Anthropic at $965B: The AI Market Just Told You Who Wins
Anthropic just hit a $965 billion valuation. The number is so close to a trillion that rounding up feels inevitable. When it crosses, and it will, the AI space will have its first official member of the ten-figure club. Not OpenAI. Not Google. Anthropic.
So let's talk about what that actually means. Not for Anthropic. For you.
The number is the message
$965 billion isn't a reward for good behavior. It's a declaration. The market is saying, very clearly, that the AI layer is consolidating. Fast.
Look at the math. Anthropic's last known raise was a $2 billion round in early 2025 that valued the company around $60 billion. Sixteen months later, we're at $965 billion. That's a 16x jump in roughly five quarters. Revenue didn't grow 16x. Users didn't grow 16x. What grew is conviction that this market has room for maybe two or three foundational players, and Anthropic is one of them.
Secondary markets tell the story. Shares that traded at $45 in January 2025 were changing hands at $680 by last week, according to data from Notice and Caplight. That's not organic price discovery. That's institutional FOMO meeting restricted supply.
Why consolidation kills the middle
Here's what nobody wants to say out loud. The "build on AI" gold rush is over for most people. When three companies control the foundational models, the margins sit with them. Everyone else becomes a reseller.
I've watched this movie before. Cloud infrastructure consolidated into AWS, Azure, and GCP. The thousand little cloud providers? Gone. Mobile OS? iOS and Android. The wreckage of WebOS, BlackBerry, and Windows Phone lines the roadside.
AI is following the same pattern but faster. The timeline from "anyone can compete" to "you need $10 billion to show up" compressed from a decade to about eighteen months.
Developers on X have been tracking this shift. @swyx pointed out two days ago that the number of startups building their own foundation models has dropped by roughly 70% since mid-2025. The ones still trying are either in stealth with massive war chests or pivoting quietly to application layers.
What builders should do now
If you're building something in AI, you have three realistic paths.
Path one: Go deep on vertical. Stop building general-purpose AI tools. The horizontal layer is owned. But industry-specific workflows with real domain expertise, real data moats, and real customer lock-in? That still works. A friend running a legal AI startup just crossed $4M ARR because he knows more about patent filing workflows than any model ever will. His product isn't the model. It's the workflow.
Path two: Own the distribution. Models are becoming commodities. Distribution isn't. If you can own a user base or a platform or a workflow that people touch daily, you have leverage. The model providers need you more than you need them. But you need to actually own the relationship, not just be a thin wrapper.
Path three: Get profitable now. Not next year. Now. The era of cheap capital for AI-adjacent startups is ending. Sequoia's most recent memo, leaked on X last week, reportedly told portfolio companies to target profitability within 18 months. That's a signal, not a suggestion.
The uncomfortable truth
Anthropic's valuation isn't your validation. It's your competition.
When the market caps three companies at over two trillion dollars combined, they're not pricing in a thriving ecosystem. They're pricing in extraction. Every dollar of margin those companies capture is a dollar that doesn't flow to application-layer startups.
The builders who will survive this are the ones who stopped pretending the model was their product. Your product is your product. The model is infrastructure. Treat it that way.
The door is closing. Not locked yet, but the latch is clicking. Build something that doesn't need it to stay open.
🪶