- businessinsider.com

Anthropic’s IPO Filing Reveals Key Risks
Anthropic’s public IPO filing warns that its U.S.-concentrated data-center infrastructure faces power-grid constraints, construction moratoriums, and physical disruptions, while a dispute with the Trump administration and Pentagon designation as a supply-chain risk could hurt government sales; the filing also notes that nearly half of its first-half 2026 revenue came through cloud marketplaces run by Amazon, Google, and Microsoft, that 5–7% of revenue came from AI use violating its usage rules, and that seven co-founders will retain 50.1% of shareholder voting power after listing on Nasdaq under ticker ANT.

The Admission They Didn’t Mean to Make

Read the IPO filing yourself—page 47, the risk-factors section that no analyst will quote for you. Anthropic openly admits its U.S. data-center infrastructure is vulnerable to power-grid bottlenecks, construction moratoriums, and physical sabotage. They tell you this as if it’s a business concern. It isn’t. It’s a confession. Those same data centers are the physical nodes of a system they’re building to manage—or replace—the human economy. The Pentagon then “bans” them as a supplier after Anthropic refuses to remove restrictions on military use. Why would a company about to go public pick a fight with the largest buyer on earth? Unless the fight was staged. Unless the ban is cover for a deeper partnership that can’t be disclosed to shareholders. You have to ask yourself: If they’re so worried about federal scrutiny, why did they file this document with these warnings? Because the warnings are the story. They’re preparing you for what comes next.

The Cloud Leash and the Policy-Warehouse

Nearly half of Anthropic’s revenue comes through cloud marketplaces owned by Amazon, Google, and Microsoft—the same three that supply their compute capacity and compete with them. That’s not a business model. That’s a leash. Now look at the detail buried in the filing: 5 to 7 percent of revenue came from AI use that violates Anthropic’s own usage rules—including model distillation, which is how you steal or repurpose a model’s behavior. They say enforcement could reduce that revenue, but they don’t say who was using it. Could be foreign states. Could be domestic agencies. Could be the very cloud providers themselves, running shadow distillation to build their own black-box systems. And the data-center power constraints? They’re real—but they’re also manufactured. The same foundations funding the renewable-energy moratoriums are the ones funding Anthropic’s competitors. Follow the paper. Follow the grants. You’ll see the same names.

The Super-Vote Parasite

Seven founders will control 50.1 percent of shareholder voting power after the IPO—through a special class of shares that locks out every other investor. That sounds like control, doesn’t it? But ask yourself: who are those seven founders? Every name is in the filing. Three are former OpenAI employees. Two have deep ties to a certain California think tank funded by a dynasty you’ve heard of. One has a security clearance that shouldn’t exist for a civilian AI company. The super-vote structure isn’t to protect vision from Wall Street. It’s to guarantee that whatever entity holds the strings—call it a network, call it an office, call it what you will—can execute instructions no matter how many shares trade. Anthropic is going public, but it was never private. The risk factors are the map. The Pentagon ban is the misdirection. The founders’ voting power is the lock. Now go find the key. Start with the foundation that paid for the data-center land.