Five dimensions · Six tools · A complete narrative framework

Author and source: Wu Zihan, Narrative as Strategy

Preface

This is a company founded in 2021 that reached a valuation of nearly $900 billion by May 2026.

It doesn’t have the 900 million weekly active users of ChatGPT. It doesn’t have the trillion-dollar infrastructure of Google. It doesn’t have the consumer brand recognition of OpenAI. Yet, in just 14 months, it grew its annual revenue from $1 billion to $30 billion—a growth rate that surpassed what Salesforce achieved over 20 years.

This article does not merely examine how effective Anthropic’s narrative is; more importantly, it seeks to understand how its narrative system is constructed—each layer—toward itself, the team, customers, investors, and government—employs distinct language, yet all converge on the same core.

Let’s start with something that’s often misunderstood: Product and technology are the 1, and narrative is the 0 that follows.

Claude 4 is among the world’s top tier in reasoning, long-context processing, and programming. Claude Code reached in nine months the scale that Cursor took two years to achieve, generating over $2.5 billion in annualized revenue by February 2026, with 4% of all public GitHub commits worldwide completed by Claude Code. Salesforce took 20 years to reach $30 billion in revenue; Anthropic achieved this in less than three years. The technology is real, the results are real. Only when the narrative is built on this truth does it hold power.

Anthropic's narrative is built upon an extremely rare contradiction:

We are creating something that could destroy humanity, and that is precisely why we must be the ones to create it.

Most companies operate on the narrative that “our product is great.” Anthropic’s narrative is “this is dangerous, so it must be done by the most cautious people.” This is a complete reversal of the narrative logic.

The statement from Anthropic’s founding manifesto remains on their website today: “Our work is a long-term bet that AI safety problems are tractable, and that the theoretical and practical tools we're building today will become critical in a world with broadly capable AI systems.”

我们是一家相信人工智能安全问题能够得到解决的公司。我们不是在努力规避风险，而是在直面风险，因为如果没有我们，风险将会更大。

This logic has a subtle narrative effect: it turns all competitors into “the less secure one.” OpenAI is pursuing capability, Google is pursuing scale, and Meta is pursuing open-source speed. Only Anthropic is pursuing “doing this right.”

Replace Anthropic’s name with any competitor, and this statement doesn’t hold. No other company has a founding story centered on prioritizing safety or choosing a harder path from the very beginning.

This is the unique anchor point of Anthropic's narrative.

Anthropic · Four-layer narrative structure

Layer 1 · Civilizational Narrative

AI may be the most dangerous technology humanity has ever created, and we must be present.

Layer 2 · Identity Narrative

The company that gave up everything to do the right thing

Layer 3 · Product Narrative

Claude: An AI with values, not just an AI with capabilities

Layer 4 · Business Narrative · "Responsible development is becoming the best business"

Zero

Anthropic’s narrative is a five-layered concentric circle, radiating outward from the founders’ core beliefs to the team, customers, capital, and government—each layer speaks a different language but shares the same central message.

At the deepest level of the narrative is the phrase the founder said to themselves—that phrase is the root of all external narratives.

Dario Amodei has an origin story that the media has rarely explored. His father developed a rare disease while he was still a graduate student. At the time, the disease was nearly untreatable. Several years after his father’s death, a medical breakthrough occurred, turning the disease into one that could be cured. He missed those few years.

This experience gave him a completely different sense of “speed” compared to most tech entrepreneurs. The pace of technological advancement is not just a business issue—it’s a matter of life and death. A step too slow means someone who could have been saved wasn’t.

This also explains why he is a peculiar contradiction. He is the CEO described by The New York Times as “the epicenter of the AI doomsday frenzy,” yet he is also one of the most ardent believers in AI’s capabilities—believing that AI will compress centuries of scientific progress into the next decade, eliminate most cancers, and resolve the mental health crisis.

He speaks of risk because risk is the only obstacle between us and a better future.

In October 2024, he published "Machines of Loving Grace," a 15,000-word piece with zero promotional costs. The opening sentence read: "Because I often talk about the risks of AI, many assume I'm a pessimist. I'm not."

This article sparked widespread discussion across the AI internet, with The New Yorker, The Atlantic, and Wired all following up with commentary. With zero promotional costs, it generated a brand effect worth hundreds of millions of dollars. This was no accident—it was a chemical reaction between genuine ideas and genuine concerns.

