$350 billion…
That’s how much money big tech will spend building artificial intelligence (AI) data centers this year alone.
That’s more than Switzerland spends running its entire government.
Of course, this propelled AI winners like Nvidia (NVDA), Broadcom (AVGO), and Vertiv Holdings (VRT) to record highs. We give thanks to the stock market gods for the profits they’ve gifted us.
When most investors see how much money big tech is spending on AI, they scream, “Bubble!”
They couldn’t be more wrong.
We were early to Nvidia, first recommending it in 2018. But I think we’re still in the early stages of the AI buildout…
It’s bigger than the entire Apollo program or the Interstate Highway System, in terms of 2025 dollars.
I often get asked, “Isn’t this 1999 fiber again?”
No.
Back then, the companies doing the spending were cash-burning startups. As a group, they were spending 3X more laying fiber optic cables than they were making in revenue. And just 2% of the fiber was being used.
Today, the world’s biggest AI spenders are also the world’s most profitable companies. And data center capacity is sold out months, even years, before it’s built. Different animal.
The other reason it’s not 1999 all over again is that a new AI spender has entered the chat. One with far deeper pockets even than big tech, and the ability to print money when needed.
Those were the words governments around the world whispered to themselves after missing the internet boom.
Take a look at this chart showing the number of $10 billion companies created in the US (left) and Europe (right) in the past 50 years:
Source: Andrew McAfee, MIT
The gap is enormous.
Most countries watched trillions of dollars in market value concentrate into Silicon Valley. Ever since, they’ve lived under a kind of digital dependency by paying American firms for cloud storage, operating systems, and the data infrastructure that runs their economies.
This time, countries are determined not to repeat that mistake. AI is too transformative and too tightly linked to national security to leave in someone else’s hands.
Imagine running America’s nuclear fleet on Chinese AI. Or letting another nation’s algorithms decide how your kids are educated. Exactly.
This is why countries are pouring billions into what Nvidia CEO Jenson Huang termed “sovereign AI.”
Sovereign AI is a country’s ability to build and run AI on its own compute, using its own data, with its own people. Basically, data centers in your own border.
Europe is standing up sovereign clouds and national supercomputers.
The UK recently flipped on Isambard-AI, its most powerful supercomputer ever.
India approved a 38,000-GPU national cluster to train models.
The UAE and Saudi Arabia are on track to spend hundreds of billions of dollars building data centers over the next five years.
As many of you know, I recently moved to Abu Dhabi. Every citizen here got a lifetime ChatGPT subscription. I met one of the guys helping OpenAI put up data centers here. Just so happens he’s Irish!
The same data center ingredients big tech are cooking with: chips, cooling, fiber, and networking.
Sovereign AI is a cool name. But really, it’s a giant wealth transfer from governments to mostly US AI companies.
Governments will build their own data centers and control their own data. But they’ll buy Nvidia chips and Arista Networks (ANET) switches.
|
I remember my friend Louis Gave—who runs the billion-dollar investing firm Gavekal, in China—saying, “A winning strategy has been to buy whatever Xi Jinping invests in.”
The same logic applies here. Politicians are telling us exactly where they’re going to pour trillions of dollars. Betting alongside them is one of the easiest trades in markets today.
As disruption investors, it’s our job to follow the money.
Imagine your government calls you up in the morning and tasks you with building fleets of data centers.
The first thing you’d do isn’t call Jensen Huang asking for a truck full of AI chips. You’d start with the hard stuff.
Buildings…
Backup power systems that turn grid electricity into usable current…
And rows of cooling pipes to suck heat out of racks that will soon be running hot enough to melt.
Only after the concrete had been poured, the power yard was humming, and the cooling loops had circulated would you call for GPUs.
That’s what “Sovereign AI” looks like in the real world. And that’s where the trillion-dollar AI boom begins.
AI will reshape industries and economies for the next 10–20 years. And I think it could be far more disruptive than the internet because it will touch so much more of our lives.
The big risk is selling AI stocks too early. Disruption rarely plays out in tidy bursts. It runs longer than most investors can imagine.
In our Disruption Investor advisory, we’re already profiting from sovereign AI via our custom-made Phase 2 AI ETF. You can access our full portfolio by joining Disruption Investor today.
Stephen McBride
Chief Analyst, RiskHedge