An 'AI Vanity' Bonfire

An epic run... Meta's bombshell... The burning has begun... The market shoots the generals... Three reasons to welcome a bear market... Can you afford to ignore my warning?

Is it over?

I (Dan Ferris) don't mean the Iran war, high diesel prices, or the riots in France.

No. The "it" investors care most about is the AI bull market that started on November 30, 2022 – the day ChatGPT launched.

Since then, the S&P 500 Index is up about 87%, and the tech-heavy Nasdaq Composite Index is up more than 125%.

During the same period, AI-chip dominator Nvidia (NVDA) is up more than 1,150%, and data-analytics firm Palantir Technologies (PLTR) is up an astonishing 2,500%. The hyperscalers – Alphabet (GOOG), Microsoft (MSFT), Meta Platforms (META), and Amazon (AMZN) – are up between 100% and 450%.

It has been an epic run. But it's starting to feel like it might be over... or at least that the beginning of the end has arrived.

I say that because the companies that have been spending big on AI infrastructure are starting to disappoint investors.

The latest example is Meta Platforms...

It reported earnings on Wednesday after market close.

The company reported $3.83 in earnings per share for the third quarter, more than triple the $1.05 it earned in the same period last year. But that was driven largely by a drastically smaller income-tax provision.

There's some noise in the operating results. Meta had a $9.9 billion charge related to an August 2026 settlement with several U.S. states over allegations that Meta purposely made Instagram and Facebook addictive to young users.

But the real bombshell was that Meta reported quarterly free cash flow of negative $2.1 billion. That's the first time Meta has burned cash instead of generating it since 2012.

On the earnings call, Meta said the quarter's $34.9 billion in capital spending was "driven by investments in servers, data centers, and network infrastructure."

The market hated the report and pushed shares down 7% yesterday.

Meta is only the latest hyperscaler to report negative free cash flow...

In July, Alphabet reported negative $5.9 billion in free cash flow for the quarter ended June 30, 2026. Unlike Meta, this was the first time reporting negative free cash flow in the company's history. Shares fell around 7% on the news. At yesterday's close, shares were down 8% since October 13.

Amazon reported negative free cash flow for the past two quarters, totaling about $27 billion – though that's more common. Amazon has reported seven such quarters out of the past 20. As of yesterday's close, its shares were down around 2% over the past six months.

No matter how good these companies' results are otherwise, the market is expressing its disapproval of their deteriorating cash flows. Perhaps the bonfire of the "AI vanities" has begun.

If we add Microsoft, Nvidia, and Palantir to this list – none of which generate negative free cash flow – only Palantir is up since the S&P 500 and Nasdaq peaked on October 6. Nvidia is the worst of the bunch, down 12% in that time.

The damage to the hyperscalers is worse than that, though.

Alphabet and Meta are both in bear market territory. Alphabet has traded as much as 20.2% below its May 13 all-time high, and Meta as much as 33.5% below its August 2025 high.

Longtime readers know I don't often bother with recent quarterly results and short-term market moves. But this is the sort of action that happens at the end of a bull market.

The market 'shoots the generals last'...

As a bull market winds down, the market shoots the soldiers in front first – the lower-quality stocks – as investors rotate into higher-quality names as safe havens. But eventually, there's nowhere to hide, and even the big, safe-haven names – the generals – peak and begin declining. When those supposedly safer mega-cap names start dropping, you know the bull is dead.

I last used that old Wall Street adage in my May 20, 2022 Digest, with the S&P 500 trading about 19% off its January 3, 2022 high and the Nasdaq about 29% off its November 19, 2021 peak.

Back then, the market topped out in a classic gradual progression.

First, the market shot the most speculative stocks (special purpose acquisition companies, clean energy, cannabis) after they peaked in early 2021. Then, it shot bitcoin and the Nasdaq once they hit their peaks in November. Eventually, it turned to the generals – the S&P 500 and Dow Jones Industrial Average each took a bullet after peaking in the first few trading days of 2022.

