Anthropic is worth a trillion dollars. Google is worth four and a half trillion. Sit with those two numbers for a second. One company makes a chat product and an API. The other one runs the internet, owns YouTube, ships an operating system on three billion phones, runs the second biggest cloud, has its own chips, its own models, and a search engine that prints money like a central bank.
And the ratio between them is 4.5 to 1.
There are only two ways to read that.
- Anthropic is wildly overpriced.
- Google is wildly underpriced.
Both cannot be true. One of them has to be. The fun part, and the part nobody wants to admit out loud, is that nobody actually knows which one.
Pretty sure nobody knows what anything is worth anymore.
The case that Anthropic is overpriced
Take the most boring version of the argument first. Anthropic has, generously, single digit billions in revenue. Its margins are not Google margins. It buys compute from Amazon and Google. It pays for the chips that the people it competes with also pay for. Its core product, a frontier model, has a shelf life measured in months, not years, because three other labs are shipping models that are within ten percent of it on most benchmarks every quarter.
A trillion dollars for that is not a valuation. It is a prediction. The prediction is that one of the big AI labs is going to capture an absurd share of the future economy and Anthropic is going to be that lab. Maybe. But the same prediction is also priced into OpenAI, into xAI, into Mistral on a smaller scale, and into whatever Meta is calling its lab this month. They cannot all be right. Most of them are going to be wrong by a factor of ten.
The case that Google is underpriced
Now flip it. Google has the data. Google has the chips. Google has the distribution. Google has the cash flow that funds the chips that fund the data that funds the distribution. If AI is the biggest platform shift in thirty years, the company that already owns the previous platform is the most obvious place for it to land.
Gemini is not behind anymore. It is honestly ahead on some axes nobody talks about. Google has YouTube, which is the largest video training corpus in existence. It has search, which is the largest behavior graph in existence. It has Android, which is the largest deployment surface in existence. And it is trading at a multiple lower than half the SaaS companies on the Nasdaq.
Four and a half trillion sounds like a lot until you write down the list of things Google can do that nobody else on the planet can do, and then it starts to sound cheap. That is the bull case in one paragraph.
Why both cases sound right at the same time
Here is what I think is actually going on. Valuation, the way we used to do it, was a function of cash flows, growth rates, and discount rates. You took the numbers, you projected them out, you discounted them back, and the number that fell out was the price. It was math. Boring math. But math.
That math does not work anymore for anything in this category. Because the inputs are not numbers, they are guesses about which company will be running the world in 2030. And guesses about who runs the world are not a discounted cash flow problem. They are a vibes problem.
Anthropic is priced on the vibe that it might be the company. Google is priced on the slightly older vibe that incumbents lose platform shifts. Both vibes are real. Both are also entirely possible to be wrong about. Nobody is doing the math. Nobody can.
We are not pricing companies. We are pricing how strongly we believe in a story. And the stories keep changing.
What gets priced when you cannot price the cash flows
When the fundamentals stop being legible, the thing that gets priced is narrative. And narrative is cheap to manufacture and very expensive to ignore. Three things end up driving the number.
1. Who is telling the story
Sam Altman walks on stage and a hundred billion dollars of perceived value moves. Dario does an interview and a chart moves. Sundar tweets one screenshot and Google adds half a Nvidia to its market cap. Founders are now load bearing structural elements of the valuation. That is not how markets are supposed to work, but it is how they actually work right now.
2. What got launched this week
A model release moves a hundred billion. A bad demo costs fifty. The half life of a narrative is shorter than a quarter. Public companies used to be valued on annual guidance. Now they are valued on whatever shipped on Tuesday. That is a different game entirely and most of the people pricing it have not noticed.
3. How scared the other side is
Anthropic is priced higher when Google looks scared. Google is priced lower when its own employees leak that the AI roadmap is messy. The valuations are coupled. They move against each other on news that has nothing to do with cash flows and everything to do with who looks like they are losing the week.
The honest answer
If you forced me to pick a side, I would say Google is the closer thing to a real number and Anthropic is the closer thing to a bet. Not because Anthropic is bad. It is excellent. But a trillion dollars on single digit billions of revenue requires a future that has not been priced anywhere else in history to actually come true. It might. It also might not. And the ratio of price to evidence is wider than anything I can find a precedent for.
But the more honest version of the answer is that I do not know. Nobody knows. The people running the funds do not know. The analysts publishing the reports do not know. The founders setting the round sizes do not know. They are all looking at each other to find out, which is how you end up with five companies priced like each one is going to win.
That is the actual interesting thing about this moment. Not that the numbers are big. Numbers have been big before. The interesting thing is that the people who used to be able to tell you whether a number was right have quietly stopped being able to tell you.
What to do with that
If you are building, the lesson is not "ignore valuations." The lesson is that valuations are a terrible signal for what is actually working. A trillion dollar valuation tells you the market believes a story. It does not tell you the story is true. The companies that will look obvious in hindsight are the ones building the boring thing that produces the cash flow nobody is currently modeling.
If you are investing, the lesson is that the easy money trade of the last three years, which was basically "buy whichever AI company got mentioned most this week," is a momentum trade dressed up as a fundamentals trade. Momentum trades end. The question is just when.
If you are watching from the sidelines, enjoy the show. We are living through the largest pricing experiment in market history and the experiment has not concluded. The trillion dollar number for Anthropic and the four and a half trillion number for Google are both real prices that real people are paying. One of them is going to look insane in five years. Maybe both.
The market is not telling you what these companies are worth. It is telling you what people believe about them right now. Those are very different things.
Pretty sure nobody knows what anything is worth anymore. And the people who say they do are the ones you should trust the least.