Claude 的公司要上市：钱多了十四倍

First, look at two numbers.

One is 16.3 billion dollars.

The other is 80.7 billion dollars.

Both are in dollars, both for the first six months of this year.

One is revenue.

One is a loss.

Revenue is fourteen times last year's.

The loss also increased a lot from last year.

A company whose revenue grew fourteen-fold - why did it also lose more?

In Anthropic's IPO document, this is the most interesting part.

The value of this document is not in the numbers.

Anthropic makes AI.

Its product is called Claude.

The document was shown to everyone on Monday.

The company wants to list on Nasdaq under the name ANT.

It could raise more than 100 billion dollars.

Its valuation could reach 2 trillion dollars.

If it succeeds, it will surpass SpaceX's June listing.

That would be the largest IPO in history.

SpaceX's valuation at the time was 1.77 trillion dollars.

In May this year, Anthropic's valuation was 900 billion dollars.

The other AI company, OpenAI, was valued at 730 billion dollars then.

OpenAI also wants new money, at a valuation that could reach 1.5 trillion dollars.

So this is not just one company's story.

It is a competition between two AI companies.

It is also an exam for all of AI.

Now back to that huge loss.

Of the 80.7 billion, 72.4 billion is not money in the company's hands.

It is just a number.

The reason is convertible notes.

A convertible note is a kind of borrowed money.

These convertible notes will later turn into shares of the company.

A share is a little piece of the company.

The higher the company's valuation, the bigger this money becomes on paper.

The company's shares also became more expensive.

The rules require the company to write a very big number.

It is not that the company actually spent 72.4 billion.

Take out this one-time money, and the money the company actually lost was only 693 million dollars.

At this time last year, that number was 2.1 billion dollars.

Take out the one-time money, and the company even had a profit of 1.6 billion dollars.

Profit was about ten percent of revenue.

The company also has a lot of cash on hand: 3.9 billion dollars.

At this time last year, the company was down 1.5 billion dollars.

So this IPO document contains two stories.

One is the 80.7 billion on paper.

The other is the 693 million the company actually lost.

Why did revenue grow so much?

Because the company's customers are using Claude.

They use it to write code, write text, and do analysis.

Last fall, Claude became much better at writing code.

Many tech companies then began using it widely.

The document specifically says that more than 1,900 customers each spend over 1 million dollars a year.

Among them, 250 customers each spend over 10 million dollars a year.

The company has a staff of 5,300 people.

The company has been around for five years and has raised about 130 billion dollars in total.

The company has a very big goal.

That goal requires a lot of money.

Much of this money will be used to buy computers.

The document says the company spent 8.95 billion dollars on computers.

Last year, that number was 2.33 billion dollars.

The company also uses other people's computers.

Over five years, it will pay about 125.2 billion dollars.

So after going public, the company will have a lot of money.

This money will quickly turn into more computers.

More computers will turn into stronger AI.

But in the same document, there is another voice.

Anthropic itself acknowledges that AI may bring great danger.

The document fully explains the risks.

They include AI possibly not doing what people tell it to do.

They also include AI being used by people to do very bad things.

Its technology also makes many governments uneasy.

Because it could get into other people's security systems.

This year, five to seven percent of revenue came from customers later found not to be following the rules.

In October, the U.S. FTC sent it a demand for information.

It also still has issues with the U.S. government.

These words were not said by outsiders.

The company itself wrote them in the document.

A company that sells AI describes AI as this dangerous in its own IPO document.

This phenomenon is obvious.

American investors look at the revenue on one side and these risks on the other.

They want to know whether this revenue can keep being this good.

The timing of this document is also very interesting.

AI has been a thing for four years now.

Investors are starting to doubt when all the money poured in will come back.

This month, Oura, which makes smart rings, delayed its listing.

A ring is something you wear on your hand.

Oura's ring can monitor your body.

SoftBank's SB Energy also delayed its listing.

They are all waiting.

If the people who want AI really become fewer, many companies will struggle.

Whether the money in AI can be recovered is the big question for this round of AI.

This document is not just numbers.

It is also asking some questions.

AI is still at a very early stage.

This stage costs a lot of money.

Whether the money can be recovered is the core question.

For investors, this is the thing they most need to think about.

The company's advantage is that its AI is strong.

But can this advantage last?

The data in the document shows what the company looks like.

The data also shows that customers are increasing year by year.

Will this trend continue?

No one can say for absolute certain that AI is safe.

Nor does anyone know the degree of danger.

The company itself acknowledges that AI carries great risk.

The company stresses that it wants to do AI well.

Who holds power over AI is also a big question.

This is a major issue.

Now AI companies want to go public and show everyone their money.

Anthropic's story is precisely trying to answer this question.

It was founded in 2021.

The people who founded it were Dario Amodei and his sister Daniela.

Plus five other people.

They all used to work at OpenAI.

They left to start their own company because they disagreed about how AI should develop.

In October 2024, Dario wrote something called "Machines of Loving Grace."

He acknowledged that most people misunderstand the benefits of AI.

And also misunderstand the risks of AI.

He also said AI could help people find new drugs and improve human health.

It could also drive the economy to do better.

The meaning of this is not just about money.

So some people call them doomers.

Of course, some people also dislike Anthropic.

Dario advocates government regulation of AI.

He also does not like making AI's technology open to everyone.

Nvidia's Jensen Huang and Meta's Mark Zuckerberg have both criticized him.

They say he wants to keep AI power in his own hands.

Some say his power is too great.

This IPO document shows people a lot of money on one side and says it carries great risk on the other.

It is like a mirror.

It reflects both the best and the most dangerous parts of this round of AI at the same time.

If your job involved an AI that is very smart and may not listen to you, would you use it?