).
That’s on top of $55 billion in bonds sold over the past year, plus roughly $40 billion in off-balance-sheet data center financing. And last night, Meta created a new class of non-voting stock that it says will be available for “capital raising.”
Bond investors noticed. Bloomberg reports that Meta’s credit risk hit a record high today on fears that tech giants will issue more debt to fund the AI buildout.
As our technology expert Luke Lango, editor of Innovation Investor, reminded us in yesterday’s Digest, “Bond markets tend to sniff out trouble before stock markets do.”
To be clear, Meta is not CoreWeave (CRWV) – the credit risk example we cited yesterday. It’s still enormously profitable. But when one of the world’s richest companies turns to borrowed money and new stock to fund its AI ambitions, it raises today’s question…
If AI busts, does it take the rest of your portfolio with it?
Yesterday’s Digest tackled two questions…
Luke’s Peak Watch Checklist addressed when the AI boom might turn. And our global macro expert Eric Fry’s “Separation” framework addressed who gets hurt (the AI Builders) and who survives (the AI Appliers).
But there’s a third question that matters to every investor…
How bad could an AI bust get?
Would it be just a tech-sector problem? Or might it drag down the whole market, and the non-AI part of your portfolio, along with it?
History points toward both outcomes, sometimes within the same bust.
When the Nasdaq peaked in March 2000, plenty of investors did exactly what you’d expect. They rotated out of tech and into unglamorous “old economy” stocks.
For a while, it worked beautifully. In 2000, the Nasdaq lost 39%, its worst year since its creation in 1971. Meanwhile, the Russell 1000 Value index, a basket of large, unexciting companies, gained 7%. Small-cap value stocks jumped nearly 23%.
That was a tech bust. It was brutal for tech investors, but it was contained.
But then it spread…
The damage leaked into the broader economy, helped along by a mountain of telecom debt, and the U.S. slid into recession in 2001. At that point, there was nowhere to hide.
Consider The Home Depot (HD) – the opposite of an internet company, selling lumber, paint, and power drills.
On the last trading day of 1999, it closed at a record $68.75. By the end of 2002, it had fallen to roughly $21, making it the Dow’s worst performer that year with a 53% drop. Home Depot’s own stumbles played a role, but the recession hammered its results. The stock didn’t revisit its 1999 high for nearly 13 years.
So, the question that matters for your non-AI holdings isn’t simply whether the AI boom ends. It’s whether an AI bust stays contained, as in 2000, or spreads, as in 2002.
The answer depends on what kind of bust it is.
To help us navigate this, let’s turn to Tom Yeung, Eric’s lead analyst at The Speculator. In his Tuesday update, Tom dissected the anatomy of a tech bust and where AI fits.
Not all busts are created equal
Tom’s key insight is that busts come in three distinct flavors, which he ranked from least to most destructive.
The first is the platform bust. A dominant company funds its expansion with its own cash and thrives, while speculation and carnage occur on the periphery.
Think Apple (AAPL) during the smartphone revolution. It became the most valuable company on Earth, while some of its chip suppliers lost half their value and one glass supplier went bankrupt.
The second is the industry-wide bust. Money floods a sector, and the losses spread throughout it, but the broader economy is spared.
Here’s Tom to illustrate:
Automobiles are the classic example. The number of U.S. automakers shrank from roughly 272 in 1909 to nine by 1941, wiping out large swaths of investor wealth without triggering a broader economic collapse.
Tom also points toward busts in airlines (1920s and 1980s), video games (early 1980s), PCs (mid-1980s), solar (2008-2013), and shale (2014-2016).
The third is the economy-wide bust – the Home Depot scenario. Here, the buildout is financed with too much borrowed money, and the losses spread far beyond the technology itself.
Tom highlights a few examples: the railroad panics of 1873 and 1893 triggered bank failures and some of the longest recessions in U.S. history… leveraged speculation in radio and utilities helped trigger the Great Depression… and the debt-fueled telecom buildout of the late 1990s played a similar role in the 2001 recession.
Notice the common thread – debt. It turns a sector’s problem into everyone’s problem.
“Three booms wearing the same trench coat”
So, which flavor of “bust” might AI bring?
Answering that one is difficult. As Tom explains:
AI isn’t a single boom. Instead, it’s all three booms wearing the same trench coat.
