The AI Trade’s $81 Billion Contradiction

The AI Trade’s $81 Billion Contradiction

Anthropic’s record-setting debut… the contradiction in the AI trade… Nvidia’s bond warning… will new governors slow data centers?… a date Eric Fry is watching

As I write on Thursday, Anthropic (ANT) is trading on the Nasdaq after pricing last night at $83 per share – above its original $70-$78 range – in a record $81.3 billion IPO.

Shares opened at $111.50 but have pulled back to roughly $95 as I write. Still, that’s about a 15% jump above last night’s pricing.

In Tuesday’s Digest, we wrote that today’s debut would serve as a barometer for AI sentiment. So, what’s the barometer telling us?

With shares trading above the offer price, everyday buyers – not just the institutions and strategic backers who got in at the offer – appear willing to pay up for AI exposure.

But note a major contradiction…

“Paying up for AI exposure” has been the last thing that investors have been willing to do over the last month.

As I write, the Philadelphia Semiconductor Index (SOX) trades at roughly 9,690, down more than 30% from its June 22 record. That’s a roaring bear market. And according to 24/7 Wall St., chip stocks are on track for their worst two-month stretch since 2008.

Or take Nvidia (NVDA), the poster child for the AI trade. Tuesday night’s strong earnings report reminded Wall Street what a genuine economic powerhouse it is – the company generated $115.8 billion in revenue, up 20% sequentially and 103% year over year. And yet, for all that operational dominance, its stock price is down about 37% in less than two months. It’s now trading at its lowest forward price-to-earnings ratio since 2011, well below historical averages of 30x–40x.

And if we look at the tech-heavy Nasdaq itself, it’s down 12% since early October, an official correction.

How can investors be clamoring for ANT while dumping the companies that supply AI’s chips?

Because AI is experiencing a massive reshuffling.

Anthropic is a buyer of computing power – chipmakers are the sellers. And right now, Wall Street is worried that the sellers are facing a major slowdown.

Earlier this week, research firm SemiAnalysis cut its forecasts for AI accelerator spending, with 2027 down 19% to $627 billion and 2028 down 28%.

Why?

Let’s go to our hypergrowth expert, Luke Lango. From his Innovation Investor Daily Notes:

The three main drivers of the AI spending slowdown (stricter policy out of Washington, tighter financial conditions on Wall Street and widening supply gaps that delay projects) are accelerating in real time.

Indeed, AI spending will be much less than expected in 2027, 2028, 2029 and 2030. That means lower AI supply chain earnings and lower AI stock prices, which is exactly what we’re seeing today.

But wait – didn’t Nvidia guide higher on Tuesday night’s earnings call? And didn’t interim CEO Colette Kress raise the company’s growth forecast for the fiscal year ending in January 2028 from 70% to 75%?

She did. Which only deepens our apparent contradiction: heavy demand for Anthropic, heavy selling of AI chip stocks all month.

At the heart of this contradiction is one thing: durability.

Nvidia’s near-term growth and sales aren’t in question. Rather, it’s how long the wider AI buildout – the chips, data centers, and power that all this spending pays for – can keep growing at this pace.

Nvidia’s results (and results for the wider chip ecosystem) are only as durable as their customers’ willingness and ability to keep spending. Luke’s three drivers all threaten that spending, and SemiAnalysis’s steepest cut estimates for 2029 and 2030 reflect the same worry.

And we’re not only seeing this fear in earnings forecasts – it’s also showing up in the bond market.

The red flags from the bond market

A stockholder can sell in a second if the story changes. A bondholder who lends to a company for 30 years is counting on its earnings still being there decades from now.

Nvidia’s June bond sale included a piece that doesn’t come due until 2056. That makes the bond market a useful read on how much confidence investors have in the AI buildout’s long-term staying power.

And according to our global macro expert Eric Fry, that market has been flashing a warning.

On Monday, Eric looked at long-term bonds from two of the AI boom’s strongest borrowers: Nvidia, which sells the chips, and Meta Platforms (META), which buys them.

Moody’s rates Nvidia one notch below the top tier, yet since late June both companies’ long-duration bond prices have fallen roughly 14.5%, and the extra yield they pay over Treasurys has widened. Nvidia’s has roughly doubled.

Eric looks at this and comes to one conclusion – the bond market is fearful of the cost of the AI buildout.

But a price tag only matters relative to what it buys. The “cost” is the bill; earnings are how the bill gets paid. Bondholders are really asking whether those earnings will keep pace with the spending. And that brings us full circle to the fear driving this market…

That AI’s earnings growth has peaked.