The Federal Reserve helped the gradual topping process by pivoting to a hawkish stance at its June 2021 meeting – a full nine months before starting to hike rates in March 2022. It gave everybody plenty of time to contemplate how higher rates might punish pandemic-era speculative excess.

Looking back, we can see an orderly bull market breakdown that took 10 to 11 months, from February/March of 2021 to January 2022. I inadvertently called the top of the Nasdaq to the day in my November 19, 2021 Digest, when I wrote...

You'll never hear me say that I believe the top of the market is imminent or that it's behind us. As I've pointed out before... market tops are processes, not one-day events. You often can't tell you're at the top until it's far in the rearview mirror.

But I will say that I believe risk in stocks – like valuations – is at an all-time high today...

This is a dangerous time for investors. Be careful. And don't say I didn't warn you.

It was the all-time high for risk and valuations in the Nasdaq. The index didn't reach that day's closing level again until more than two years later. The bear market bottomed in 2022, with the S&P 500 down 25% and the Nasdaq down 36%.

The point is, a market top is a process, not an event. If today's bull run really is over, the topping process is happening faster this time.

The Dow Jones Industrial Average – deemed less relevant in the AI boom – peaked on August 5, and as of yesterday's close, traded about 3% below that level. The Russell 2000 Index – the lowest-quality names – peaked on August 14 and trades about 10% lower. The Nasdaq and S&P 500 both peaked on October 6. The Nasdaq is down 6%, while the S&P 500 is down about 2%.

Source: Bloomberg

A two-month topping process would be eight or nine months shorter than what we saw in 2022. It's as though, this time around, the Fed's hiking cycle has hit the market in the face with a shovel. But maybe that's just because the AI boom is focused on a relatively small group of stocks.

If I had to make the case for a longer topping process, I could say it started on June 20, 2025. That's when CoreWeave (CRWV), the "neocloud" company that's building its own data centers, hit its last all-time high. Yesterday, the stock closed roughly 70% below that.

The stock has been crushed as the market doubts whether it will ever generate free cash flow. CoreWeave has borrowed more than $50 billion to build data centers, networking hardware, and Nvidia chips it rents to customers. You need plenty of cash flow to pay off that kind of debt, especially when a bunch of it costs between 8.5% and 9.75%.

According to data compiled by Bloomberg, the company's numerous junk-rated bonds all trade at discounts to par between 15% and 30%. The bond market (aka the "smart money") deems the company's debts very risky. It knows that if CoreWeave defaults, the stock will be worthless.

In short, the market absolutely hates CoreWeave's fundamentals.

Some of my colleagues have been debating the stock's prospects, but avoiding it is a no-brainer for me. It's burning cash like crazy and there are plenty of profitable companies trading at attractive valuations elsewhere. There's no reason to try to catch a falling knife.

Maybe CoreWeave's 2025 peak means the biggest low-quality name is done, and maybe the Russell 2000's August 2026 peak means the entire lower-quality universe of stocks is done.

Digest readers know I don't make predictions...

But if I had to guess, I'd say the bull has another year or two to run, just because markets mostly go up. So the odds favor being bullish at any given moment over being bearish.

Even so, maybe we'll look back on this moment as the canary in a coal mine. Maybe we'll say it was when the market started to wake up from its blue-sky dreams of massive AI wealth and start demanding to see free cash flow instead of endless spending fueled by borrowing.

Plenty of folks will say I'm an idiot for even contemplating the bull's end right now, given that corporate earnings are soaring.

But I don't know of any current bulls who predicted a diesel-fuel crisis last December, before the Iran war was on anybody's mind. And nobody was bearish on the Nasdaq in November 2021 when I said the risk was at an all-time high and that you'd better be careful. So I'm comfortable being the lone voice crying into the wilderness. It's the trade-off you agree to make if you want to remain aware of growing risks and potential market inflection points.

One last thing...

This might sound a little perverse, but I'll be relieved – and even a bit happy – when the bear market finally arrives. I say that for three reasons...

