Author: Fu Sheng
We've previously discussed Anthropic's strengths and weaknesses, as well as the three records it has set.
Just this Monday, there was another major development in the Silicon Valley AI circle. Anthropic, beating OpenAI to it, confidentially filed for an IPO. Morgan Stanley and Goldman Sachs are joint lead underwriters, with a potential listing as early as October this year.
After the H round raised $650 billion, the valuation reached $965 billion—just a step away from one trillion.
By the time it actually goes public, a trillion is a given, and it could even reach $1.5 to $2 trillion. If it truly hits $2 trillion, it will surpass SpaceX to become the company with the highest Pre-IPO market capitalization in the world.
Many friends have left messages in the background saying, "Fu Sheng, are you exaggerating too much?" and even a fellow entrepreneur directly asked me: When will this wave of AI bubble burst?
My answer has always been the same: Once you understand the situation, you won't think this is a bubble.
Is It Really a Bubble?
That entrepreneur's logic was straightforward: What company is worth $1 trillion after just four or five years? Wasn't this how the 2000 internet bubble burst?
I said, don't jump to conclusions first. The term "bubble" has been shouted about so much that no one really thinks about one question: What are the similarities and differences between the 2000 bubble and today? Pull out these two lines and compare them, and the answer will present itself.
History Is Similar, But It Doesn't Repeat Itself
If, just because Nasdaq crashed in 2000, one claims that AI in 2026 must have a bubble, then that thinking system needs an upgrade.
What's similar? The sentiment. A new wave of technology emerges, capital rushes in, valuations soar, outsiders exclaim they don't understand—this psychological script is indeed identical to back then.
But the differences are the core to judging whether there is a bubble. If you only see the first half and draw a conclusion, that's like marking the boat to find the sword (i.e., being myopic).
Telling Stories and Doing Accounting Are Two Different Businesses
What did internet companies in 2000 rely on for valuation? A domain name, a PPT, a "dream price-to-earnings ratio." No revenue, no profit, not even paying customers, relying purely on stories and imagination to prop up stock prices. Back then, analysts debated, "How big could this company become in the future?" and this "future" had no revenue data to support it.
What Anthropic relies on today is an open hand.
As we discussed before, Anthropic is currently the privately held company with the fastest revenue growth, highest output per employee, and highest valuation in human history.
First, look at revenue growth. Its ARR: early 2025, $1 billion. By the end of 2025, $9 billion. By this May, $47 billion. Internal documents show a year-end target of $100 billion—from $1 billion to $100 billion in less than two years. No second company in business history has ever charted this curve.
And this isn't a bloated figure achieved by burning money. Anthropic's Q2 2026 revenue is estimated at $10.9 billion, marking the first time it achieved operating profit, about $560 million, meaning it started making money even before going public.
Next, look at output per employee. Anthropic currently has about 3,000 employees. Calculated at $47 billion in annualized revenue, output per employee exceeds $10 million.
A programmer equipped with Claude Code can potentially accomplish the work of an entire team from before. And Claude Code, launched less than a year ago, has reached $2.5 billion in annualized revenue as a single product, capturing 54% of the AI programming market.
Finally, look at the valuation logic. Anthropic's main business is selling API subscriptions to enterprises. If you view it as a traditional SaaS company, SaaS valuation has a classic formula: Price-to-Sales ratio equals market capitalization divided by annualized revenue, multiplied by a multiple. Enterprise subscription revenue is very stable; as long as the renewal rate is above 95%, the capital market generally assigns a 10x multiple. If it truly reaches $100 billion by year-end, 10x is $1 trillion.
You could argue that this valuation model itself isn't very reasonable, but you can't say it has no model. The internet bubble was purely priced by imagination. Today, Anthropic's books are laid out piece by piece. Its client list is solid: 8 out of the Fortune 10 are using it, and over 1,000 large enterprises spend over $1 million per year on Claude. Netflix, Spotify, KPMG, Salesforce—all on the list.
Back in the internet bubble days, analysts asked, "How big could this company become?" Today, analysts ask Anthropic, "How much did you make this quarter, and how much more can you make next quarter?" This is the fundamental difference.
From the Carbon-Based Economy to the Silicon-Based Economy
After explaining the two reasons, I still feel it's insufficient. Behind Anthropic's IPO lies a larger, deeper trend: the human economy is transitioning from carbon-based to silicon-based.
I recently said in an internal company meeting that we should rename the HR department to IR, from Human Resource to Intelligence Resource. A company's intelligence level and competitiveness no longer depend solely on how many people it has, how many smart brains, but also on how much computing power it has, how many models, how much AI capability to scale repetitive work.
This concept isn't mine. Companies in Silicon Valley have already validated this logic. Nvidia's Vice President of Applied Deep Learning, Bryan Catanzaro, publicly stated two months ago that his research team has entered a stage where "investment in computing power surpasses investment in manpower," with the cost of computing resources far exceeding employee salaries.
Sam Altman recently mentioned a very interesting phenomenon: many companies exhausted their full-year AI budget in Q1 this year. Companies have discovered that every dollar invested in model capability and computing power directly increases product competitiveness—development efficiency, customer response speed, data analysis depth, things that used to rely on piling up human resources, now run at scale with AI.
The human economy is shifting from being carbon-driven to being driven by a dual-engine of carbon and silicon. This is what is truly happening behind Anthropic's IPO. It's not just a company ringing a bell; it's a price anchor point for a new economic paradigm.