Now, Eric doesn’t think this means the AI megatrend is ending – he sees AI adoption accelerating. But he believes something fundamental changed in the AI economy this year, a development he calls “The Crossing.” He’s urging investors to recognize that the companies that created so much wealth in AI’s first chapter may not deliver the biggest gains in the next.

So, where does he see those gains coming from?

The companies that effectively put AI technology to work in their respective sectors.

We’ll bring you more details over the coming days, but go ahead and mark your calendar for Monday, Nov. 23, at 10 a.m. Eastern – that’s when Eric will dig into this in his free Life After Nvidia presentation.

He’ll explain what he’s seeing today, what it means for yesterday’s AI winners, and where the greatest opportunities lie ahead. You can reserve your spot right here.

The political risk behind the AI spending debate

If the market is bracing for cuts to AI spending, one question may matter more than any earnings report…

Will the new political majority actually slow data center construction as is widely anticipated?

The case that it will is stronger than many investors realize…

Democrats won full control of Congress in this month’s midterms. They also flipped governorships in Texas, Ohio, Georgia, Nevada and Iowa – all data center hot spots – often on campaign promises to rein in the industry.

Meanwhile, Data Center Watch found that 32 of 34 winning governor candidates backed stronger regulation or restrictions, and 12 of 34 backed bans or moratoriums.

Voters appear to agree. About three-quarters oppose a new data center in their own community, per the AP’s VoterPoll.

Texas is the test case – and it’s racing the clock

Texas Governor-elect Gina Hinojosa, who defeated Gov. Greg Abbott this month, has promised “a moratorium – a ban – on all data center approvals” when she takes office Jan. 19.

But according to Barron’s, Texas regulators aren’t waiting. Abbott paused approvals in August pending a grid audit, but the state has since resumed permitting.

In just the past few days, it approved permits for 2.9 gigawatts of power plants serving data centers – about as much power as three large nuclear reactors generate. Energy research firm Cleanview projects that 6 to 10 more gigawatts could be approved before Hinojosa is sworn in, on top of 24 gigawatts already permitted.

In other words, the industry is racing to lock in projects before the rules can change. That cuts two ways…

A big pipeline is already permitted, which cushions the near term. But it also shows how seriously the people closest to the money are taking the risk, and it pushes the real uncertainty out to projects that haven’t been approved yet.

But there’s also the argument against a data center moratorium

A U.S. Chamber of Commerce analysis – from a business group, to be fair – estimates that one mid-sized data center generates nearly $700 million in state GDP and supports more than 5,200 construction jobs a year.

Given these dollars, even some Democrats are weighing the economics…

For example, in April, Maine Gov. Janet Mills vetoed what would have been the nation’s first statewide data center moratorium. She said a pause was warranted, but the bill made no exception for a $550 million project in the town of Jay, which she said would bring more than 800 construction jobs and substantial property tax revenue. She said she would have signed it with that exception.

It’ll be interesting to watch what these democratic legislators do when the choice is between following through on campaign promises and watching potential data center jobs and tax revenues disappear.

So, how much of the risk here is already priced into AI stocks?

Returning to Luke, he presented his readers with a chart from Bank of America’s Global Fund Manager Survey showing how quickly sentiment has flipped.

He noted that when asked whether an AI hyperscaler will announce a capex cut in the next six months, only about 14% of fund managers said yes in September. That rose to about 30% in October and 52% in November, while the share saying no fell from about 80% to 40%.

Here’s Luke:

That’s a big shift in a short time, and it cuts both ways.

In the near term, it means the market is bracing for cuts, and estimates are likely to follow. But it also means the cut is increasingly priced in. When a fear becomes the majority view, it has less power to surprise.

That’s one reason we expect the market to eventually look past the cuts, even if the bottom is still months away.

Luke goes on to say that if the hyperscalers’ 2027 guidance in late January comes in better than this survey implies, that’s the kind of upside surprise that could flip sentiment fast.

Altogether, this points to three calendar dates to circle

The first is the closest, and brings us back to Eric: Monday, Nov. 30.

He believes this day could bring an important clue about the next phase of AI. Why? He’ll explain in his free Life After Nvidia presentation on Monday, Nov. 23, at 10 a.m. Eastern, but it relates to our wider conversation.

The second is Jan. 19, when Hinojosa takes office in Texas and we start learning how far the campaign promises really go.

The third is late January into early February, when the hyperscalers report quarterly earnings and update their 2027 spending plans.

In the meantime, ANT’s trading will be our day-to-day barometer. We’ll keep you updated on that – and on all three dates.

Have a good evening,

Jeff Remsburg


Article printed from InvestorPlace Media, https://investorplace.com/2026/11/the-ai-trades-81-billion-contradiction/.

©2026 InvestorPlace Media, LLC