On April 29, it was revealed that Anthropic was negotiating a new round of financing, with a valuation that may exceed US$900 billion. If the transaction is completed, this less than four-year-old company will surpass OpenAI in one fell swoop and become the most expensive AI unicorn on the planet. $900 billion. What does this number mean? In terms of A-shares, its market value is higher than Kweichow Moutai. In Silicon Valley, it knocked over the valuation throne that OpenAI had worked so hard for ten years. Even more intriguing is the timeline. Just a few months ago, Anthropic was valued at $60 billion.
Google and Amazon have successively invested in it at a valuation of US$350 billion, with a total commitment of up to US$65 billion. From 60 billion to 900 billion in less than a year,15 times.
Anthropic valuation timeline generated by Claude
Capital doesn’t lie. When the smartest money in the world rushes wildly in the same direction, something must have changed.
What has changed?
The answer is shocking——
According to data from market intelligence company Sensor Tower, ChatGPT uninstalls increased by 132% year-on-year in April this year. Last month, its uninstall rate was even higher, up 413% year-on-year.
The number of uninstalls of ChatGPT skyrocketed563%. During the same period, Claude's downloads surged every week199%.
In many countries, Claude once topped the iPhone free application rankings.
Claude's single-day downloads in the United States exceeded ChatGPT for the first time. On February 28, the app topped the list of free apps in the US App Store and remained there until March 2, with the ranking jumping by more than 20 places in just one week. Claude also topped the iPhone free app charts in Belgium, Canada, Germany, Luxembourg, Norway and Switzerland
Users are voting with their feet. And he shot with his foot on the gas pedal.
OpenAI: The cracks in the empire begin from within
On the surface, OpenAI is still huge.
The GPT series has hundreds of millions of users, Codex has just launched a new wave of craze, and the Stargate project claims to spend US$500 billion to build computing infrastructure.
But the cracks in an empire often crack from within first.
If the loss of users is a pain in the skin, then the shrinkage of "Stargate" is a bone-breaking wound in OpenAI.
According to media surveys, the actual progress of the Stargate project is far less glamorous than the PPT.
$500 billion. 10 nuclear power plants. Humanity's only path to the future. Now, it's a shrinking rental contract.
The British project has been stopped. The Norwegian project was cancelled. Flagship base in Texas, abandoned.
Ultraman said this is called "flexible mentality." Partners say this is called "burning bridges across rivers". SoftBank is angry, Oracle is settling accounts, and Microsoft is silently picking up the slack beside the rubble.
The subtext behind the scenes could not be clearer: When OpenAI begins to shirk its infrastructure responsibilities, it loses control of the physical world.
The pace of fund arrival, the location selection of the computing power center, and the coordination of partners—every link is being worked on.
"The progress is much slower than expected."
At the same time, OpenAI is experiencing a silent brain hemorrhage.
Dario Amodei—Anthropic’s founder and CEO—is himself the former vice president of research at OpenAI.
He took away not only himself, but also a group of OpenAI's core security researchers.
When talking about leaving OpenAI, he said bluntly: "Instead of staying and arguing about other people's vision, it is better to bring people you trust to realize your own vision."
This exodus never stopped. In the past two years, OpenAI's alignment team and security team have continued to move to Anthropic.
What is a company's most valuable asset? It’s not the number of users, it’s not the valuation, it’s the group of people who can define the next generation model.
When this group of people choose to leave, the direction itself is the answer.
Anthropic: From “Safety Lab” to “The Most Expensive Unicorn”
Anthropic’s rise is completely different from the conventional narrative in Silicon Valley.
It didn’t start by burning money to grab users.
Its starting point is a paper on AI safety, and its selling point is "Constitutional AI"-using constitution to constrain model behavior. For a long time, the mainstream view in Silicon Valley has been that these people are too idealistic and cannot do much.
Then, Claude 3.5 Sonnet was released.
The programming ability crushes GPT-4o, long context understanding is far ahead, and the hallucination rate is greatly reduced.
The tide in the developer community changed overnight. The same sentence is everywhere on Reddit, Hacker News, and X: "I quit ChatGPT Plus and switched to Claude."
It’s not just one person who says this, it’s thousands of people who say it.
What's more critical is the enterprise side. Claude's API call volume on AWS has more than doubled in the past six months.
More and more enterprise customers are beginning to migrate their core business from GPT to Claude - not because it is cheap, but because it is easy to use.
Google saw this and threw money at it. Amazon saw this and threw money at it. The two giants have committed a combined US$65 billion. The bet is not on Anthropic's current situation, but on its ability to define the next generation of AI.
But, did OpenAI really lose?
Don’t rush to write an epitaph for OpenAI yet.
The iteration speed of the GPT-5 series is still astonishing. As soon as 5.5 was launched, the shadow of 5.6 appeared in the background log.
Codex is taking off in an all-round way as an agent tool, and the moat of the developer ecosystem cannot be broken overnight.
Another trump card for OpenAI is scale. Hundreds of millions of monthly active users, deep integration with Microsoft, and corporate customers around the world—these existing advantages will not evaporate just because of a round of financing news.
Historically, valuation transcendence has never meant victory or defeat has been decided.
When Facebook went public in 2012, many people felt that Google's social dream was shattered. Looking back a decade later, Google's search empire was never really threatened.
The cruel thing about the AI competition is that today’s throne may not count after the next round of models are released.
The end of Qingping when the wind rises: Something big is happening
Zooming out, this is not a "who wins and who loses" story.
This is a story in which "the power structure of the AI industry is being rewritten."
Two years ago, OpenAI was the only big star.
A year ago, Google Gemini started to catch up.
Today, Anthropic has reached the top position in terms of valuation.
At the same time, xAI is burning money, Meta is disrupting the situation, and open source AI is catching up.
There is no longer just one winner. AI is moving from "one super-dominance" to "separation by a group of heroes."
In this melee, there is only one variable that really determines the outcome -Who can make the next generation model first?.
It’s not about whose PPT looks better, or who has more financing.
AI is not magic, it is heavy industry. It requires astronomical electricity bills, it requires absolute credibility, it requires ruthless financial discipline.
Anthropic’s 900 billion valuation is essentially the market betting that the next “iPhone moment” may not be in the hands of OpenAI.