First, the higher market valuations and speculative excess go, the further the market could fall when it's all over.

Just consider... after 1929, the Dow fell nearly 90% and didn't make a new high for 25 years. Japanese stocks peaked in December 1989... fell around 80%... and didn't make a new high for 35 years.

Stocks were never more expensive than in late 1999. The 2000 to 2002 bear market took the Nasdaq down around 80%. It didn't make a new high until 2015.

A bear market starting today might not be as bad as one that starts at higher valuations with more baked-in excesses a year or more from now.

Second, a top means the market is punishing AI vanity stocks for spending too much money on investments that might not earn a great return – if any. That means the market thinks burning cash to make these investments is a big mistake. It doesn't like all the borrowing these companies are doing to fund potentially poor investments.

That is exactly what markets should do. They should take fundamentals into account. They should assess the likelihood that returns on massive AI capital spending will disappoint.

Punishing the hyperscalers for burning cash makes perfect sense. In fact, the return of making sense might even be enough to help us avoid a massive bust. It could slow the hyperscalers down and reset expectations, perhaps without a massive credit crisis and bear market.

Third, we need an occasional washout to keep stocks priced attractively enough for good long-term compounding. Before the bear markets that began in early 2000 and late 2021/early 2022, stocks hit their most expensive levels in recorded history. Without a bear market, it's unlikely they'd have performed very well since those dates.

Valuations today have surpassed those of late 2021. Again, plenty of folks will tell you this is nonsense and that stocks are cheap if you know how to analyze earnings correctly.

Maybe they're right, but I'm not convinced.

Maybe it's silly to suggest the market suddenly cares about fundamentals and that it will start selling the generals. Maybe I'm wrong. But I'll wait for the market to tell me. And since every boom goes bust eventually, can you really afford to ignore what I'm telling you?

New 52-week highs (as of 10/22/26): Automatic Data Processing (ADP), American Financial (AFG), Amgen (AMGN), American Water Works (AWK), Berkshire Hathaway (BRK-B), Cintas (CTAS), Exelixis (EXEL), FactSet Research Systems (FDS), Hagerty (HGTY), IQVIA (IQV), Coca-Cola (KO), Eli Lilly (LLY), Lonza (LZAGY), Altria (MO), Nasdaq (NDAQ), Pfizer (PFE), Qualys (QLYS), Republic Services (RSG), SharkNinja (SN), State Street SPDR Portfolio S&P 500 Value Fund (SPYV), Travelers (TRV), Visa (V), Veeva Systems (VEEV), Veralto (VLTO), and State Street Health Care Select Sector SPDR Fund (XLV).

In today's mailbag, feedback stemming from yesterday's Digest about AI spending and "scarcity of capital"... Do you have a comment or question? As always, e-mail us at feedback@stansberryresearch.com.

"The article today stated possibly 80% of the 'neoclouds' could go bankrupt. Try 90%. The hyperscalers are so tunnel-visioned at this point they're all sprinting (FOMO) to the edge of a cliff with CapEx. Then there's Anthropic. A $2 trillion value slapped on a company who LOST $81 billion in the first half of the year and has never shown a profit...

"Will AI disappear after the bubble inevitably bursts? Of course not. Just like trains, planes, cars and the Internet didn't. They changed our lives and AI will as well. But they left a lot of bankrupt people and institutions along the way. Is this time different? Not a chance in Hell.

"Keep up the good work!" – Subscriber Jim V.

Corey McLaughlin comment: Thanks for the note, Jim, and the perspective.

Just one clarification: The research we cited yesterday from Gartner didn't say 80% of neoclouds would go bankrupt, per se. Some might. But the firm is projecting 80% of these companies will "cease independent operations" because of cost pressures by 2030 and suggests some could combine with others or be taken over by larger businesses.

But point taken about the tunnel-vision, FOMO-fueled spending trend, which has been going on for years now.

Good investing,

Dan Ferris
Medford, Oregon
October 23, 2026