I used to think investing in AI meant picking the smartest company in the room.
Find the one with the best model. The flashiest demo. The CEO who talks like they’re one product launch away from rewriting reality. Buy the stock, sit back, and let the future compound.
That worked—briefly. Then I realized something uncomfortable.
The real money isn’t always in the intelligence.
It’s in the infrastructure that makes intelligence possible.
And once that clicked, I stopped looking at AI like a software story and started seeing it for what it really is: an industrial revolution disguised as code.
The Illusion of the “AI Company”
Everyone wants to own the next breakthrough model.
It feels intuitive. Intelligence is the product, right?
But here’s the problem: intelligence is becoming commoditized faster than people want to admit.
Models improve, competitors catch up, open-source alternatives emerge, and suddenly what looked like a moat starts to feel like a temporary lead.
Meanwhile, something else is happening quietly in the background.
Compute demand is exploding.
Not growing. Not trending. Exploding.
Every new model, every iteration, every “slightly better” version requires exponentially more compute. More GPUs. More energy. More cooling. More data movement.
Which means the real bottleneck isn’t ideas.
It’s infrastructure.
The Picks and Shovels Moment (Again)
We’ve seen this movie before.
During the gold rush, the people selling picks and shovels often made more consistent money than the miners chasing gold.
AI is starting to look eerily similar.
Everyone is chasing intelligence. But the companies enabling intelligence—the ones building the compute ecosystem—are quietly collecting tolls on every step of the journey.
You don’t need to win the AI race if you’re supplying the race.
You just need everyone else to keep running.
What “Infrastructure of Intelligence” Actually Means
When I say infrastructure, I don’t mean one thing. I mean an entire layered system that makes AI possible.
And once you start breaking it down, you realize how many different ways there are to play this.
1. Compute Hardware
This is the obvious one.
GPUs, accelerators, specialized chips—these are the engines of AI.
Training a large model isn’t just expensive—it’s absurdly expensive. We’re talking tens or hundreds of millions of dollars in compute for frontier models.
Which means the companies designing and manufacturing these chips are sitting at the center of the ecosystem.
They don’t need to guess which AI model wins.
They just need demand to keep increasing.
And right now, it is.
Aggressively.
2. Data Centers (The New Oil Fields)
Data centers used to be boring.
Now they’re the new oil fields.
Massive facilities packed with compute, optimized for power, cooling, and latency. The physical backbone of the digital world.
AI has turned data centers into strategic assets.
You’re not just renting server space anymore—you’re securing access to intelligence capacity.
And as demand outpaces supply, pricing power starts to shift.
This is where things get interesting from an investing perspective.
Because data center operators aren’t just real estate plays anymore.
They’re infrastructure monopolies in slow motion.
3. Cloud Platforms (The Gatekeepers)
You can have the best chips in the world, but if you can’t access them easily, it doesn’t matter.
That’s where cloud providers come in.
They abstract away the complexity and sell compute as a service.
Simple in theory. Brutal in practice.
Because once you build on a cloud platform, switching becomes painful. Expensive. Time-consuming.
So customers stay.
And the cloud providers quietly become gatekeepers to AI development.
Every model trained, every inference run, every experiment conducted—there’s a meter running somewhere.
And someone is collecting.
4. Networking and Data Movement (The Hidden Bottleneck)
This is the part most people ignore.
Moving data efficiently between compute nodes is just as important as the compute itself.
If your GPUs can’t talk to each other fast enough, your entire system slows down.
So companies specializing in high-speed networking, interconnects, and data transfer are suddenly critical.
Not flashy. Not headline-grabbing.
But absolutely essential.
And in investing, “essential but overlooked” is where some of the best opportunities tend to hide.
5. Energy and Power Infrastructure (The Constraint No One Can Ignore)
Here’s the inconvenient truth: AI runs on electricity.
A lot of it.
Training large models consumes massive amounts of power. Running them at scale consumes even more.
