From Silicon to Minds: So This Is What a Mind Costs
- Rich Washburn
- 5 hours ago
- 7 min read
There was a number on my screen that made me stop.
$1.75 trillion.
That was one of the valuation levels traders on Polymarket were assigning meaningful probability to Anthropic reaching.
One point seven five trillion dollars.
At some point numbers like that almost lose their meaning.
A million is understandable...A billion gets fuzzy...A trillion is basically a word we use when there are too many zeros to visualize anymore. And $1.75 trillion for an artificial intelligence company sounds, on its face, completely insane. But the longer I looked at it, the more I started thinking we may be asking the wrong question.
Maybe the question isn't: How can an AI company possibly be worth $1.75 trillion? Maybe the question is: So this is what a mind costs. Not a human mind.
I'm not making an argument about consciousness, sentience or personhood. I'm talking about something much more practical. A system that can reason.
Write.
Code.
Research.
Analyze.
Plan.
Use tools.
Review its own work.
Correct mistakes.
Maintain context.
And increasingly take an objective and work toward it.
Something mind-like enough to be economically useful. And for the first time in history, we know how to manufacture that.
We Used to Buy the Pieces
For most of the computing age, we bought pieces.
We bought silicon. > Then processors. > Then computers. > Then servers. > Then software. > Then cloud compute. > Then storage. > Then data. > Then models.
Each generation moved us further up the abstraction stack. But most of the time, the machine was still waiting for us to tell it exactly what to do. Excel could calculate the spreadsheet. It couldn't decide why you needed the spreadsheet. Photoshop could manipulate the photograph. It couldn't look at the campaign and decide what image would work better.
A compiler could execute the code. It couldn't sit beside you and say: "I understand what you're trying to build. There's probably a cleaner way to do it."
Traditional software consumes instructions. Increasingly, AI can consume intent.
Tell the machine what you are trying to accomplish and it can begin figuring out some of the steps between here and there. That's not merely faster software. That's another layer entirely.
Silicon became compute.
Compute became software.
Software became intelligence.
And intelligence is becoming agency.
From silicon to minds.
The Trillion-Dollar Valuation Isn't Just the Model
This is also why I think it is misleading to look at something like a $1.75 trillion Anthropic valuation and imagine somebody putting that price tag on a chatbot.
They're not.
The thing being valued is much larger.
Because there is an enormous physical and digital machine underneath that little blinking cursor.
GPUs.
Memory.
Fiber.
Networking.
Storage.
Data centers.
Cooling plants.
Transformers.
Substations.
Power generation.
Software.
Models.
Training.
Inference.
Data.
Context.
Memory.
Tools.
Researchers.
Engineers.
Distribution.
Capital.
And extraordinary quantities of electricity.
All of it comes together so that someone sitting at a desk can type: "Here's what I'm trying to do." And something on the other side can respond: "Okay. Let's figure it out."
The model isn't the whole product.
The infrastructure isn't the whole product. The software isn't the whole product. The electricity isn't the whole product. They're organs.
What is valuable is the capability that emerges when you assemble all of them.
The mind is the product.
We Have Never Really Had an Economic Object Like This
Human intelligence has always been scarce for a very simple reason. It comes attached to humans.
If you need more expertise, you hire another person.
If you need more engineering, you hire engineers.
More research? Researchers.
More analysis? Analysts....More writing? Writers.
And there are very good reasons that employing someone does not mean owning their mind.
People have their own lives.
Their own motives.
Their own judgment.
Their own limits.
They sleep.
They disagree.
They leave.
They can tell you no.
And they absolutely should be able to.
Artificial intelligence creates a completely different economic object. Not a human being that can be owned. A manufactured cognitive system that can be accessed and directed.
That difference is enormous.
For centuries, cognition was inherently tied to biological scarcity. You could educate it.
Employ it.
Organize it.
Incentivize it.
But you couldn't manufacture another million instances of it simply by building another data center. Now we can...Or at least we are beginning to. That means cognition itself is becoming an industrial product.
Electrons In. Intelligence Out.
This is where all of the enormous infrastructure spending suddenly makes more sense.
Why are companies racing to secure gigawatts of electricity?
Why are we building gigantic GPU clusters?
Why is there renewed interest in nuclear power?
Why are transformers, substations, transmission capacity, cooling technology and semiconductor supply suddenly strategic concerns?
Because those are inputs.
The more interesting question is: What is the factory producing?
Increasingly, the answer is cognition.
Electrons go in.
Intelligence comes out.
Not consciousness.
Not humanity.
Capability.
Coding capability.
Research capability.
Financial analysis.
Scientific reasoning.
Design.
Writing.
Operations.
Strategy.
