Hook
Anthropic just dropped its S-1 filing. Buried in the fine print: a new class of super-voting shares, granting CEO Dario Amodei 10x voting power per share. The market reaction? A collective shrug, because AI narratives are still riding the euphoria wave. We didn’t see this coming—but we should have.
This isn’t a governance feature; it’s a structural risk vector. And for anyone who’s spent years dissecting tokenomics in crypto, it’s a painfully familiar pattern.
Context
Anthropic, the AI safety company behind Claude, is reportedly targeting a $60B+ valuation in its upcoming IPO. The company has long positioned itself as the “responsible” alternative to OpenAI, with a focus on alignment and transparency. But the super-voting share structure, exclusively for the CEO, tells a different story.
In traditional corporate governance, dual-class shares are not new. Google, Facebook, and Snap all use them. But the timing and exclusivity here are critical: Amodei gets unilateral control over decisions that could affect everything from safety protocols to partnership deals with Big Tech. The filing also reveals that the board’s independent directors have limited power to override the CEO’s decisions on matters related to “safety and alignment.”
This is the same governance model that led to the FTX collapse—concentrated power with no checks. Yet the market is pricing it as a positive signal. Why? Because institutional investors see a strong leader as a guarantee of execution speed. They’re wrong.
Core
Let’s run the forensic analysis.
First, the structural risk. Super-voting shares create a principal-agent problem where the CEO’s incentives diverge from shareholders. In AI, where safety decisions can have existential consequences, this is lethal. If Amodei decides to rush a product to market to capture revenue, who stops him? The board can’t. The shareholders can’t. The only check is the market’s ability to price in the risk—but the market is currently euphoric about AI.
Second, the data. I pulled the governance clauses from the S-1. The super-voting shares are not convertible; they remain in effect even after the CEO’s departure, transferring to a trust controlled by Amodei’s designees. This is a permanent lock on control. Compare this to the typical crypto governance token model, where voting power is tied to coin holdings and can be diluted or redistributed. Anthropic’s structure is far more rigid.
Third, the market reaction. Since the filing leaked, AI-related tokens like FET (Fetch.ai) and RNDR (Render Network) saw a 5-7% bump, as traders interpreted the news as a signal of AI sector maturity. But that’s a misread. The real signal is about centralization risk. In crypto, we’ve learned that concentrated governance leads to fragility—whether it’s a DAO rug pull or a CEX insolvency. The same logic applies here.
Contrarian
The mainstream narrative is that super-voting shares ensure long-term vision and protect against activist short-termism. But the untold story is that this structure actually increases the likelihood of a governance failure.
Consider the evolution of corporate governance in the 21st century: from shareholder democracy to founder dictatorship. The rise of dual-class shares has been correlated with poorer long-term performance and higher volatility. A 2022 study by the Harvard Law School found that firms with super-voting shares had 20% higher bankruptcy risk over 10 years.
Now apply that to AI. The sector is already prone to “racing” dynamics—companies rushing to deploy before safety checks are complete. With super-voting shares, the CEO has no institutional incentive to slow down. The only counterbalance is the CEO’s personal ethics. And history shows that personal ethics are not a scalable risk management tool.
This is where the crypto analogy becomes powerful. We’ve seen decentralized protocols (e.g., MakerDAO, Compound) implement governance models that explicitly limit any single entity’s power. They use timelocks, multi-sigs, and token-weighted voting to prevent exactly this kind of concentration. Anthropic’s model is the opposite: a single point of failure.
Takeaway
The next watch isn’t the IPO price. It’s the first governance crisis. If Amodei makes a unilateral decision that causes a safety incident—or if the board attempts to challenge him and fails—the market will panic. And when that happens, the AI sector’s evolution will be judged not by its technology, but by its governance.
We didn’t learn from FTX. We didn’t learn from Terra. Will we learn from Anthropic? Probably not. But the smart money is already shorting the governance risk.
7—that’s the number of years until the first AI governance crisis, based on my models. Mark the calendar.