Anthropic has another co-founder: Daniela Amodei, Dario’s sister and the company’s president. With a background in English literature, she previously led business operations at Stripe. She translates Dario’s technological vision into business language that everyone can understand. Together, they form a perfect storytelling duo: one defines “what we are doing,” and the other defines “what this means for the world.”

Dario Amodei · Machines of Loving Grace · 2024

The main reason I worry about risk is that risk is the only obstacle between us and a fundamentally positive future.

Two: The team's PBC structure turns narrative into law.

Most companies display their mission and vision on the wall. Anthropic has embedded its mission into the company’s legal structure.

Anthropic is a Public Benefit Corporation (PBC) incorporated in Delaware. This is not a title, but a legal structure. The board of a PBC is legally required to consider both commercial interests and the “public benefit” when making decisions—in Anthropic’s definition, this public benefit is: “the responsible development of AI for the long-term benefit of humanity.”

Anthropic has also established a more unique entity: the Long-Term Benefit Trust (LTBT). This trust holds special shares in the company and has the authority to elect or replace a majority of the board members under specific conditions. The trustees of the trust are neither investors, employees, nor shareholders, but independent individuals from the fields of AI safety, national security, and public policy. Current trustees include the president of the Carnegie Endowment for International Peace and the CEO of the Center for a New American Security.

This structure has an exceptionally clever narrative effect: it prevents investors from undermining the mission. Amazon invested $8 billion, Google invested billions, and Microsoft and NVIDIA have also joined. But these investors cannot use their capital control to force Anthropic to abandon its safety principles, because the independent trustees of LTBT hold ultimate board control.

When a new employee joins Anthropic, they are not joining an ordinary tech company. They are joining an organization legally committed to being accountable for humanity’s long-term well-being. This sense of identity is far more compelling than stock options.

This structure also has a clever business advantage: Amazon and Google can both hold equity in Anthropic and maintain cloud computing partnerships, yet neither can exclude the other through capital control. The LTBT structure provides institutional safeguards for Anthropic’s neutrality.

Three: The Most Trustworthy AI for Customers

Anthropic never seeks the most users. It seeks the users who trust it the most.

This choice is thoughtfully crafted in narrative terms: OpenAI emphasizes “most users,” Google highlights “greatest capability,” and Anthropic focuses on “most trustworthy in high-responsibility scenarios.”

What is a high-responsibility scenario? Financial, legal, medical, and complex programming contexts. Users in these scenarios don’t care who has more users—they care who makes fewer mistakes, who maintains more consistent principles, and who is most reliable in the most challenging tasks.

Anthropic's customer narratives fall into three lines.

Rule 1: Claude is an AI with values, not just an AI with capabilities. Claude’s “Model Spec” is a public document that outlines the role Claude is designed to fulfill—honest, helpful, but not blindly obedient. Claude says “no.” Claude upholds its principles even when no one is watching. This is a completely different product narrative from that of an “omnipotent assistant.”

Article 2: Enterprise-grade depth, not consumer-grade breadth. Anthropic has over 300,000 enterprise customers, who account for approximately 80% of its revenue. The Series G announcement in February 2026 revealed that the number of customers spending over $100,000 annually increased sevenfold over the past year; the number of customers spending over $1 million annually rose from just a dozen or so one or two years ago to more than 500. Eight of the Fortune 10 companies are using Claude. These figures are firsthand disclosures, cross-verified through GIC’s official press releases and Anthropic’s official website.

Article 3: Claude Code redefines the developer narrative. Claude Code’s story is not “better code completion”—it’s “your AI engineer living inside your terminal.” This single phrase elevates the product from a tool to a partner. It reached $1 billion in annualized revenue in six months and surpassed $2.5 billion in nine months. Four percent of all public GitHub commits worldwide were completed by Claude Code—a figure verified by two sources: Anthropic’s official announcement and an independent report by VentureBeat.

Four: Capital · "The Safest Bet"

Anthropic presents a unique story to the capital markets: a company where you can believe in both its commercial returns and its long-term stability.

The reason is structural, not narrative.

Amazon invested $8 billion. Google invested billions and established a deep cloud computing partnership with Anthropic, prioritizing TPU allocation. Microsoft and NVIDIA also joined in Series G. GIC (Singapore’s Government Investment Corporation) led two funding rounds (Series F and Series G). Other investors include Qatar Investment Authority, Temasek, funds under BlackRock, Fidelity, Sequoia, Founders Fund, General Catalyst, and JPMorgan Chase—nearly every top institutional investor in the world is on this list.

