Why NVIDIA’s Biggest Bottleneck Isn’t Chips

My dad worked construction most of his life. Came home most nights with dust in his hair, cuts on his hands, and stories about the latest project he was working on.

One evening, when I was maybe twelve, he held up a piece of scrap copper pipe.

"You know what this is?"

"A pipe," I said, barely looking up from my homework.

"This is civilization." He tossed it to me. "Every light that turns on, every motor that runs, every signal that travels, it all moves through this. They can invent all the fancy computers they want. But without this nothing works."

I didn't understand then.

I do now.

A $500 Billion Trade Deal

Last week, two announcements came in within hours of each other. Most probably filed them under "tech news" and moved on.

Taiwan Semiconductor Manufacturing Company, the company that physically makes the chips Nvidia designs, powering every AI system on Earth, announced they're increasing capital expenditure to $54 billion for 2026.

That's a 40% jump from last year.

Their CEO, C.C. Wei, spent four months personally meeting with customers and their customers' customers before making that call.

His conclusion? "AI is real. Not only real, it is starting to grow into our daily life."

Prior to their announcement, the U.S. and Taiwan signed a $500 billion trade deal specifically designed to bring 40% of Taiwan's semiconductor supply chain to American soil.

Commerce Secretary Howard Lutnick made the terms clear: companies that build here get tariff-free imports. Companies that don't? 100% tariffs.

Wall Street and the media called these AI or tech stories.

They missed what's actually happening.

Every single one of those billions has to become something physical. Data centers. Semiconductor fabs. Power infrastructure. Transmission lines. Cooling systems.

And every single one of those things requires copper.

The 18-Year Time Bomb

Robert Friedland built Ivanhoe Mines into one of the most successful mining ventures in history. The man has forgotten more about commodities than most analysts will ever learn.

Last October, speaking at USC's Energy Business Summit, he said something that still echoes in my head:

Let that sink in.

Ten thousand years. Every copper wire in every building ever constructed. Every electrical system in every vehicle ever manufactured. Every transmission line stretched across every continent.

Every piece of electronics your grandparents, your parents, and you have ever touched.

We need to match all of it. In 18 years.

And that's before accounting for the AI buildout, the data centers, the electrification of transportation, and the grid upgrades required to power it all.

Friedland wasn't done: "You people have no idea whatsoever what we're facing. You're dreaming."

I've spent enough time around mining executives to know when someone's talking their book and when someone's genuinely alarmed.

Friedland was alarmed.

The Copper Shortage

S&P Global just published a study called "Copper in the Age of AI: Challenges of Electrification."

Their conclusion: primary copper supply will peak in 2030, then decline. By 2040, we're looking at a 10 million tonne supply gap.

The entire world currently produces about 23 million tonnes of primary copper per year.

We're talking about a shortage equivalent to nearly half of current global production.

Can't we just mine more?

Here's the problem:

• Copper exploration budgets have collapsed by almost 50% in real terms since 2012

• New mines take 15 to 20 years to develop

• The easy copper has already been found

• Average ore grades have dropped from 1.4% in 2010 to 0.65% today

At 1.4% ore grade, producing one tonne of copper requires processing about 71 tonnes of rock. At 0.65% grade, that jumps to 154 tonnes.

That's more than double the material movement, double the energy consumption, double the water usage, all to produce the same amount of copper.

What about substitution?

Aluminum has only 60% of copper's conductivity. In data centers, where space is premium and heat management is critical, S&P Global describes copper as "non-negotiable."

You can't substitute your way out of a structural shortage when the alternative quite literally can't do the job.

And what about recycling?

We're already recycling everything we can. Only 4 million tonnes of the 30 million tonnes consumed annually comes from recycled sources.

The low-hanging fruit has been picked.

I've learned to watch what money does, not what people say.

In February 2024, Friedland predicted copper inventories were "like a powder keg ready to explode."

Since then:

• Copper prices have risen 58%

• The Global X Copper Miners ETF has gained 128%

• Copper miners have decisively broken out against the S&P 500

• The inflation-adjusted copper price has broken through a 15-year trendline

A prominent research team following this story noted that copper miners are outperforming the S&P 500 by over 40 percentage points in just the last three months.

That's the equity market pricing in sustainable increases in operating margins. That's institutional money repositioning before the rest of the world catches up.

If you’re bullish on AI, you have to be bullish on copper

There's no version of the AI future that doesn't run on copper. Every data center, chip fab, and every power line feeding electricity to the systems running these models, and even the cooling system keeping servers from melting… all run on and require copper.

The TSMC announcement is about spending $54 billion on physical infrastructure.

The $500 billion U.S.-Taiwan deal is all about construction and factories and transmission lines and grid upgrades.

The AI revolution is really a copper story, and all the tech CEOs know it.

Meanwhile, exploration budgets are collapsing while hyperscaler capex goes parabolic. One chart shows copper exploration spending falling for over a decade. The other shows data center investment going vertical.

Those two lines are on a collision course.

And the collision is called "shortage."

Here’s How We Play This

Since October 2025, I've held two copper-focused positions in my portfolio. As of this writing, they're up 37% and 54% respectively.

Those specific recommendations are available to Premium Members.

But here's what I can tell you: I'm not done.

The supply constraints are real, and the demand drivers are accelerating. The timeline for new supply is measured in decades, not quarters.

This week, I'm opening a new position in the copper space.

Today is Martin Luther King Day, so I’m sending out the recommendation while the market is closed.

That way, when you read the recommendation below and decide you’d like to get in, you can do it as soon as tomorrow morning when the market opens.

It's based on everything I've laid out here, the AI buildout, the infrastructure requirements, the supply crisis that can't be fixed fast enough.

Premium Members get the details below, including: the specific company, the entry price, the reasoning, and the risk management approach.

My father was right about that copper pipe.

Civilization runs on it.

The difference between now and when he was wiring buildings is that demand is exploding while the supply is constrained. The easy deposits are gone. The timeline to bring new supply online is longer than the timeline of the demand surge.

Every headline about AI, every announcement about data centers, every deal to build chip fabs, they're all copper stories.

The market is starting to figure this out, and the breakouts are happening. The repricing has begun.

I know where I'm placing my bet.

The new copper recommendation drops below for Premium members. If you've been waiting for a reason to join, this is it.

Double D

🔓 Premium Content Begins Here 🔒