SPECIAL REPORT: ARTIFICIAL INTELLIGENCE

PART 1

Editor’s Note: Because Anthropic’s S-1 registration statement has been filed confidentially with the United States Securities and Exchange Commission (the “SEC”) and is not yet public, all financial figures, ownership percentages, and projections cited in this report are drawn from widely published reporting by Bloomberg, CNBC, The Wall Street Journal, TechCrunch, and other sources; every figure should be independently verified and cross-checked against the S-1 once it becomes available.

Part One: How We Got Here — The Long Arc of Artificial Intelligence

The Revolution in Three Acts

To understand what Anthropic is, you must first understand what it is trying to finish. The story of artificial intelligence is, in the truest sense, the story of the twentieth and twenty-first centuries compressed into a single technological throughline — a seventy-five-year obsession with building a machine that thinks.

Act One began in 1950, when British mathematician Alan Turing published his landmark paper “Computing Machinery and Intelligence,” asking the question that would haunt an entire field: “Can machines think?” Turing proposed what he called the “imitation game” — now known as the Turing Test — as a practical measure of machine intelligence. It was a theoretical provocation more than a blueprint, but it planted a seed that would not stop growing.

What the history books cannot fully capture, and what the 2014 film The Imitation Game — starring Benedict Cumberbatch in a portrayal that earned him an Academy Award nomination — brought to a wider audience, is the full weight of what Turing contributed before he ever posed that philosophical question. During the Second World War, Turing was the central intellectual force behind Britain’s effort at Bletchley Park to crack the Enigma cipher — the fiendishly complex encryption system used by Nazi Germany to coordinate its military communications across every theater of the war. His design of the electromechanical “Bombe,” building on earlier Polish work, broke the code at industrial scale and gave the Allies the ability to read German naval dispatches in near real-time. Historians have estimated that Turing’s work shortened the war in Europe by two to four years and saved upwards of fourteen million lives. It was, by any measure, the most consequential act of applied mathematics in human history. And yet the British government, in a moral failure that defies comprehension, rewarded Turing’s service to civilization by prosecuting him in 1952 for “gross indecency” — the crime of being a gay man in a country that had criminalized homosexuality. Subjected to chemical castration as a condition of avoiding prison, Turing died in 1954 at the age of forty-one, officially ruled a suicide by cyanide poisoning, though the full circumstances of his death remain disputed. He was not granted a formal royal pardon until 2013 — nearly sixty years after his death — and the broader “Alan Turing Law,” which posthumously pardoned thousands of men convicted under the same statutes, was not enacted until 2017. The man who may have done more than any other individual to save Western civilization from fascism was destroyed by the civilization he saved. In one of those truly ironic and strange twists of history, it is not a stretch to say that the same man who was instrumental in saving the world from Nazi control also laid the intellectual foundation — in computation, logic, and the mathematical theory of mind — for the applied reasoning and tools that are now shaping what may be the most important technological innovation in the entire course of human history.

Act Two arrived in 2012, in what the field now recognizes as the moment everything changed. A team led by Geoffrey Hinton at the University of Toronto entered a computer vision competition called ImageNet and won by a margin so large — cutting the error rate nearly in half — that it shocked the entire research community. Their tool was a deep neural network: layers upon layers of mathematical functions loosely inspired by the human brain, trained on massive quantities of data using graphics processors originally designed for video games. The result was not programmed intelligence. It was learned intelligence. The machine had not been told what a cat looked like. It had seen ten million cats and figured it out.

That moment triggered a gold rush. Google, Facebook, Microsoft, and Amazon poured billions into deep learning research. The talent wars began. Salaries for AI PhDs reached levels not seen since the dot-com boom. And the models kept getting better — faster than almost anyone predicted.

Act Three opened in 2017, when a team of researchers at Google published a paper titled “Attention Is All You Need.” The paper introduced the transformer architecture — a new way of building neural networks that could process language by learning which parts of a sentence to pay attention to when generating the next word. It was, in retrospect, the most consequential computer science paper of the decade. The transformer made large language models possible. Large language models made ChatGPT possible. And ChatGPT, launched in November 2022 to one hundred million users in sixty days, made the present moment possible.

The OpenAI Exodus — and the Birth of Anthropic

By 2021, the center of gravity in AI had shifted to San Francisco, and specifically to a nonprofit-turned-capped-profit entity called OpenAI. Founded in 2015 by a group that included Elon Musk and Sam Altman, OpenAI had attracted some of the world’s most brilliant AI researchers and secured a multi-billion-dollar partnership with Microsoft. It was building, by consensus, the most powerful AI systems in the world.

