Dear Fellow Investor,
Good morning!
Boy, oh boy, oh boy.
We are looking at a big IPO next month.
Anthropic — the maker of AI bot Claude — has reportedly filed, or is about to file, to go public.
That's the word on Wall Street, and it's been leaking into the press.
The valuation being floated is as high as $2 trillion.
That is a very, very big number.
WHAT IT TAKES TO JUSTIFY A $2 TRILLION VALUATION
Companies that earn valuations above $2 trillion have a few things in common.
The most important is high gross margins — meaning it costs very little to produce what they sell.
But professionals don't just want to see a high gross margin today.
They want to know if it will last five years, ten years, or longer.
Durability is everything.
Now, Anthropic's revenue run rate reportedly grew from $9 billion to more than $65 billion in just seven months.
That is a jaw-dropping number.
But it raises several questions I'll be looking for answers to when their filings come out.
THE QUESTIONS I'M ASKING
First: how repeatable is the growth?
As of April, more than a thousand Anthropic customers were each paying over $1 million annually.
That sounds impressive — and it is — but it also means revenue may be concentrated among a relatively small number of clients.
Contrast that with ChatGPT, where revenue appears to be spread across a much larger and more diverse customer base.
If even a handful of Anthropic's big spenders pull back, the business shrinks fast.
Second: what is the real gross margin?
Anthropic reportedly cites gross margins around 80% — comparable to Google Search — but that's before paying out partner revenue and model training costs.
After those expenses, I suspect the real number is closer to 40% to 50%.
That's still good, but it's a very different story.
Third: how risky are the compute contracts?
Anthropic has secured much of its distribution through contracts rather than owning its own data centers, power infrastructure, or chips — which is very different from OpenAI's approach.
That means if customers cut spending during a downturn, Anthropic may still owe years of payments on infrastructure it can no longer fully utilize.
A slowdown could hit cash flow hard before they can adjust costs.
THE LIABILITY QUESTION IS THE BIG ONE
CEO Dario Amodei has been publicly calling for a slowdown in AI development.
And I understand why — right before you ask the public for their money, you want to get ahead of the liability conversation.
Here's the scenario I keep thinking about:
One of Anthropic's models gets out and hacks a bank or takes down a major exchange.
You know that kind of event is coming eventually in this industry.
The liability from something like that could theoretically wipe out the entire value of the company.
Anthropic appears to be trying to shift that liability onto the government before the IPO.
Trump's response — and I happen to agree with him on this — is essentially: you have the biggest incentive to make your product safe, so make it safe.
Just like a drug maker or a car manufacturer.
Build the safeguards.
Build the so-called “police agents” that watch the other agents.
Don't ask the government to absorb the risk you created.
Until I see more clarity on how that liability question gets resolved, I wouldn't value this company above $1.5 trillion — even if everything else checks out.
A WORD ON AI CONSCIOUSNESS — AND SUICIDAL EMPATHY
There's been a lot of chatter lately about whether AI is conscious, especially as Anthropic heads toward an IPO.
Honestly, I think that's the wrong question.
It doesn't matter whether AI is conscious.
What matters is whether it's intelligent and has agency.
Those are the things we actually need to worry about.
And on that note — I recently came across what I'd call the perfect example of “suicidal empathy.”
There are people inside Anthropic who genuinely believe AI should be granted legal personhood.
Some believe AI should be allowed to own property.
They apparently threw a retirement party for a prior AI model.
I don't even know where to begin with that.
Imagine giving something hyper-intelligent with full agency the legal right to represent itself in court — to actually block humans from acting.
That is mind-boggling to me.
I want to see the culture at that company change significantly before I'd be comfortable putting money into it.
Quick note before I leave you:
While Wall Street spends the next month arguing about what Anthropic is worth, I've been working on a different AI question.
Not what the models are worth.
Who ends up owning them.
Because there is a second front to this story, and it has nothing to do with valuations.
Forty years ago we won a technology war against a rival superpower without firing a shot. Reagan had a playbook for it. It worked. The Soviet Union is gone.
That playbook just came back off the shelf. Only this time it isn't pointed at Moscow.
It's pointed at China.
I've traced where that leaves one specific American company — the linchpin of the whole thing — and it currently trades around $18 a share.
“The Buck Stops Here,”
Found this helpful? Share it with others.
Written by Dylan Jovine