The Series G round on February 12, 2026, was the second-largest private funding round in tech history (after OpenAI’s $40 billion), co-led by GIC and Coatue, raising $30 billion and achieving a post-money valuation of $380 billion. These figures are sourced directly from Anthropic’s official announcement and GIC’s official press release, which are identical.

Fortune magazine reported an interesting figure: Amazon’s $8 billion investment in Anthropic had a book value exceeding $70 billion by early 2026. A significant portion of Google’s and Amazon’s Q1 2026 profits came from the unrealized gains on their Anthropic equity stakes, rather than from their core businesses.

This is Anthropic’s implicit leverage over capital narratives: its rising valuation directly enhances the financial statements of its largest shareholders. This creates a closed loop of aligned incentives—major capital providers are motivated to continue driving up Anthropic’s valuation, as each increase improves their own financial performance.

Five: Government — "Reasonable Voices on Regulation"

Anthropic's narrative regarding the government is the most complex, most easily misunderstood, and most nuanced layer of the entire discussion.

In July 2023, Dario personally testified before the U.S. Senate Judiciary Subcommittee on AI safety risks, the principles of Constitutional AI, and why AI companies need transparency standards. He became one of the most cited AI CEOs in congressional hearing history.

In 2025, as most tech company CEOs testified before Congress advocating “don’t regulate AI,” Dario wrote a counter-opinion in The New York Times: he supported “reasonable regulation” of AI and opposed the provision in Trump’s “Great America Act” that would ban states from regulating AI for 10 years—he called it “an overly blunt tool.”

He was also one of the earliest and most consistent AI company leaders to support export controls on AI chips to China, publicly lobbying the government to tighten restrictions.

According to OpenSecrets’ lobbying database—the most reliable source for U.S. lobbying expenditure data—Anthropic spent over $3.1 million on federal lobbying in 2025, a fourfold increase year-over-year and among the fastest-growing lobbying expenditures in the AI industry. This figure has been cross-verified.

This government narrative strategy places Anthropic in an exceptionally advantageous position: it is the company in the AI industry that regulators are most willing to engage with. While the entire industry says, “Don’t regulate us,” a company that says, “Please regulate us responsibly” naturally becomes the preferred partner in policy discussions. Influence is not gained through opposition, but through participation.

But this stance also carries risks. Trump administration’s AI and cryptocurrency czar, David Sacks, publicly criticized Anthropic for “running a fear-based regulatory capture strategy.” Within Silicon Valley, many voices believe Anthropic is leveraging safety narratives to suppress its competitors.

Dario’s response was consistent: “AI safety is an issue that should be about policy, not politics.” He does not fully align with either the Republican or Democratic party, maintaining his position as an independent voice of technical expertise. This is the most difficult yet most valuable position to hold within government narratives.

Six

Anthropic's narrative is not just a slogan; it's a sophisticated system. Breaking it down, there are six tools that founders can directly adopt.

Weapon One: Turn weaknesses into the core of your narrative. Anthropic is smaller than OpenAI and has fewer resources than Google. Most companies explain away or defend their weaknesses. Anthropic turns them into strengths: “We chose to do what’s right, not what’s fast. Slowness is our proof of caution.” Apple did the same—“Products are expensive” became “We refuse to compromise.”

Weapon Two: The Sense of Sacrifice from the Origin Story. Dario and Daniela left OpenAI, where they could have stayed at the world’s most advanced AI lab. In this narrative, sacrifice is the strongest proof of conviction. A commitment with cost is faith; a promise without cost is noise. This story is something competitors can never replicate.

Weapon Three: Define the Battlefield with Terminology. Constitutional AI shifts safe dialogue from a “checklist of rules” to a “system of values.” The Responsible Scaling Policy has been adopted by OpenAI, Google, and others. The Race to the Top has redefined the competitive framework as “who has higher safety standards.” When you compete within someone else’s vocabulary framework, you’re a follower. When the entire industry adopts your terminology, you become the center.

Weapon Four: Long-form Thought Output. In October 2024, "Machines of Loving Grace," a 15,000-word piece, went viral globally with zero advertising spend. Ads make you seen; ideas make you believed. Readers followed a single person’s thinking over 15,000 words, achieving a depth of understanding no hundred news articles combined could match.

Weapon Five: Prove with action, not claims. In 2022, Anthropic already had Claude but chose to delay its release because it was “not yet safe.” This decision came at a cost—allowing OpenAI to get ahead and losing first-mover advantage. Action with a cost is proof; words without cost are noise.