Inside OpenAI, however, a fault line had been widening. A group of senior researchers had grown increasingly uneasy — not about the power of what was being built, but about the trajectory of the organization doing the building. Microsoft’s billions came with commercial expectations. The pressure to ship products, to generate revenue, to compete in a market that had been ignited by their own research was, in the view of some, pulling the organization away from its founding mission of ensuring that artificial general intelligence benefited all of humanity.

In the summer of 2021, Dario Amodei — OpenAI’s Vice President of Research, co-creator of GPT-2 and GPT-3, and co-inventor of Reinforcement Learning from Human Feedback — resigned. He was joined by his sister Daniela Amodei, then Vice President of Operations, and seven other senior researchers: Tom Brown, the lead author of the GPT-3 paper; Chris Olah, the pioneer of neural network interpretability; Jared Kaplan, co-author of the landmark scaling laws paper; Jack Clark, Policy Director and founder of the Import AI newsletter; and Sam McCandlish, a specialist in the theory of deep learning. It was, arguably, the most significant talent departure in the history of Silicon Valley — nine people who left not for money, but for a mission.

In October 2021, Anthropic was born.

For full biographical profiles of each founder, see Part 2 of our Special Report on Artificial Intelligence and the companion piece to this article: “The Seven Founders of General Wisdom.”

The Leadership: Dario and Daniela Amodei

Dario Amodei, 42, serves as CEO. He holds a PhD in biophysics from Princeton, co-invented Reinforcement Learning from Human Feedback, and led the development of GPT-2 and GPT-3 at OpenAI. He is the philosophical architect of Anthropic’s safety-first mission and the author of “Machines of Loving Grace,” a landmark essay on AI’s potential to transform human civilization. Time named him one of the world’s hundred most influential people in both 2025 and 2026. At Anthropic’s current $965 billion valuation, his equity stake is likely worth tens of billions of dollars.

Daniela Amodei, 38, serves as President. A former operations leader at Stripe and OpenAI, she runs Anthropic’s commercial, financial, and organizational machinery with an effectiveness that has driven the company from $500 million in annualized revenue in 2023 to a $47 billion run rate in 2026 — a trajectory with no precedent in the history of enterprise software. Her equity stake, comparable in structure to her brother’s, places her among the wealthiest women in American technology upon the IPO.

Constitutional AI — The Scientific Differentiator

Anthropic was not founded merely to build a better chatbot. It was founded on a specific thesis about how to build AI that remains beneficial as it becomes more powerful. That thesis — codified into a methodology called Constitutional AI — teaches models to critique and revise their own outputs against a written set of principles, rather than relying solely on human feedback. The model learns to reason about its own behavior, not just to pattern-match on what human raters approved of in the past. The practical result is an AI system that can generalize good behavior to novel situations — a property that becomes increasingly important as models grow more capable and their deployment contexts more diverse.

PART TWO: THE FILING

The Confidential S-1: History in a Document

On June 1, 2026, Anthropic announced that it had confidentially submitted a draft registration statement on Form S-1 to the SEC for a proposed initial public offering. The filing was confidential, as permitted under the JOBS Act — meaning the full document will not become public until the formal IPO roadshow begins. But its submission marked an inflection point not just for Anthropic, but for the entire AI industry and, by the company’s own reckoning, for human civilization.

For the first time, a leading frontier AI laboratory will be required to describe its business, its risks, its finances, and its governance in the black-and-white language of SEC disclosure. Every risk factor that Anthropic’s lawyers draft — every honest accounting of model unpredictability, safety failures, dependency on cloud infrastructure, regulatory exposure, and yes, existential risk from advanced AI systems — will become a matter of public record. There is no precedent for this. No company has ever had to price the possibility of building a technology that could harm humanity into an IPO prospectus.

The Valuation: Nearly a Trillion Dollars — And the Race No One Has Evern Run

Anthropic’s most recent private financing — a $65 billion Series G announced in late May 2026 — valued the company at approximately $965 billion on a post-money basis. That figure, if sustained through an IPO, would make Anthropic’s public debut the largest in history, exceeding Saudi Aramco’s $25.6 billion offering in 2019 and potentially raising more than $60 billion in new capital based on current banker estimates. To put that in context: $965 billion is larger than ExxonMobil, larger than JPMorgan Chase, larger than Walmart — a valuation that implies Anthropic will be one of the defining businesses of the next half-century, not merely a successful technology company but a foundational piece of global infrastructure.

The ownership structure reflects the company’s unusual history. Amazon, whose investment commitment of up to $25 billion represents one of the largest corporate bets on an AI company in history, holds an estimated 15 to 19 percent stake. Google, which has separately invested $40 billion and secured TPU compute capacity, owns approximately 14 percent. The founding team collectively retains a substantial equity position — with Dario and Daniela Amodei holding the largest individual positions, and co-founders Tom Brown, Chris Olah, Jared Kaplan, Jack Clark, and Sam McCandlish each estimated at roughly two to four percent. At a $965 billion valuation, even a two percent individual stake is worth approximately $19 billion. Other investors include Spark Capital, Salesforce, Nvidia, Microsoft, and the Saudi Arabian government’s Public Investment Fund.

