Anthropic's IPO Structure Leaves Public Shareholders With Little Real Power

Anthropic's pitch has always centered on safety: build powerful AI, keep it under control, make sure humanity benefits. As the company prepares for one of the most closely watched IPOs in recent memory, that control mandate has produced a governance structure that effectively excludes one group from decision-making, the people buying its stock on public markets.

Dual-class structures that let founders retain control without majority economic ownership are common enough in tech listings. Alphabet, Meta, and others have run variants of this playbook for years. Anthropic's version goes further. The company has built a four-layer governance system that, by design, keeps most major decisions, including board composition, out of common shareholders' hands.

The first layer is Anthropic's legal status as a public benefit corporation, a for-profit structure that allows the board to weigh its stated mission, AI safety, alongside shareholder returns. That status offers legal cover: directors making safety-driven decisions that arguably hurt short-term shareholder value are less exposed to investor lawsuits. The current board is chaired by president Daniela Amodei and includes CEO Dario Amodei, her brother, along with Netflix cofounder Reed Hastings, former Microsoft and GM CFO Christopher Liddell, Novartis CEO Vas Narasimhan, and Spark Capital partner Yasmin Razavi. One seat sits vacant.

The second layer sits above the board itself. A Long-Term Benefit Trust, known as the LTBT, holds a special class of stock that gives it power to elect and remove a majority of Anthropic's directors, called Class T directors. This is not a shareholder body in any conventional sense. It is chaired by Neil Shah, a global health expert and former managing director at GiveWell, alongside former Federal Reserve chair Ben Bernanke, national security expert Richard Fontaine, and Harvard physicist Christopher Stubbs. One LTBT seat is also vacant, and existing trustees will pick the replacement themselves, with no shareholder input at all.

Four Layers of Control

Layer three belongs to the founders. Anthropic's seven cofounders jointly hold a special stock class that, when they form a majority among themselves, grants 50.1% of shareholder voting power and the right to elect three board members. If the founders split without a majority, their shares revert to voting in proportion with common stock, a fallback that matters more than it might appear.

The fourth layer is a check on the trust itself. When the LTBT nominates or removes one of its four "Specified" directors without cause, meaning without alleging actual misconduct, the board's nominating committee can block the move and kick the decision to a vote of Anthropic's employees. That detail is the most telling piece of the entire structure. It formalizes something Silicon Valley has already watched play out in public.

OpenAI's board briefly fired Sam Altman in what became known as "the blip," over concerns he wasn't being fully candid with directors. Employee loyalty to Altman, not board authority or trust oversight, forced a reversal and a reinstatement, followed by a new governance framework. Anthropic's structure essentially writes that dynamic into its bylaws in advance. Employees, not public shareholders, hold a real check on both the trust and the board in certain scenarios.

For investors evaluating ANT ahead of listing, the practical takeaway is straightforward: cracks in founder unity could hand common shareholders more influence than the structure initially suggests, while employee votes can override the LTBT on specific board matters. Neither scenario puts meaningful power in the hands of ordinary stock buyers under normal conditions.

None of this control is permanent, at least on paper. Ten years after the IPO, founders can voluntarily end their control, or it lapses automatically once fewer than three of the original seven remain active. The LTBT's authority persists longer: it continues until 85% of common shareholders vote to dissolve it, or the trust itself steps down with board approval. Only after both special classes disappear does ownership translate into actual control for public shareholders, a scenario the company presumably hopes arrives through normal succession rather than crisis.

The governance question lands alongside genuinely strong growth numbers. Anthropic's S-1 filing, made public this week after weeks of leaked drafts, shows revenue of $16.3 billion for the first half of 2026, up sharply from $1.2 billion in the same period a year earlier. That growth came with an operating loss of $693 million, a meaningful improvement from the $2.1 billion loss posted in the prior comparable period. Investors weighing those figures against the governance structure face a split decision: the business is scaling fast, but buyers of ANT stock are purchasing economic exposure, not a voice in how the company is run.

The Bottom Line

Anthropic's governance architecture is a rational response to a real tension, building frontier AI while trying to avoid the pressures that pure profit-maximization can create. Whether that tension justifies locking common shareholders out of board elections for a decade or more is a separate question, and one public investors should weigh carefully before buying in. The revenue growth is real and substantial. So is the governance risk. Anthropic's leadership has bet that safety-minded control structures, backed by a trust of outside experts and reinforced by employee veto power, will outperform a model where shareholders have more direct say. Public markets will now get the chance to price that bet, even if they can't vote on it.