War Drums and AI's 'Scarcity of Capital'

Bombs after midterms?... Norway's $30 billion bond hit... Meta goes negative... Jensen Huang's 'scarcity of capital'... A pre-IPO look at Anthropic... Join Whitney Tilson and Jeff Brown on October 28...

The war drums get louder...

The U.S. has ordered nonemergency government staff and their families out of Saudi Arabia as of late yesterday. Iran-backed Houthi militants keep hitting the kingdom, most recently near Saudi Aramco's Shaybah oilfield. Falling debris there killed two workers.

And that's the calm part of the news...

Multiple reports, citing unnamed sources, say U.S. and Israeli military officials have drawn up specific plans for renewed strikes on Iran after next month's U.S. midterm elections. The attacks would target Iran's energy infrastructure, nuclear facilities, and missile and drone stockpiles.

Signs point to it. President Donald Trump told reporters last night that something will happen "fairly soon." And Defense Secretary Pete Hegseth met with Israeli Prime Minister Benjamin Netanyahu this week in an unannounced visit to Israel.

Let us put two and two together...

The midterms are less than two weeks away... the window for a "deal" with Iran appears to be long closed... and the market is starting to reflect it again. Brent crude, the international benchmark, traded today near $113 a barrel. A month ago, it was around $100.

Meanwhile, money keeps getting more expensive...

The effects of the war, as we've been telling you for months and more so recently, have already been showing up in things like inflation numbers... banks' balance sheets... and riots in Europe.

Here's another (big) example of the influence...

This morning, Norway's sovereign wealth fund – the world's largest, at about $2 trillion – reported a roughly $30 billion loss in the third quarter. The fund's stocks were roughly flat, but its fixed-income holdings lost 3.5% due to rising interest rates.

Not coincidentally, the folks who run the fund also said last month that they want to cut back on owning U.S. Treasurys.

Today, the 10-year Treasury yield actually slipped to around 5%, a one-month low. Higher yields have begun to attract buyers. But that's still 5%, a level not seen often in about 20 years.

Today, though, stocks were the bigger story, stemming from a combination of earnings reports and more AI questions... The tech-heavy Nasdaq Composite Index was down nearly 3%. But in a good sign for market breadth, more than 300 of the S&P 500 Index holdings were up, and the equal-weight version of the benchmark U.S. index rose slightly.

Earnings, Meta's curious move, and the 'capital' problem...

As we wrote on Tuesday and again yesterday, this has been a strong start to earnings season, just as the market expected.

S&P 500 companies are on pace for roughly 29% earnings growth in the third quarter, which would be the third straight quarter above 25%, according to FactSet.

Today brought more of the same... at least in most sectors.

Lockheed Martin (LMT) raised its full-year outlook on the back of a record $292 billion backlog. Railroad operator CSX (CSX) and chemicals maker Dow (DOW) both beat estimates. American Airlines (AAL) posted record revenue, though it forecast rising losses because of fuel costs.

But the tech reports are getting most of the attention and reaction.

Tesla (TSLA) beat estimates last night but slashed its 2026 capital-spending forecast to about $20 billion from more than $25 billion. And CEO Elon Musk lowered his long-term vehicle production target to 5 million a year by decade's end. Shares fell 3% today.

The bigger story is Meta Platforms (META)...

Its revenue grew 26% to $64.6 billion. But Meta spent $34.91 billion on capital expenditures in the quarter, more than the $32.79 billion of cash its operations brought in. That left free cash flow at negative $2.12 billion, versus positive $10.6 billion a year ago.

This might sound familiar...

Alphabet (GOOGL) went free-cash-flow negative last quarter. Now it's Meta.

But, like Alphabet, Meta is not slowing down its spending plans. Meta raised the floor of its 2026 capital-expenditures forecast to between $135 billion and $145 billion. There's already a cost, too... And we see it in Meta's quarterly filing with the U.S. Securities and Exchange Commission ("SEC").

In Meta's SEC filing, it announced plans to issue new nonvoting Class C shares in January. That's effectively a 2-for-1 split that also lets Mark Zuckerberg raise cash without giving up control.

Describing what Meta management told the company's independent financial advisers at Moelis, the filing says (emphasis ours)...

... management had previously presented three forecast cases to the Board during a regularly scheduled Board meeting in May, and that, in each case, these long-term forecasts showed the Company's need to raise hundreds of billions of dollars over a period of several years beginning as early as fall 2026.