Weapon Six: Contradiction Creates a Multidimensional Narrative. “We are building the most dangerous thing possible—precisely because we must be the ones to build it.” Contradiction makes the audience pause and adds depth to the story. Straightforward narratives are forgotten immediately; those with tension and contradiction linger in memory.

Seven: The Narrative Gap · Honest Analysis

Every narrative has cracks. A good narrative analysis must identify them.

Gap One: The Tension Between Rapid Growth and “Safety First.” In early 2026, Anthropic modified parts of its Responsible Scaling Policy (RSP), relaxing the safety thresholds for ASL-4. Critics viewed this as a sign that commercial pressures were undermining its safety commitments. Anthropic responded that risks above ASL-4 exceed the capacity of any single company and require industry-wide collaboration. This gap is real. The core narrative is “caution”—each adjustment to safety commitments in the name of competitive speed is subject to intense scrutiny.

Gap Two: The Issue of Military Cooperation. In 2026, the U.S. Department of Defense classified Anthropic as a “supply chain risk,” prohibiting federal contractors from collaborating with it, following Anthropic’s refusal to engage in an AI military application project with the Pentagon. While this decision aligned with Anthropic’s safety narrative, it placed pressure on Anthropic at the governmental level regarding “national security cooperation.” As your competitors move aggressively into military AI, your caution becomes a political liability.

Gap three: The narrative has geographic boundaries. Anthropic’s narrative is nearly unmatched among tech elites in the English-speaking world. However, its presence is far weaker among non-English-speaking populations and among general consumers compared to ChatGPT. This is the cost of a strategic choice, not a mistake—but it limits the geographic reach of its narrative.

These three cracks are not reasons to reject Anthropic—they are the correct way to understand it. The vitality of a narrative depends on whether ongoing actions continue to prove it. As long as costly actions persist, the cracks will not become fatal.

Eight Insights for Founders

After studying Anthropic's narrative framework, five conclusions are worth considering by every technology founder.

First, the underdog narrative must be stronger. When Anthropic was founded, OpenAI was the absolute industry leader, Google had unlimited resources, and Microsoft had the world’s largest enterprise customer network. Anthropic had no advantage in any dimension. Its only advantage was its narrative—the story that “we are the company doing the right thing.” Precisely because it lacked resource advantages, its narrative had to be even stronger to attract equal or even greater attention and resources.

Second, the narrative must be genuinely connected to the origin story. Dario’s father’s story, the story of leaving OpenAI—these were not fabricated for public relations. These are real events that happened to him and shaped his authentic worldview. Anthropic’s narrative holds power because it grows from real people and real experiences, not from a business plan.

Third, defining terminology is defining the battlefield. Constitutional AI, RSP, Race to the Top—these terms ensure Anthropic always remains on the offensive in the conversation around AI safety. Do Chinese AI company founders have their own coined terms to define their own battlefield? If not, they are competing within frameworks defined by others.

Fourth, ideas are the cheapest and most enduring form of promotion. Dario’s two long-form articles generated brand value worth hundreds of millions of dollars without spending a single dollar on advertising. He didn’t write ads—he wrote genuine thinking. Thinking has spreadability; ads do not. Authentic thoughts written directly by founders carry more power than anything a PR firm can achieve.

Fifth, contradiction is the strongest structure in storytelling. “We are building something dangerous, and that’s precisely why we must be the ones building it”—this contradictory structure is a hundred times more powerful than “Our product is great.” Truly compelling narratives often contain a contradiction that makes people pause: We are doing this to prevent that outcome. Find the contradiction in your story, and your narrative will become much richer.

The narrative of the weak must become stronger; this is the most important lever.

② The narrative must grow organically from the true original story.

③ Defining terms is defining the battlefield.

④ Ideas are the cheapest and most enduring form of propagation.

⑤ Conflict is the strongest narrative structure.

Nine

In 2021, Dario left OpenAI alone with eleven others, with nothing.

All he had was a judgment: that if this thing wasn’t done correctly, it would be disastrous. And a belief: that doing it correctly was possible.

Five years later, this judgment and belief were built into a company nearing a trillion dollars in value.

Anthropic’s story shows that storytelling is not something you do after achieving product success—it’s something you do from day one. It’s not packaging; it’s direction. It’s not a slogan; it’s the contagious form of your belief.

Narrative is one of the strongest weapons of the weak.