Amazon and Google's Anthropic Stakes Valued at $308 Billion in IPO at $83 a Share
Anthropic is going public at $83 a share, which values the stakes of Amazon (14.7%) and Google (13.7%) at about $308 billion combined. Both companies sell Anthropic cloud services and chips, so their investments used to flow straight back to them as payments for its bills. Now public investors and bondholders will cover losses of tens of billions of dollars per quarter. For developers building on Claude, this lowers the risk of a service shutdown. Keeping a backup model is still a good idea.
AI-processed from TechCrunch; edited by Hamidun News
Anthropic is going public at $83 a share, and the prospectus's biggest surprise is who collects a large share of the money. At that price, the stakes of Amazon (14.7%) and Google (13.
7%) are worth about $307.9 billion combined, or roughly $159 billion and $149 billion respectively. The catch is that both companies are also Anthropic's main suppliers.
Imagine investing in a customer who immediately pays that money back to you for your services: revenue grows, but no new money enters the system. Critics of the AI boom have questioned exactly this kind of "circular" financing. If you build products on Claude or are choosing which API to build a business on, this matters: pricing, rate limits, and the risk that a model provider runs out of money all depend on how sustainable its funding is.
Who gets the money
Amazon has invested $18 billion in Anthropic. Google has invested $12.3 billion since 2023 and promised another $30 billion if Anthropic meets certain targets.
Beyond them, Nvidia, AMD, Microsoft, Broadcom, Micron, and TSMC are taking part in a $26 billion private placement at the IPO price. Anthropic already has multibillion-dollar strategic agreements with all of them except TSMC: the company buys their products.
Why the "circle" may break
When Anthropic trains a model or serves Claude to users, it pays cloud companies, and they get the money back. From the outside, this looks like record cloud revenue and a record fundraise for a startup, but in essence the money moves from one pocket to another.
The IPO changes the picture. The suppliers' investments are now liquid, meaning they have paid off. Anthropic still loses tens of billions of dollars per quarter, but the people paying for that will no longer be suppliers and private investors. They will be buyers of the shares on the exchange and holders of corporate bonds. These sources do not have the conflict of interest that raised questions before.
After the listing, the company expects about $220 billion in cash, cash equivalents, and short-term investments, including IPO proceeds and the $26 billion private placement. According to Gil Luria of D.A. Davidson, Anthropic raised about $141 billion and will receive about $91.7 billion net at $83 a share. The company can pay for part of its future obligations with stock.
Is there enough money to pay for compute
According to regulatory filings, Anthropic has already committed to spend $900 billion, mostly on compute and infrastructure. Some of the contracts allow termination. Joey Brookhart, head of the Tokenomics team at SemiAnalysis, estimates that paying for compute over the next four years requires about $280 billion in cumulative revenue. He considers that achievable: in the third quarter of 2026, revenue was $17 billion, or about $68 billion on an annualized basis.
Brookhart expects public status to make Anthropic a more reliable borrower, although it was recently reported that lenders may not view it as an investment-grade company. He also objects to describing the arrangement as purely circular and calls it beneficial for the ecosystem. Luria distinguishes Anthropic from "neoclouds" like CoreWeave, which he believes depend far more heavily on Nvidia's support.
Risks and takeaways for practitioners
Both analysts name the main risk plainly: Anthropic could lose the race. Its growth now rests on its status as the leader in generative AI. If OpenAI, Google DeepMind, or even cheap open-weight models overtake it, the situation could change quickly. Luria is blunt: if everyone is using someone else's models, a company valued at $1.5 trillion is "worth nothing."
The risk for Amazon, Google, and Microsoft is much lower: even if another model wins, it will still run in their data centers.
The practical takeaway: Anthropic now has a substantial safety cushion, and the chance of sudden trouble at Claude's provider has dropped. But that is no reason to tie yourself to a single model. Leadership can change, so it is better to design your product's architecture so the model can be swapped without rewriting the entire application.
- Nov 19, 2026Google's Gemini 4 Neon runs twice as fast as Carbon and costs 17% less per task
- Nov 19, 2026The same Gemma 4 wrote a backend that passed 80% of checks in opencode, but only 41% in OpenHands
- Nov 19, 2026OpenAI's coding agents would cost $2400 a day per researcher at API prices
- Nov 19, 2026A GitHub Script Catches App Store and Google Play Rejection Reasons Before Review: 542 Stars in a Day
Key Questions Answered
Who gets the money?
Amazon has invested $18 billion in Anthropic. Google has invested $12.3 billion since 2023 and promised another $30 billion if Anthropic meets certain targets. Beyond them, Nvidia, AMD, Microsoft, Broadcom, Micron, and TSMC are taking part in a $26 billion private placement at the IPO price. Anthropic already has multibillion-dollar strategic agreements with all of them except TSMC: the company buys their products.
Why the "circle" may break?
When Anthropic trains a model or serves Claude to users, it pays cloud companies, and they get the money back. From the outside, this looks like record cloud revenue and a record fundraise for a startup, but in essence the money moves from one pocket to another. The IPO changes the picture. The suppliers' investments are now liquid, meaning they have paid off. Anthropic still loses tens of billions of dollars per quarter, but the people paying for that will no longer be suppliers and private investors.
What are the key details on Is there enough money to pay for compute?
According to regulatory filings, Anthropic has already committed to spend $900 billion, mostly on compute and infrastructure. Some of the contracts allow termination. Joey Brookhart, head of the Tokenomics team at SemiAnalysis, estimates that paying for compute over the next four years requires about $280 billion in cumulative revenue. He considers that achievable: in the third quarter of 2026, revenue was $17 billion, or about $68 billion on an annualized basis.
Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews
Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.