As early as fall 2026. As in, now.

Meta folks later called that a theoretical figure, not a financing plan. But the writing is on the wall: The company created three internal forecasts, all saying the same thing... And Meta just created itself a new share class to sell.

Today, Meta shares fell 7%.

This made comments from Nvidia CEO Jensen Huang this morning stand out even more...

CNBC asked Huang about the "neocloud" business model – companies like CoreWeave (CRWV) that rent out Nvidia (NVDA) chips to AI developers. Huang replied that demand is strong, but capital isn't...

There's a scarcity of capital. The demand [for] the capital is much greater than the availability of the capital.

Umm, excuse me?

The point is something my colleague Alan Gula talked about in Monday's issue when he discussed CoreWeave. "It's central to the AI boom and is a forerunner in GPU-backed financing," Alan wrote.

Huang said that Nvidia wants lenders to treat its chips as collateral – a "financeable asset." In other words, the chipmaker's customers can't fund everything they want to buy, so they're asking Wall Street to lend against the product.

This setup can work. But given the size of the build-out and boom we're talking about, it raises risks.

An analysis from research firm Gartner last week put a number on it.

According to Gartner, as much as $3.5 trillion of AI-infrastructure spending could be exposed to write-downs by 2030 if AI software doesn't prove profitable enough. The firm warned that more than 80% of "neocloud providers" may not survive as independent businesses by 2030.

Here's what we do know: No matter what anyone says about AI changing everything, somebody has to pay for it. And with the 10-year Treasury yield around 5% and war risks rising, bills aren't getting cheaper.

So, here's our takeaway for your portfolio: In the AI trade right now, align yourself with the cash.

Before you buy or add to any AI-related stock at this point, be picky. Companies selling AI tools – or selling the picks and shovels to customers who can actually pay – are in a stronger spot than those raising new debt or issuing new stock to keep up.

We're talking about businesses that already make plenty of money... and are using AI to make even more.

Think of established software companies sitting on years of proprietary data that AI models can't scrape from the public Internet. The best of them dominate a niche industry and count the biggest names in that industry as customers.

Companies like that don't need to raise hundreds of billions of dollars and build capital-intensive data centers to be part of the AI boom. They can pay for AI upgrades out of the cash they already generate... and AI makes their moats deeper by helping their customers get more out of data nobody else has.

Our Stansberry's Investment Advisory team recommended a stock that fits this description to a T earlier this month: It's the dominant software provider to one of the world's most research-heavy industries, with more than 1,500 clients and operating margins of more than 40%.

Editor Whitney Tilson says shares of this business could triple from today's prices. Existing Investment Advisory subscribers and Stansberry Alliance members can read the full write-up here.

We just got a good (if preliminary) look at another AI heavyweight's numbers...

The Wall Street Journal reported Tuesday on a draft IPO prospectus from Anthropic, the maker of the Claude AI models. Anthropic could go public as soon as next month at a valuation reported near $2 trillion.

The Journal says Anthropic's revenue hit $16.3 billion in the first half of 2026, up from $1.1 billion a year earlier. But the company also had a net loss of around $81 billion for the first half of the year.

The company has also committed $450 billion to mostly "noncancelable" cloud-computing agreements over the next seven to 10 years. That means Anthropic had better start turning a profit sometime soon.

These numbers only tell you so much, though, and they won't be confirmed until Anthropic's formal IPO paperwork is filed and published...

So, in this week's Mosaic Trader, our colleague Josh Baylin took a different approach to evaluating the company – by looking at how people are actually using Anthropic's tools. And he found a shift in momentum that could change the way you look at this business.

Josh's big takeaway: Developers are increasingly treating Anthropic as a platform that other people build on, not just an AI chatbot or a coding tool. They are shifting toward "plugins and add-ons" – including ones built for everyday office, finance, and legal work.

One way Josh measures that is GitHub "stars." GitHub is a development platform where programmers store and share code, and a star is its version of a bookmark or a "like" – a way to track, save, and show appreciation for a project. Here's what Josh told subscribers...

We track the same data for the other big names in AI. Over the past four weeks, here's how new GitHub stars split among six of them...

Anthropic drew about 35% of new stars over the past month. It saw more developer momentum during that period than Microsoft (MSFT) did, even though Microsoft has about 5 times as many total GitHub stars as Anthropic does.

In other words, more people are paying attention to the smaller player than the established giant.