At the top are cash-rich mega-caps that still fund most of their AI spending from operating cash flow. They look a lot like Apple in the 2010s.
In the middle are companies in crowded, competitive niches spending other people’s money: frontier AI labs, AI chip startups, and humanoid robot makers. Tom notes that, like the early automakers, not all of them can survive.
At the bottom are companies like CoreWeave, built almost entirely on borrowed money. Here’s Tom:
These are the most dangerous types of AI bets because they’re relying on enormous amounts of borrowed capital to build capacity no one knows will be needed.
This framing also explains why the AI bubble debate produces such contradictory answers from smart people. Each side is usually describing a different part of the trench coat.
Tom’s base case, and the signposts that would change it
For now, Tom believes we’re set up for the mildest version:
If the AI trade were to sour today, we think the selloff would most closely resemble the “platform bust” of the smartphone revolution.
The most speculative companies would take the damage, while the broader economy escaped largely unscathed.
His reasoning is that today’s market is anchored by the Magnificent Seven and other well-funded tech giants. Together, they account for nearly half of the S&P 500, creating a cash-rich buffer.
Of course, the “mild” version could be anything but mild for many AI stocks. As we highlighted in yesterday’s Digest, Eric’s view is that some AI Builders could fall 50% or more. Today alone, I’m looking at many AI Builders down double digits.
Tom believes the more speculative trades could lose everything:
We will likely see dozens of frontier AI labs and advanced robotics firms shutter, and perhaps some highly leveraged “neocloud” companies as well.
But what might cause the damage to spread – changing a platform bust into something worse?
Tom flags two signposts. The first is the speculative middle tier going public:
Every IPO like this moves more of the eventual losses from venture funds to ordinary shareholders, which is how the PC and solar booms turned into industry-wide busts.
That’s already underway…
SpaceX (SPCX) went public in June in a record IPO. OpenAI has confidentially filed for its own. And Anthropic is reportedly targeting a listing as soon as mid-November at a potential valuation near $2 trillion, which would rank among the largest IPOs ever.
If this trend continues, Tom says it would be wise to trim exposure across the whole AI sector.
The second signpost – in this case, pointing toward an “economy-wide bust” – is the big one: the Mag 7 being forced to borrow to keep the buildout going.
To be clear, big, profitable companies borrow all the time, often simply because debt is cheap. Tom’s warning is about the moment borrowing stops being a choice and becomes a necessity, when AI spending outruns the cash these companies generate.
This isn’t as far-fetched as you might think. A Goldman Sachs analysis projects that annual spending on AI infrastructure will reach $1.6 trillion by 2031. By Tom’s math, that’s six times the cash the Mag 7 companies currently hold.
Here’s Tom on where that could lead:
If this keeps going for long enough, even the best Mag 7 firms will end up looking like fiber optic firms in the 1990s.
They will be forced to borrow enormous sums of cash, creating that perfect environment for a new credit crisis.
Coming full circle
Which brings us back to Meta.
Last night, the hyperscaler stepped into the early stage of Tom’s second signpost. Its capex now exceeds the cash its business generates. It holds more debt than cash. And it just created a new class of stock it can sell to raise money.
One company’s cash-burning quarter isn’t a credit crisis. But it’s the direction worth watching as the rest of the Mag 7 report in the coming days.
But for the moment, Tom’s bottom line is reassuring for the broader market, if not for your AI stocks:
For now, the stage is set for a comparatively mild AI “platform bust,” and that’s the scenario your portfolio should be built for.
Building for that scenario means avoiding the leveraged Builders at the bottom of Tom’s trench coat and owning the businesses positioned to benefit from cheaper AI. That’s the heart of Eric’s Separation thesis that we covered in yesterday’s Digest.
For help on what this means in practice, Eric’s Apogee system recently flagged four AI Appliers: overlooked companies that are using increasingly cheap AI to improve their businesses.
And in his new presentation, Eric reveals his No. 1 pick to buy ahead of November 4, the day after the midterms, when he expects the Separation to accelerate. He’s sharing the name and ticker for free, with no signup required.
Click here to watch Eric’s presentation and get the free name.
We’ll keep watching these signposts, and what they tell us about which kind of bust could be headed our way – or already here.
Have a good evening,
Jeff Remsburg
(Disclosure: I own HD, AAPL)