And as AI adoption accelerates, energy becomes a limiting factor.
Not compute. Not algorithms. Energy.
Which means utilities, power generation, and energy infrastructure are now part of the AI story.
That’s a sentence I never thought I’d say five years ago.
But here we are.
Why This Changes How I Invest
Once I started thinking about AI as an infrastructure play, my entire approach shifted.
I stopped trying to predict which model would win.
I stopped obsessing over product demos and benchmark scores.
And I started asking a different question:
Who gets paid no matter who wins?
That’s the core of infrastructure investing.
You’re not betting on outcomes—you’re betting on activity.
As long as AI development continues—and I see no reason why it wouldn’t—the demand for compute infrastructure keeps growing.
And the companies enabling that growth keep collecting.
The Power of Indirect Exposure
There’s something psychologically satisfying about owning the “obvious” winners.
The big names. The companies everyone talks about.
But obvious doesn’t always mean optimal.
Indirect exposure—owning the infrastructure behind the trend—often provides a different risk-reward profile.
Less dependent on any single company’s success.
More tied to the overall direction of the industry.
It’s less exciting.
And often more effective.
The Risk Everyone Ignores
Now, before this turns into a love letter to AI infrastructure, let’s talk about the risks.
Because they’re real.
1. Overbuild
If everyone rushes to build data centers, manufacture chips, and expand capacity, you can end up with oversupply.
Which means pricing pressure.
Which means margins compress.
We’ve seen this in other industries. There’s no reason AI infrastructure is immune.
2. Technological Shifts
What if a new architecture dramatically reduces compute requirements?
What if efficiency improvements outpace demand growth?
What if edge computing changes where and how AI is processed?
Infrastructure bets assume continued demand growth.
If that assumption breaks, the thesis weakens.
3. Regulation and Geopolitics
AI is becoming strategically important.
Which means governments are paying attention.
Export controls. Supply chain restrictions. National security concerns.
These factors can impact companies in ways that have nothing to do with fundamentals.
And that’s a layer of complexity you can’t ignore.
The Part That Actually Excites Me
Despite the risks, this is one of the most compelling investment themes I’ve seen in a long time.
Not because it’s trendy.
Because it’s foundational.
We’re not just building better software.
We’re building the physical and digital infrastructure for a new kind of economy—one where intelligence itself becomes a resource.
And like any resource, it requires extraction, processing, distribution.
That’s infrastructure.
My Framework Going Forward
This is how I think about it now:
- Layer 1: Chipmakers and hardware providers
- Layer 2: Data centers and physical infrastructure
- Layer 3: Cloud platforms and service providers
- Layer 4: Networking and data movement
- Layer 5: Energy and power systems
Each layer has its own dynamics. Its own risks. Its own opportunities.
But they’re all connected.
And the more I study them, the more I realize this isn’t a single trade.
It’s an ecosystem.
The Quiet Advantage of Boring Businesses
There’s a certain irony here.
The most transformative technology of our time is being powered by some of the most “boring” businesses imaginable.
Utilities. Real estate. Networking hardware.
Not exactly the stuff of viral headlines.
But that’s often where the opportunity lies.
Because while everyone is chasing the story, fewer people are studying the structure behind it.
Final Thought: Intelligence Needs a Body
We talk about AI like it’s this abstract, almost mystical force.
Something that exists in the cloud, detached from physical reality.
But that’s not true.
Intelligence needs a body.
It needs chips. Servers. Buildings. Power lines. Cooling systems.
It needs infrastructure.
And that infrastructure is where a significant portion of the value is being created—and captured.
So while everyone else debates which AI company will dominate, I’m increasingly comfortable standing one step back.
Watching the entire system.
And investing in the parts that make the whole thing possible.
Because if there’s one thing I’ve learned, it’s this:
You don’t have to predict the future perfectly.
You just have to understand what the future will require.
And right now, the future requires a lot of compute.
Comments
Post a Comment