It's important to note, though, that over the past 12 weeks, Microsoft still led – accounting for 39% of new GitHub stars to Anthropic's 24%. So the past month is a surge, not yet a new normal. But it is a meaningful surge and one we'll monitor closely.

Josh is careful not to overstate it. "Stars" measure attention, not dollars. A developer can like a project and never use it or make money from it. But as he wrote...

It tells us that the people who build software keep choosing Anthropic's tools and that they're increasingly using them for more than code.

Soon, public investors may get their turn to vote Anthropic up or down, like its users.

What to watch...

When Anthropic files its prospectus publicly, Josh says he'll be zeroing in on three things...

• Audited revenue. So far, Anthropic's revenue numbers come from news reports about what it told investors. "Audited" means outside accountants have checked the books. We want to see confirmation that sales are growing fast and profits are holding up.
• Share count. This is how many slices the company is cut into. The more shares there are, the smaller each investor's slice. We want to know how many shares exist today and how many more could be created later, like through stock given to employees.
• Lockup terms. Early investors and employees usually have to wait a set time, often about six months, before they can sell their shares. When that wait ends, a wave of selling can push the stock price down. We want to know when that date will be and how many shares could hit the market.

And one more...

We'll also want to know if the recent shift in developer momentum represents a new floor or just a burst timed to the headlines. Time will give us the answer there. If plugins and industry add-ons keep outgrowing Claude Code itself, that's a sign the platform story is real.

Existing Mosaic Trader subscribers and Stansberry Alliance members can read Josh's full issue here... and check up on his open recommendations, too.

In the meantime, as we sort through war risks and AI questions, there are opportunities...

If you want to put that "follow the cash" idea to work, more new research from Josh is a good place to start. It looks past the tech giants writing the checks to the smaller suppliers who get paid either way.

Josh has been digging into a story he believes most investors are misreading: what Elon Musk is actually building with his Starlink satellite service, Grok AI chatbot, and SpaceX's planned orbital data centers.

Josh calls it the "Starphone" – a device he believes could take on Apple's iPhone and the big wireless carriers at the same time. It would connect through satellites instead of cell towers, with the heavy AI work done up in space.

Musk has publicly denied building a phone. But Josh points to the paper trail... billions of dollars spent on wireless spectrum, a "Starlink Mobile" trademark, satellite-chip deals with major chipmakers, and fresh approvals from the Federal Communications Commission.

As Josh sees it, nearly every piece is already in place. And he believes investors in the right supplier stocks could make as much as 50 times their money by the time the Starphone launches.

He lays out the whole story, and how he believes you can get positioned before it goes mainstream, in a new presentation. Click here to watch it now. He even names one stock for free in his presentation – one he thinks could double over the next few years.

Whitney Tilson and Jeff Brown converge on the 'AI Super Chip'...

Elsewhere, as we mentioned yesterday, Whitney is sitting down on camera with legendary tech analyst Jeff Brown from our corporate affiliate Brownstone Research to talk about what could be the biggest investment opportunity of 2027...

Their subject is a new AI "super chip" they say is 50 times faster than Nvidia's – and a tiny California company, plus five little-known stocks they call the "Fabulous Five," that they believe could hand investors 10 times their money.

Whitney has expressed caution over AI becoming a "bubble." But he and Jeff agree that there's still money to be made, even if the broad market struggles.

Join Whitney and Jeff on Wednesday, October 28 to get the full story in a free presentation. They'll also share two free recommendations – no small thing, considering that the last time Whitney and Jeff gave away a free pick together, it quadrupled in just over a year.

Click here to reserve your free spot now.

And stay tuned to these pages for more. In this weekend's Masters Series, you'll hear directly from Jeff – who, among other winning calls, recommended Nvidia at a split-adjusted 63 cents before it rose more than 35,000%.

New 52-week highs (as of 10/21/26): Automatic Data Processing (ADP), American Financial (AFG), Alpha Architect 1-3 Month Box Fund (BOXX), Cintas (CTAS), Chevron (CVX), Exelixis (EXEL), FactSet Research Systems (FDS), IQVIA (IQV), Saturn Oil & Gas (SOIL.TO), Travelers (TRV), Visa (V), Veeva Systems (VEEV), Waters (WAT), and Zebra Technologies (ZBRA).

Have a comment or question for us? As always, send your notes to feedback@stansberryresearch.com.

All the best,

Corey McLaughlin
Baltimore, Maryland
October 22, 2026