The One IPO I'd Buy Early

The $2 Trillion IPO

Anthropic, the maker of Claude, plans to publicly unveil its IPO prospectus shortly after Labor Day.

On one side, you have the most ambitious AI company heading toward public markets. Claude has evolved from a chatbot into something businesses can actually build around.

You can write software, analyze data, automate workflows, connect APIs and, with Claude Code, build surprisingly complex products.

On the other side, Anthropic is reportedly preparing to tell investors that its potential market opportunity exceeds $30 trillion, topping SpaceX's already “ridiculous” $28.5 trillion estimate.

And as a user, I also think Claude is currently the most reliable LLM.

But that alone doesn’t make it a great investment.

Help us become the fastest growth stock newsletter:
Share HappyStocks with Friends

Anthropic develops the Claude family of large language models.

Its products range from the consumer Claude app to Claude Code, enterprise AI tools and an API that allows developers to integrate Claude directly into their own software.

Anthropic isn't trying to make money from people paying $20 per month for a chatbot.

It wants Claude sitting below other companies' products, software and workflows. That's potentially a much larger business.

Developers are the ones that burn tokens and ultimately bring in revenue.

And I believe the AI industry is just getting started.

Anthropic just mapped out which jobs AI could potentially replace.

The exact revenue mix isn't publicly audited yet, so treat estimates cautiously.

Roughly 80% of revenue comes from business customers, with the remainder coming from consumer subscriptions.

Anthropic monetizes through direct API usage, enterprise contracts, Claude Code and consumer plans. Enterprise customers generally pay for seats plus actual model usage.

Claude Code is interesting.

Businesses aren't just asking Claude questions anymore. Developers are letting it read codebases, write features, fix bugs and execute multi-step engineering tasks.

That moves AI from a productivity tool toward actual labor replacement.

I wouldn't call it wide. Not yet.

Training a frontier model like Claude requires enormous amounts of data, compute, engineering talent and capital. Those are meaningful barriers.

But Anthropic isn't alone.

OpenAI, Google, DeepSeek and open-source ecosystems give users plenty of alternatives.

Models can also improve quickly. Today's leader can become tomorrow's second-best model within a few releases.

Anthropic's opportunity is to turn model quality into enterprise integration and switching costs.

Once Claude is deeply embedded across a company's codebase and workflows, replacing it becomes much harder.

That's where the moat could eventually form.

And I think it does form.

Swapping a model out sounds easy on paper. In practice, once Claude is wired into a company's codebase, internal tools and approval workflows, ripping it out means re-testing everything that touches it.  That's not a pricing decision anymore.  That's a migration project.

I also don't think this ends with one winner. I see several, with Google and Anthropic leading.

Anthropic reportedly believes its TAM (total addressable market) exceeds $30 trillion.

For perspective, that's roughly a quarter of annual global GDP.

So yes, Anthropic is basically arguing that Claude could eventually address an economic opportunity comparable to the output of the United States.

I wouldn't put $30 trillion into a valuation model.

TAM is not revenue.

The number is better viewed as a statement about how broadly Anthropic thinks AI could penetrate software, labor, research and enterprise operations.

Nevertheless, they’re still far from capturing their TAM.

Anthropic's annualized revenue run rate crossed $47 billion in May, according to the company.

By the end of July, it had reportedly surpassed $65 billion.

Preliminary Q2 revenue was more than $11.5 billion, over 14x the year-ago level and more than double Q1's $4.73 billion.

A run rate isn't actual annual revenue. It essentially takes current revenue performance and annualizes it.

Still, going from $47 billion to $65 billion in roughly two months tells you how quickly demand is moving.

What the market has underrated is the margin profile.  Anthropic runs roughly 80% to 85% gross margins on inference, and posted its first adjusted quarterly operating profit of $559 million this year (leaked info per Bloomberg).

Behind that land grab is a real business.  The engine is profitable. The money is going back out the door by choice, not by necessity.

This is one of the biggest risks to the entire Anthropic investment thesis, and it's pretty simple.

Anthropic charges customers to use Claude.

But there are also open-source AI models that companies can download, modify and run themselves. Think models from Meta, DeepSeek and others, often accessed through tools like Ollama.

That creates a potential pricing problem.

Imagine Anthropic charges a company $100 to complete a certain amount of AI work.

If an open-source model can eventually do 90% or 95% of the same job for $20, some customers will ask why they're paying the premium.

The bull case is that enterprises won't always choose the cheapest model.

They care about reliability, security, support, compliance, uptime and integration with their existing systems. A bank or pharmaceutical company may happily pay more if Claude consistently performs better and is easier to deploy safely.

