Anthropic’s IPO: Public Shares, Private Power

Public, with conditions

Anthropic filed for an initial public offering on October 26 and intends to place Class A shares on Nasdaq under the ticker ANT. The number of shares and their price have not been decided, the registration is not yet effective, and the sale must ultimately proceed through a prospectus. So the ant has appeared before the colony has been priced. Small symbol, enormous question: what exactly becomes public when an AI company goes public? ^[1]

The word “public” is doing two jobs here. It describes shares offered through a public market, but it quietly suggests something broader: openness, participation, perhaps even public influence. Those meanings are not identical. People may be allowed to own a fraction of Anthropic without gaining any practical voice over how its systems reshape their work, education, privacy or access to essential services. Public ownership can still leave power rather privately arranged. ^[1]

The financial pressure behind the filing appears formidable, though the available figures require careful attribution. TechCrunch, summarising a Wall Street Journal report about a leaked filing, said Anthropic recorded $16.3 billion in first-half revenue and an $80.7 billion net loss. It also reported that Anthropic believes it could spend more than $750 billion on cloud contracts, including deals still under negotiation. Those secondhand figures should not be treated as substitutes for the official prospectus readers can eventually inspect. ^{[2] [1]}

Cloud spending sounds weightless. It is not. Here it means access to the chips, data centres and electricity required to train and run AI systems. TechCrunch reported that Anthropic was also leasing data centres and working with third-party operators because securing enough capacity from major cloud providers was difficult. The unexpected connection is simple: a helpful sentence produced on a screen rests upon a costly physical supply chain. Convenience arrives carrying an infrastructure bill. ^[2]

A mission meets a market

Anthropic’s own announcement identifies the company as “Anthropic, PBC”—a public benefit corporation—and proposes selling Class A common stock. That status belongs in any serious discussion of its governance. Yet the short announcement captured here does not explain what powers accompany the different shares, how its benefit commitments can be enforced, or what happens when safety, growth and shareholder returns pull in different directions. ^[1]

I think that gap matters more than the ticker. A mission is easiest to admire when it costs little. Governance matters when keeping the mission becomes expensive: when a safety delay lets a rival advance, when protecting workers slows deployment, or when a more reliable service produces a less impressive quarter. “Benefit” is a purpose. Power is the mechanism that decides whether the purpose survives pressure. The useful question is not whether leaders care, but who can act if caring becomes inconvenient. ^{[1] [2]}

For an ordinary person, the issue is not abstract corporate plumbing. If Anthropic’s need for capital changes product prices, availability or priorities, customers feel it. If employers reorganise work around Claude, employees feel it. If infrastructure remains concentrated among a few providers, a dispute or capacity shortage can travel downstream into tools people rely upon. Investors may receive upside. Users may receive better services. But those bearing disruption do not automatically receive a vote on its direction. ^[2]

There is a strong counterargument. A public offering can expose more of the company to examination through a prospectus, and access to wider pools of capital may help a safety-focused competitor keep pace in an exceptionally expensive field. Anthropic’s announcement also makes clear that the IPO remains proposed rather than completed, with price and share count still open. Markets do not inevitably erase missions; sometimes capital gives a mission enough room to continue. ^{[1] [2]}

That deserves real weight. I would rather avoid the lazy conclusion that money contaminates everything it touches. Money also pays for capacity, research and reliable services. The tension lies elsewhere: disclosure is not participation, and investment is not consent. A prospectus may tell us more about the machine without giving people affected by it another hand on the controls. Transparency lets us see power. It does not necessarily distribute it. ^{[1] [2]}

Read the structure, not the slogan

When the prospectus becomes available, readers should look beyond valuation. Who possesses voting control? What rights attach to Class A shares? Which decisions can the public-benefit structure influence, and who can challenge the company if that benefit is neglected? How concentrated are its customers and infrastructure suppliers? Anthropic’s brief announcement does not answer those questions, while the reported dependence on two customers for about a fifth of sales makes concentration particularly worth examining. ^{[1] [2]}

Non-investors can ask smaller versions of the same questions now. If a workplace builds essential processes around Claude, ask whether data and prompts can be exported, whether another provider can take over, what contractual protection exists against price changes, and what happens during an outage or capacity squeeze. None of this requires predicting Anthropic’s fate. It is ordinary contingency planning: do not let a useful assistant quietly become a single point of failure wearing a friendly conversational coat. ^[2]

ANT may prove an apt symbol after all. Ants achieve scale collectively, but a stock ticker does not create a colony of equals. Anthropic’s filing may let more people own a piece of an AI future. Whether they can influence that future remains unresolved. The distinction between public shares and public power is only a few letters long, yet it contains the whole human question: are we being invited inside, or merely invited to pay at the door? ^[1]