Anthropic adopts a PBC plus long-term trust governance structure in its bid for IPO

Dario Amodei, who once served as OpenAI's Vice President of Research, co-founded Anthropic, and the company has joined the ranks of the world's top AI firms on the strength of its Claude large models.

When news of Anthropic's IPO push came out, most people's attention landed on its revenue growth, model capabilities, and the trillion-dollar valuation expectations. Few noticed its governance structure, which departs from the usual Silicon Valley model.

That is hardly surprising: when Silicon Valley tech companies go public, the playbook for control has long been fixed—mostly dual-class AB shares, with voting power held in the founders' hands.

Anthropic did not adopt the traditional model; instead, it uses a Public Benefit Corporation (PBC) layered with a long-term independent trust.

As a result, power at Anthropic lies neither with the founders nor with the investors, but is handed to a group of outside trustees who hold no shares and receive no profits; former Federal Reserve Chair Ben Bernanke is on that list of trustees.

This design collides with a long-standing underlying consensus in capital markets: a listed company should prioritize shareholder returns.

Has the value of the equity investors hold already been quietly discounted? This is not an armchair legal question; it is a real test that Anthropic's IPO places before Wall Street.

How it unfolds will also leave a reference sample for every large-model company that plans to go public later.

Breaking down the dual-layer structure: the trust holds the real power

In U.S. capital markets, the vast majority of tech companies take the form of a traditional business corporation, a C Corp.

For such traditional companies, the law imposes a hard requirement: when management and the board make major decisions, their primary goal must be to make shareholders more money. If management forgoes short-term high profits out of considerations such as safety or social impact, shareholders can sue the board and seek damages.

A Public Benefit Corporation (PBC) is likewise a for-profit enterprise; it can raise funds and issue equity normally, can list and pay dividends in the future, and competes in the market like any ordinary commercial company.

Its most core change is that it rewrites the board's legal responsibilities.

Anthropic's entire governance architecture is a stack of two layers. One is the Public Benefit Corporation as the legal vehicle; the other is the long-term benefit trust that holds actual control. Together they form this highly unusual constraint mechanism.

Under PBC rules, management need not single-mindedly prioritize shareholder returns when making decisions; it may legitimately weigh public safety and social impact.

AI company governance paths diverge sharply: Anthropic chooses a middle road

Actually, this corporate form was not invented by Anthropic. Outdoor brand Patagonia, crowdfunding platform Kickstarter, and ice cream brand Ben & Jerry's all adopted the PBC structure early on.

Across the world's top AI sector, leading companies have taken three distinctly different paths in choosing their corporate structures; the differences among OpenAI, Anthropic, and xAI are especially stark.

At star company OpenAI, the entity that conducts commercial operations externally is also registered as a PBC. OpenAI uses a special structure: a nonprofit foundation sits at the top and holds control, with a for-profit subsidiary beneath it conducting commercialization, thereby constraining capital's profit-seeking behavior.

Compared with OpenAI, Anthropic's path is more of a compromise. The two appear to be aimed at the same goal of restraining capital, but their underlying architectures differ in essence.

OpenAI's top layer is a nonprofit foundation; Anthropic's main entity remains a for-profit Public Benefit Corporation (PBC).

Anthropic merely carves out corporate control and hands it to outside trustees of an independent trust. The OpenAI foundation directly holds equity; Anthropic's trust holds no shares and only holds voting control.

In 2023, Elon Musk founded xAI, initially to take on OpenAI. The company briefly registered as a PBC, then chose to abandon that and revert to an ordinary commercial corporate structure.

With xAI shedding this added mission constraint, all decisions revolve around commercial growth, capital operations are flexible, expansion speed is not bound by institutional limits, and the safety baseline relies on corporate self-discipline, with no hard constraints from law or trust.

Anthropic sits between the two. It retains market-based financing channels and will not have its hands tied by a nonprofit entity as OpenAI does; at the same time, through the independent trust mechanism, it turns its safety mission into a hard constraint, reducing the chance that short-term capital impulses override long-term risk control.

Anthropic is trying to balance two things that are hard to reconcile: freedom for commercial expansion and the public responsibility attached to AI technology.

This divergence conceals an underlying industry disagreement: is a super large model just an ordinary business, or a technology with infrastructure attributes and inherent public risk? Different judgments give rise to completely different corporate systems.

Why do large-model companies attach so much importance to corporate responsibility?

In reality, the technical risks of large models spill over strongly; if commercialization is pushed forward blindly, it will give rise to a series of social-level risks.