So as an investor, I wouldn't ask whether open source will exist. It will.

The real question is whether Claude can remain good enough to justify premium pricing?

The answer depends on how you define "good enough." It's really about the broader enterprise value proposition.

Why Claude Wins the Premium…

 Mega-corporations (banks, healthcare) will happily pay a premium for Anthropic’s strict security, compliance guarantees, and legal indemnification.

Downloading a "free" or cheap model isn't really cheap. Running it at scale will require expensive GPUs, the best engineers, and continuous maintenance. Claude is plug-and-play.

For complex tasks, like AI agents navigating software or executing multi-step business logic, Claude’s superior reasoning and reliability justify the extra cost.

Why Open-Source Wins…

With open-source models like Meta's LLaMA closing the performance gap, premium models will struggle to justify costing 5x more for a mere 2% improvement.

Companies can use Claude to train and fine-tune their own smaller, ultra-cheap open-source models, essentially using Anthropic to build its own cheaper replacement.

Once open-source tooling (like Ollama or cloud-hosted options) becomes simple to deploy, corporate finance departments will force IT to switch to the cheaper alternative to cut costs.

My Verdict:  As an investor, you shouldn't worry about Claude losing the mass market to open source; it was always going to lose the mass market.

The real metric to watch is Anthropic's enterprise retention in high-stakes verticals. If Anthropic can maintain a market lead in complex reasoning, agentic reliability, and enterprise-grade safety features, it should easily justify its premium pricing to the world's largest companies. If that lead shrinks to parity, the pricing pressure will become a serious threat.

We're getting surprisingly close.

Claude Code can already write and modify software, work across large codebases and automate substantial portions of development.

Anthropic is also expanding Claude into enterprise workflows, science and increasingly sophisticated agentic tasks. Its recent releases continued pushing further into coding and research.

The question is shifting from "Can AI answer this?" to "Can AI complete this?"

That's a much larger market.

The numbers we have are still secondhand, and nothing here has been audited yet.

So, when the prospectus drops, I'm watching one thing: whether the margin picture holds. If it does, this is a profitable company choosing to spend. If it doesn't, the story changes and so does the price I'd pay.

The bull case comes down to three things.

Growth is exceptional. The addressable market is enormous. And the product is genuinely good.

Anthropic has also demonstrated strong enterprise traction. As of February, it said more than 500 customers were spending over $1 million annually, with eight of the Fortune 10 using Claude.

If AI becomes a meaningful layer of corporate infrastructure, Anthropic has a legitimate chance of becoming one of its largest providers.

Price matters.

Anthropic's May funding round valued it at $965 billion. Now the IPO conversation is centered around a potential valuation approaching $2 trillion.

At $65 billion of annualized revenue, $2 trillion would equal roughly 31x run-rate sales.

That's a lot of future growth and success already priced in.

Then there's compute. Anthropic has committed roughly $517 billion over the next decade to lock up future capacity.

That's the largest capital land grab in corporate history.  It's a bet that demand keeps compounding long enough to fill that capacity.  If it does, they own the supply. If it doesn't...well...you know how that story goes.

Over five years, I think the key question is whether Claude becomes infrastructure or remains a product.

Infrastructure gets embedded.  Infrastructure creates switching costs.  Infrastructure gets very difficult to remove.

Over ten years, the range of outcomes becomes enormous. Anthropic could become one of the world's most important software companies, or AI models could commoditize enough that much of today's perceived value moves elsewhere in the stack.

Regulation is another risk.

In June, the Commerce Department ordered Anthropic to cut off Fable 5 and Mythos 5 for any foreign national, worldwide, effective immediately. Anthropic couldn't verify nationality in real time, so it pulled both models for every customer that night. The controls were lifted June 30. About eighteen days of downtime. 

That actually cuts against my own moat argument.

I said switching costs lock enterprises in. Eighteen days of forced downtime is what convinces a CTO to wire up a second model as a fallback. Which lowers the switching cost.

I still think the moat holds. A backup model isn't the same as a migration. But the outage gave every large customer a reason to build one.

You've probably seen the "automate your entire business with Claude" videos on social media.

The interesting part is that it's increasingly doable.

But I wouldn't take a $30 trillion TAM literally.

Professional investors will probably treat that number more like a mission statement than a financial forecast.

I almost never buy a company right when they go public. Anthropic is one of the few exceptions where I'll probably buy sooner than later. 

Growth this steep usually shows signs of some deceleration. I don't see it soon.   Also, the underlying economics are healthier than the valuation suggests. 

So yes, I'm likely paying up for future growth here.

Happy Investing,

Ralph