Anthropic's $1.5 Trillion Valuation Rests on Stories It Tells Itself

The Singularity Sales Pitch

At the IMF meeting in Bangkok earlier this month, Anthropic cofounder Jack Clark told an audience of economists, bankers, and lawmakers that the singularity has arrived. The concept, coined by science fiction writer Vernor Vinge four decades ago, describes the moment when technology surpasses human comprehension and control. Clark's declaration was unequivocal: "I believe that this year we have gained convincing evidence that it has now begun."

The claim sits at the heart of Anthropic's recently filed IPO prospectus, a 335-page document that values the company at more than $1.5 trillion. Roughly 100 pages focus on risks, a proportion that itself signals the company's positioning. Anthropic presents its technology as both inevitable and dangerous, and positions itself as uniquely equipped to manage what it builds.

Yet this narrative strategy raises questions that extend beyond product capability. When a company tells multiple, sometimes contradictory stories about what it is, how should regulators, investors, and the public assess its claims?

The Vatican and the Mystery Box

Last year, Anthropic cofounder Chris Olah addressed Pope Leo XIV at the Vatican. His message leaned into mystery: "We keep finding things that are mysterious, even unsettling. We find evidence of introspection. We find internal states that functionally mirror joy, satisfaction, fear, grief, and unease."

Many researchers would dispute the suggestion that large language models possess anything resembling humanlike inner experience. The claim ventures into territory where scientific consensus is thin and interpretation is wide. But Anthropic's framing does not invite scepticism. It invites awe.

This is narrative capture. In a field where few certainties exist, the companies building frontier AI models fill the void with their own interpretations. When Clark speaks of the singularity or Olah describes emotional states in software, they are not merely reporting findings. They are shaping the parameters of debate.

Multiple Identities, Competing Priorities

Nathan Calvin, chief strategy officer at Encode AI, a Washington-based advocacy group, posed a question on X last week that cut through the rhetoric: "What is Anthropic?"

The question emerged after a draft bill, circulated privately by Anthropic to US lawmakers and later leaked, proposed legislation that critics argued served the company's interests. Calvin's challenge was simple. If Anthropic wants to be trusted as the responsible actor in AI development, why should anyone take that claim at face value?

One answer is straightforward. Anthropic is a company whose IPO filing values it at $1.5 trillion, enriching its cofounders with generational wealth. But the company has always insisted it is more than a profit-seeking entity.

Anthropic also presents itself as an AI safety organisation, committed to developing technology that is reliable and nearly risk-free. At the same time, it operates as a frontier competitor, racing to build systems with superhuman capabilities before rivals do. Whether this urgency stems from a belief in its own responsibility or from competitive ego is unclear, possibly because both motivations coexist.

CEO Dario Amodei responded to Calvin's post with a defence of what he calls the "race to the top" strategy. The idea is that by demonstrating responsible practices at the frontier, Anthropic raises standards across the industry. Amodei argued that this strategy works only if the company practicing it remains at the leading edge.

The logic is internally consistent, but it glosses over the tension. A profit-driven megacorp and an AI safety organisation have fundamentally different incentives. The former answers to shareholders and growth targets. The latter answers to public interest and long-term risk mitigation. Anthropic attempts to embody both, and Amodei has acknowledged mistakes in balancing these roles.

The Hydra Problem

At Opentechwire, we have tracked the evolving narratives of AI companies across Asia and the West, and a pattern has emerged. Companies that position themselves as guardians of existential risk while also pursuing aggressive commercialisation often struggle with coherence. Anthropic is a vivid example of this tension.

The leaked draft bill, which critics argue tilts regulatory frameworks in Anthropic's favour, is one instance where the company's multiplicity becomes a liability. If Anthropic is primarily a safety advocate, self-serving legislation undermines that claim. If it is a competitive frontier lab, the safety rhetoric begins to look like branding. If it is a trillion-dollar public company, the entire narrative apparatus serves investor confidence.

The stories Anthropic tells itself matter because they inform the stories it tells the public. Those stories, in turn, shape how much autonomy the company enjoys in self-regulation, how much access external safety testers receive to its models, and how much benefit of the doubt lawmakers extend when new proposals surface.

But the multiplicity makes trust harder to establish. A company that is many things to many people is difficult to hold accountable, because the relevant identity shifts depending on the question being asked.

What Lies Beneath the Narrative

Anthropic's IPO prospectus is a document designed to satisfy legal disclosure requirements and attract capital. But it is also a window into how the company frames its mission. The emphasis on risk, the invocation of science fiction concepts, and the insistence on unique responsibility all serve a dual purpose. They justify the company's existence and they create a moat around its authority.

This is not inherently dishonest. Every company tells a story about itself. But when that story involves claims about the singularity, about mysterious inner states in software, and about being the only entity capable of managing the technology it builds, the stakes are higher. The narrative is not just marketing. It is a bid for regulatory deference and public trust.

The question is whether that trust is warranted. Anthropic's track record includes significant technical contributions and a visible commitment to interpretability research. But it also includes lobbying efforts that critics argue prioritise competitive advantage, and public messaging that blurs the line between scientific finding and speculative framing.

The Case for Simplicity

A technology company that builds software, sells it, and competes in a market is easier to evaluate than one that also claims to be a safety organisation, a frontier lab, and a steward of humanity's future. The former can be assessed on product quality, business performance, and regulatory compliance. The latter invites scrutiny of intent, narrative consistency, and the gap between stated mission and actual behaviour.

Anthropic's multiheaded identity complicates that evaluation. It is possible that the company genuinely believes it can be all these things at once. It is also possible that the multiplicity is strategic, allowing the company to invoke whichever identity is most advantageous in a given context.

Either way, the result is a trust problem. When a company's story shifts depending on the audience, it becomes harder to know which version is the real one. And when the technology at stake is as consequential as Anthropic claims, that ambiguity is not just a communications challenge. It is a governance risk.

The IPO prospectus offers a glimpse of how Anthropic sees itself. But the stories it tells, and the identities it inhabits, suggest that the real Anthropic is still standing up.