If an ordinary commercial company's governance framework were adopted, capital could easily pressure management into loosening model permissions and launching high-risk businesses for the sake of short-term financial reporting. The PBC leaves room for the board at the legal level, allowing it to place AI safety above short-term profit.

But Anthropic judged that a PBC alone is not enough. A PBC can only protect the board legally; it can hardly stop shareholders or founders from adjusting the direction of development under enormous commercial temptation. To this end, the company additionally established the LTBT long-term independent trust, building a dual constraint mechanism.

The PBC is merely a legal vehicle; on its own it cannot constrain capital. What truly holds real power, and is also the most controversial, is the LTBT long-term benefit trust.

The trustee team consists of externally selected experts. They do not hold Anthropic equity, cannot benefit from the value increase of a listing, and do not participate in corporate dividends; their compensation only covers their time costs.

The trustees have two core powers: first, the right to replace a majority of board members, giving them core say at the personnel level; second, when the company engages in major transactions such as large-scale non-routine borrowing, selling more than half of its equity, or transferring core AI models that reach a threshold, it must report to the trust five days in advance.

After studying the relevant clauses, many commercial law scholars have pointed out that this five-day reporting requirement is not a simple notification; once implemented, it will form a soft veto power.

Once the trust questions whether a transaction poses a safety risk, it can block the transaction through personnel appointments and removals, sustained pressure, and other means.

This differs fundamentally from the AB share model: AB shares rely on founders holding high voting rights, and the power remains inside the company; Anthropic, by contrast, hands this core power to a third party independent of shareholders and the founding team.

Long-term barriers and an unavoidable valuation discount

Anthropic's structure is equivalent to surrendering part of shareholders' inherent rights in advance.

This structure has already produced two completely opposing views within institutions, and the difference in tolerance between the primary and secondary markets is especially pronounced.

The bullish logic is clear. As a third-party gatekeeper, the trust can avoid the control struggles and internal friction between founders and investors common among Silicon Valley companies, ensuring strategic continuity.

Amid the trend of continuously tightening global AI regulation, a governance structure with built-in safety constraints is itself a compliance moat. It is hard for capital to pressure the team into taking technical risks for quarterly reports, and slow variables such as R&D investment and safety assessments can continue to receive resources.

Institutions in the primary market are more willing to pay for this long-term narrative. The secondary market's concerns are equally realistic, and are also the biggest source of uncertainty for the future share price.

The core contradiction is the structural dilution of shareholder rights. Investors buy shares, yet can hardly exert influence over transactions capable of changing the company's fate.

The switch for major matters rests in the hands of a group of trustees who are not accountable to shareholders. Once the trust vetoes a merger or licensing plan that could bring huge returns on safety grounds, shareholders can only passively accept the outcome, lacking effective checks and balances.

Traditional due diligence can calculate revenue, gross margin, and the competitive landscape, but the market cannot build a valuation model around the trustees' subjective judgments.

This gives rise to an unavoidable paradox: the company has top-tier technology and a fast-growing business, possessing all the traits of a high-growth target, but the rights carried by its equity are incomplete. The market will ultimately assign a discount for this deficiency in rights.

The primary market bets on long-term potential and can accommodate this discount; most secondary-market funds operate under clear assessment cycles and are naturally more cautious about this uncertain governance risk.

This cognitive gap will run through the pricing game for a long time after Anthropic's listing.

A governance experiment for the entire industry; the test is still ahead

From a longer-term perspective, the significance of this matter has already gone beyond the success or failure of Anthropic's own IPO.

In the medium term, as the share price fluctuates after listing and major transactions are successively implemented, discussions about the trust's excessive power and opaque rules will recur. Once the trustees' decisions come into direct conflict with investor interests, shareholder lawsuits, institutional position adjustments, and valuation reassessments are all foreseeable plotlines.

If Anthropic's mechanism can run smoothly and maintain a stable balance between commercial expansion and risk control, it has the chance to become a new model for leading large-model companies.

Capital markets have operated for a century on a foundation built on the logic of shareholder primacy. Wave after wave of technological change has reshaped products and business models, yet this underlying governance logic has rarely faced substantive challenge.

We are now witnessing a slow institutional adjustment. Should capital surrender part of its power for public safety, and how should commercial returns and long-term risks be weighed? For now, the industry has no standard answer.

Through this IPO, Anthropic is testing the waters on behalf of the entire industry. Where this experiment ultimately leads is not only tied to this company's share-price story, but will also define the posture in which future giant AI companies should enter the public capital markets.

This article comes from the WeChat public account "Wall Street Tech Eye"; author: Du Jie; editor: cc Sun Congying; published by 36Kr with authorization.