Customer service.
Planning.
Millions and eventually billions of specialized artificial cognitive workers operating at different levels of sophistication. That's what the infrastructure is for.
We are building factories for minds.
And Now Bring That $1.75 Trillion Number Back to Your Desk
This is where the entire conversation gets almost ridiculous.
Because somewhere in the financial world, people are contemplating whether one company involved in manufacturing this capability might eventually be worth $1.75 trillion.
Meanwhile, you can sit at your desk and access a meaningful fraction of that capability for about twenty bucks a month. Think about the asymmetry of that for a second.
Trillions of dollars of semiconductor development.
Data-center construction.
Power infrastructure.
Research.
Engineering.
Training.
Software development.
Capital investment.
Human knowledge.
All collapsed into a box on your screen that costs roughly what people casually spend on streaming subscriptions. And we're still debating whether ordinary professionals should bother learning how to use it. Of course they should.
That doesn't mean everybody needs to become an AI engineer. It doesn't mean you need 47 agents running around your computer. It doesn't mean every task needs AI.
And it certainly doesn't mean blindly trusting whatever a model tells you.
It means something much simpler: if this much capability has suddenly become this inexpensive to access, ignoring it entirely is becoming an increasingly difficult position to defend. The hurdle isn't whether AI can perform every part of your job.
That's the wrong test. Ask whether it can give you a 10% advantage.
Can it save you thirty minutes?
Can it read something before you do?
Can it improve a first draft?
Can it find the flaw in an argument?
Can it give you ten approaches instead of the two you thought of?
Can it explain something outside your expertise?
Can it write the little piece of software you otherwise wouldn't have built?
Can it help you think?
If the answer to even a few of those is yes, the economics become absurd.
You Don't Need the Whole Trillion Dollars
And this may be the most important part.
You don't need to capture $1.75 trillion worth of value. You need to capture twenty dollars worth. That's it.
If an AI subscription saves you one hour in an entire month, there is a decent chance the math already works. If it helps you make one better decision, the math may work.
If it helps a salesperson write one better proposal...
A programmer solve one problem faster...
A student understand one difficult concept...
An executive see something in a report they missed...
A small business owner automate one annoying process...
A researcher explore one additional hypothesis...
Then suddenly we're talking about one of the strangest bargains in the history of technology. This isn't like buying a million-dollar mainframe in 1970.
It isn't like building a data center. You don't need a research lab. You don't need a rack of GPUs. You don't even need to understand how a transformer model works.
Someone else spent the billions.
Someone else built the infrastructure.
Someone else trained the system.
Someone else secured the electricity.
Someone else assembled the factory.
You get to rent the mind.
The Biggest Divide May Not Be Between Humans and AI
I suspect we're eventually going to stop talking so much about AI replacing people. That's dramatic, but it misses something happening much sooner.
The more immediate divide may simply be between people who have learned to work with machine intelligence and people who haven't.
Two accountants.
Two attorneys.
Two marketers.
Two engineers.
Two entrepreneurs.
Two students.
Same education.
Similar experience.
Similar intelligence.
Except one of them has learned how to surround themselves with inexpensive artificial cognition. The other hasn't.
You don't need science fiction to understand what happens next. One person begins every problem alone. The other doesn't. That compounds.
Maybe slowly at first. Then all at once.
From Silicon to Minds
For fifty years, we kept improving the machine. Faster processors.
More memory.
More storage.
Better networks.
Better software.
More compute.
Then somewhere along the way, the machine stopped merely calculating what we told it to calculate. It began participating.
We learned how to turn silicon into something resembling useful cognition. And now markets are beginning to wrestle with what that capability might ultimately be worth.
Maybe it's hundreds of billions.
Maybe it's $1.75 trillion.
Maybe today's valuations will eventually look absurd.
Maybe they'll look cheap.
I have no idea.
But I think there's a much smaller and much more useful question.
If some of the smartest investors and largest companies on Earth believe the ability to manufacture machine intelligence could create trillions of dollars of value, and you can put a piece of that intelligence beside you at your desk for roughly twenty dollars a month, what exactly are you waiting for?
You don't have to worship it.
You don't have to fear it.
You don't have to understand every piece of it.
And you certainly don't have to hand it the keys.
But you should probably learn how to use it. 😉
Because we spent the last fifty years figuring out what silicon was worth. Now we're beginning to put a price on the minds it can produce. And somehow, for the moment, yours is available by subscription.
Rich Washburn is a technologist, strategist, and Founder & Chief AI Architect of ARIA AI Labs, working at the intersection of AI, infrastructure, communications, and capital. He also serves as Managing Partner and Chief AI Officer at Eliakim Capital.
Comments