Cue up The A-Team reference…

Last week, Bank of America’s latest Fund Manager Survey crowned semiconductors the most crowded trade on the planet.

A day before the top chip company, Nvidia $NVDA ( ▼ 1.64% ) reported blowout earnings, no less.

My response? I love it when a plan comes together!

While Wall Street keeps screaming about a crowded chip trade we’re printing profits. With no end in sight.

What’s The Big Skinny? During last week’s episode of The Big Skinny, we revisited the performance of our Top 10 AI Stocks Not Named Nvidia. They’re averaging a 70% gain (and counting). Then we sat down with Dan Newman, founder of The Futurum Group. Forget crowded, he’s convinced the AI trade is just getting started. And he handed us five more AI buys… with upside as high as 78%. [Watch our full conversation here.]

The bears have been parroting this Barron’s headline since the day it dropped on March 19…

Here’s the rub — BofA’s Fund Manager Survey is a sentiment thermometer. It’s not a supply-curve reader.

Pundits poll the same crowd every month, hand them a multiple-choice question, then dress the answer up as analysis. Noise. Recycled noise.

May’s survey only doubled down – 73% of managers re-named long global semis the most crowded trade.

Enter Newman. A supply-curve reader.

His team runs recurring surveys of 800+ enterprise IT decision-makers and tear down the latest accelerators.

They just dropped a 100-page initiation report called “The $10 Trillion Ticket” covering 14 semiconductor and infrastructure names.

When I asked Newman what he made of the “crowded” call right after Nvidia’s blowout, he didn’t hedge:

Translation? The “crowding” everyone’s whining about isn’t accurate.

To the contrary, the entire semiconductor complex is under owned relative to the demand wall barreling toward it.

And he’s right. Consider what’s actually happening underneath the AI hood:

Against this bullish backdrop, what’s Dan’s take?

The BofA Fund Manager Survey isn’t a sell signal. It’s a confession that most of Wall Street is still underweight the biggest infrastructure build in modern history.

This is the chart the bears don’t want you to see or accept.

Read the base case twice. Twenty-seven gigawatts (GW) today. One hundred and ninety GW by 2030. That’s a 7x expansion in five years.

Now do the dollar math…

Newman pegs the buildout at ~$43 billion per GW. Run that across the 190 GW base case and you land on $2.7 trillion in cumulative data center capex.

Mind you, the ultimate insider, Jensen Huang’s number is even bigger at $3 to $4 trillion.

Either way, we’re talking about trillions, plural. And we haven’t even witnessed the first trillion in spending.

Here’s the punchline buried inside the chart – 70% of the buildout is for inference, not training.

Agentic AI is force-feeding tokens through the system at a clip no analyst saw coming. Case in point:  

The denominator (capex) is in every headline. The numerator (token demand) isn’t in any of them.

Newman’s blunt: “Structural shortages are going to last for years.”

Memory, logic, ASICs, optics, CPUs. Every link in the chain. Capacity additions don’t catch demand until 2030 at the earliest — and that assumes perfect execution. Nobody executes perfectly.

So tell me again — where’s the “crowd”?

I’ve been banging the table on the semi/AI trade for the better part of a year – and it keeps printing.

The Top 10 non-Nvidia plays we shared Thursday are up an average of 70%.

The semiconductor sector has been the best-performing sector of the last decade. So don’t even think about selling iShares Semiconductor ETF $SOXX ( ▲ 3.94% ).

And then Dan Newman walked in with a fresh 100-page initiation report with 14 more AI names to buy. So you grab a pen and listen to our full conversation for the full details, including:

_{For entertainment/educational purposes only. Not financial, legal, or tax advice; not a recommendation or solicitation. Terms & Conditions: TheBigSkinny.com.}