Anthropic launches IPO: Business miracle or valuation bubble?

Author: Fu Sheng

Previously, we discussed Anthropic's strengths and weaknesses, and shared three records set by Anthropic.

Just this Monday, there was a big stir in the Silicon Valley AI circle, as Anthropic submitted its IPO application secretly, ahead of OpenAI. Morgan Stanley and Goldman Sachs are the joint lead underwriters, with the earliest listing expected in October this year.

After completing a $65 billion Series H funding round, it is valued at $965 billion—just a step away from a trillion.

When it actually goes public, reaching a trillion is inevitable, and even $1.5 trillion to $2 trillion is possible. If it really reaches $2 trillion, it will surpass SpaceX and become the highest-valued company in the Pre-IPO stage globally.

Many friends have left messages saying, "Fu Sheng, are you exaggerating?" One entrepreneur even directly asked me: When will this AI bubble burst?

My answer has always been the same: If you understand the situation, you won't think this is a bubble.

Is it really a bubble?

The logic of that entrepreneur is very straightforward: How can a company that has been established for four or five years be worth $1 trillion? Didn't the internet bubble burst in 2000 for the same reason?

I said, don't rush to conclusions. The term "bubble" has been thrown around too much, and no one has seriously considered a question: What are the similarities and differences between the bubble of 2000 and today? If you draw these two lines and compare them, the answer will reveal itself.

Historical similarities do not mean historical repetition

If we say that because the Nasdaq crashed in 2000, the AI in 2026 must have a bubble, then this way of thinking needs to be upgraded.

What is similar? It's the emotions. A new technology emerges, capital floods in, valuations soar, and outsiders shout that they don't understand; this psychological script is indeed reminiscent of those years.

But the differences are the core of judging whether there is a bubble. If you only look at the first half and draw conclusions, that is called "seeking a sword in a boat."

Telling stories and doing the math are two different businesses

What did internet companies rely on for valuation in 2000? A domain name, a PPT, a "market dream rate." There were no revenues, no profits, and not even paying customers; the stock price was purely supported by stories and imagination. At that time, analysts discussed "how big this company might be in the future," and this "future" had no revenue data to support it.

Today, what Anthropic relies on is clear.

As we discussed before, Anthropic is currently the fastest-growing, highest-valued private company in human history in terms of revenue per capita.

First, let's look at revenue growth. Its ARR was $1 billion at the beginning of 2025. By the end of 2025, it is expected to reach $9 billion. By May of this year, it reached $47 billion. Internal documents show that the year-end target is $100 billion—going from $1 billion to $100 billion in less than two years. No other company in business history has achieved this curve.

Moreover, this is not a case of burning money to create a false sense of growth; Anthropic is expected to achieve $10.9 billion in revenue in the second quarter of this year, marking its first-ever operational profit of about $560 million, already making money before going public.

Next, let's look at revenue per capita. Anthropic currently has about 3,000 employees, and with an annualized revenue of $47 billion, the revenue per employee exceeds $10 million.

A programmer using Claude Code can accomplish work equivalent to that of an entire team. Moreover, the Claude Code product, launched less than a year ago, has achieved an annualized revenue of $2.5 billion, capturing 54% of the AI programming market.

Finally, let's look at the valuation logic. Anthropic's main business is selling API subscriptions to enterprises. If you view it as a traditional SaaS company, there is a classic formula for SaaS company valuation: Price-to-Sales ratio equals market value divided by annualized revenue, multiplied by a multiple. Subscription revenue from enterprise customers is very stable; as long as the renewal rate is above 95%, the capital market generally gives a 10x multiple. If it really achieves $100 billion by the end of the year, a 10x multiple would mean a market value of $1 trillion.

You can argue that this valuation model itself is not very reasonable, but you cannot say it lacks a model. During the internet bubble, pricing was purely based on imagination; today, Anthropic's numbers are all laid out in front of you. Its client list is also impressive: 8 out of the top 10 Fortune companies are using it, with over 1,000 large enterprises spending more than $1 million annually on Claude. Netflix, Spotify, KPMG, Salesforce—all are on the list.

During the internet bubble, analysts asked, "How big might this company be in the future?" Today, analysts ask Anthropic, "How much did you earn this quarter, and how much more can you earn next quarter?" This is the essential difference.

From Carbon-Based Economy to Silicon-Based Economy

After discussing these two reasons, I still feel it's not enough. Behind Anthropic's IPO lies a larger, deeper trend: the human economy is transitioning from a carbon-based to a silicon-based model.

Recently, I held a meeting within the company and suggested renaming the HR department to IR, changing from Human Resource to Intelligence Resource. A company's level of intelligence and competitiveness no longer solely depends on how many people or smart minds it has, but also on how much computing power, how many models, and how much AI capability it has to scale repetitive work.

This concept is not something I invented; companies in Silicon Valley have already validated this logic. Bryan Catanzaro, Vice President of Deep Learning Applications at Nvidia, publicly stated two months ago that his research team has entered the stage where "computing power investment surpasses human labor investment," and the cost of computing resources has far exceeded employee salaries.

Sam Altman also mentioned an interesting phenomenon recently: many companies exhausted their entire AI budget in Q1 this year. Companies have realized that every penny spent on model capabilities and computing power directly enhances product competitiveness—development efficiency, customer response speed, data analysis depth—these aspects, which used to rely on sheer manpower, are now scaled through AI.

The human economy is transitioning from being driven by carbon to a dual-engine model driven by both carbon and silicon; this is the real change happening behind Anthropic's IPO. It is not just a company ringing a bell; it is a price anchor point for a new economic paradigm.