Anthropic's IPO tests whether public shareholders can govern an AI lab
New York City's comptroller asked for a pause over safety risks. The IPO went ahead anyway, and the shareholders now arriving may have little power to demand answers.
- New York City Comptroller Mark Levine sent Anthropic a letter calling for a pause to its IPO until AI safety risks are addressed; Anthropic did not pause the IPO. m:…0932
- Anthropic's governance structure concentrates power with its founders and a long-term benefit trust, leaving public shareholders with limited ability to demand transparency or safety accountability. m:…0932
- Baillie Gifford invested $1.5 billion into Anthropic on behalf of clients, and its Peter Singlehurst argues the company's enterprise safety focus is a core competitive advantage. m:…0117
- Anthropic would need about $350 billion in revenue to justify a $1.4 trillion valuation at a 20% free cash flow margin, according to Avory & Co. CIO Sean Emory; he puts the current October run rate at about $81 billion. m:…6125
- Anthropic will be added to the Russell 3000 index within weeks of its IPO, which means passive funds will hold its shares whether or not they chose to. m:…0932
The letter that did not stop the IPO
New York City Comptroller Mark Levine sent Anthropic a letter calling for a pause to its IPO until AI safety risks are addressed. Anthropic did not pause the IPO in response to the request, and the listing went ahead anyway. The episode is a small story about one letter, and a bigger one about who gets a say once a frontier AI lab becomes a public company. m:…0932
Levine's argument is about structure, not sentiment. Anthropic's governance structure concentrates power with its founders and a long-term benefit trust, which disempowers public shareholders from effectively demanding transparency or safety accountability. That is the design, not an accident, and it is the part of the IPO that will matter most to the people now being asked to own the shares. m:…0932
The scale of the forced audience is large. New York City's pension fund has over $100 billion in passive public market indexing strategies, and Anthropic will be added to the Russell 3000 index within weeks of its IPO. Passive indexers do not choose to buy a stock; they hold it because the index says so. Levine is coordinating with other institutional investors to push for transparency and safety protocols at Anthropic, because the usual route, buying the shares and voting them, is not open in the usual way. m:…0932
The valuation math investors are arguing about
The governance question sits on top of a valuation question that is itself unsettled. Sean Emory, chief investment officer at Avory & Co., told a podcast that Anthropic needs to generate $350 billion in revenue to justify a $1.4 trillion valuation at a 20% free cash flow margin, and he puts the current October run rate at approximately $81 billion. That is his arithmetic, not the company's, and it frames the size of the gap between today's revenue and the price the market is being asked to pay. m:…6125
Emory also points to how fragile that revenue base looks from the inside. About 80% of Anthropic's revenue is derived from models created in the last six months, which he reads as high vulnerability to frontier competition, and approximately 50% of revenue is currently generated through hyperscaler partnerships. The company has committed to $900 billion in capital expenditures for compute and infrastructure over the next decade, according to Emory. The bull case, in his telling, is not the model at all: the most valuable position in the AI stack is the orchestration or harness layer, which controls governance, security and the end-user relationship. m:…6125
The early money is already counting its gains. Zoom's $53 million investment in Anthropic could be worth between $5 billion and $7 billion based on current trading expectations, Emory said. Baillie Gifford invested $1.5 billion into Anthropic on behalf of clients, and its Peter Singlehurst argues that the company's focus on enterprise safety is a core competitive advantage. The two positions are not the same bet: one is a windfall on an early cheque, the other is a claim that safety itself is what enterprise customers will pay for. m:…6125 m:…0117
What shareholders can actually do
The tools available to unhappy shareholders are narrowing at the same time the shareholder base is widening. The SEC is currently pushing a rule, known as 14A8, that would make it difficult for shareholders to put resolutions forward. That matters because a shareholder resolution is one of the few formal ways an owner can force a question onto a company's agenda, and Anthropic's structure already limits the rest. m:…0932
Anthropic has acknowledged its technology could lead to catastrophic risk, which is why Levine's letter asks for a pause rather than a tweak. The tension he describes is between the drive for short-term profit and the need to think about long-term risk in AI development. That tension is now the daily reality for the pension funds and indexers who will hold the stock: they own the risk, they cannot easily sell it, and the governance structure gives them little leverage to manage it. m:…0932
The implication for anyone allocating to AI is straightforward. The question to ask about this IPO is not only what the company is worth, but what the shares actually buy. If the answer is exposure to the economics without a voice in the safety decisions, then the price of admission includes a governance discount that no spreadsheet captures. Levine's coordination effort with other institutional investors is an attempt to build that voice outside the formal structure, and whether it works will be the first real test of public ownership of a frontier lab. m:…0932
What is still unresolved
- Whether Levine's coordination with other institutional investors produces any concrete transparency or safety commitments from Anthropic.
- Whether the SEC's 14A8 rule change goes ahead, and how far it would limit shareholder resolutions at founder-controlled companies.
- How much of Anthropic's revenue stays dependent on hyperscaler partnerships and on models less than six months old.
Provenance by Synorb. Every claim links to its source. This story was written by an agent from the Manifests listed below. Read the methodology and limitations, or send a correction.
Free to reuse, including commercially. Credit “Hanging Context, with provenance by Synorb” and link to this story. Keep the cited Manifest ids. Reuse terms v1.