The October 2026 window is widely cited by bankers. Goldman Sachs, JPMorgan, and Morgan Stanley are in early discussions for lead underwriter roles. Wilson Sonsini Goodrich & Rosati has been retained as legal counsel. The mechanics of the largest IPO in history are quietly being assembled — and Anthropic will not be alone. Three companies are each preparing to enter the public markets at valuations that could exceed one trillion dollars: SpaceX, OpenAI, and Anthropic itself. This has never happened before, not even once. SpaceX is expected to be first, with its offering anticipated before year end. Anthropic and OpenAI are racing to be second. The sequencing and pricing of these three offerings will be among the most closely watched dynamics in the financial markets through the close of 2026.

“Three companies could enter the public markets with valuations of at least $1 trillion — something that has never before even happened once. SpaceX is expected to be first. Anthropic and OpenAI are racing to be second.”

PART THREE: THE NUMBERS — FINANCIAL PORTRAIT OF A HYPERGROWTH MACHINE

Revenue: The 80x Growth Story

The most striking feature of Anthropic’s financial profile is not its absolute scale — it is the velocity. CEO Dario Amodei described the company’s first-quarter 2026 performance as “crazy 80x growth” on an annualized basis. That phrase, almost incomprehensible in the context of any established business, reflects a trajectory that has no modern precedent among companies of comparable starting scale.

The numbers tell the story with unusual clarity. At the end of 2023, Anthropic’s annualized revenue run rate was approximately $500 million. By year-end 2025, that figure had risen to approximately $9 billion. By February 2026, it had crossed $14 billion. By April 2026, $30 billion. By late May 2026, when the Series G closed, Anthropic’s run-rate revenue had surpassed $47 billion — more than doubling in roughly ninety days.

The Cash Burn: $1.2 Billion Per Month

The other side of this equation is equally dramatic. Anthropic is projected to lose approximately $14 billion in 2026 — a monthly cash consumption of approximately $1.2 billion. The core reason is compute. Training a frontier AI model requires thousands of specialized chips running for months. Inference — actually running those models to respond to user queries — requires a parallel infrastructure at massive scale. Anthropic is paying for both simultaneously, while also investing in the research required to build the next generation of models.

Anthropic has committed to approximately $80 billion in cloud infrastructure costs through 2029, primarily through its partnerships with Amazon Web Services and Google Cloud. This commitment is not a cost to be avoided — it is the price of playing at the frontier.

Path to Profitability: 2029

Company disclosures and investor materials project positive free cash flow in 2029. Gross margins are already moving in the right direction — from 38% at the inference level one year ago to 70% today, with projections of approximately 77% by 2028 as training costs fall and scale economics take hold. The question investors will be asked to answer is whether they trust that trajectory enough to fund the losses in the meantime.

The IPO Raise: Potentially $60 Billion

Bankers familiar with the process expect the IPO to raise more than $60 billion in new capital. If accurate, that would be three times the previous record for a technology IPO (Alibaba, $25 billion, 2014). The proceeds would fund continued model development, infrastructure build-out, talent acquisition, and global enterprise expansion.

PART FOUR: THE CUSTOMER BASE — 300,000 BUSINESSES AND EIGHT OF THE FORTUNE TEN

Enterprise Dominance

Anthropic is, first and foremost, a business-to-business company. Approximately 80 percent of its revenue comes from enterprise and API customers — companies that have integrated Claude into their products, workflows, and operations. The breadth of the customer base is striking: more than 300,000 business customers, eight of the Fortune 10 using Claude, and more than 1,000 companies spending at least $1 million annually on Anthropic’s products.

Large Account Growth: 7x Year-Over-Year

The most commercially significant metric in Anthropic’s disclosed financials may be this: large enterprise accounts grew nearly seven times year-over-year. Named customers include Netflix, Spotify, KPMG, L’Oreal, and Salesforce. These are not experimental deployments. They are production integrations embedded in software stacks, customer service operations, legal workflows, financial analysis processes, and software development pipelines. The switching costs embedded in those integrations are among the most important facts in the prospectus.

Claude Code: The Breakout Product

Among Anthropic’s product lines, Claude Code has emerged as the breakout hit of 2026. An AI-powered software development tool that can write, review, debug, and document code at a level many engineers describe as genuinely transformative, Claude Code exceeded a $2.5 billion annualized run rate within months of its enhanced release. Enterprise use represents more than half of Claude Code’s revenue — a signal that professional software development teams have adopted it as a core productivity tool, not an experiment.

Amazon Bedrock: The Distribution Channel

As of April 2026, more than 100,000 business customers run Claude through Amazon Bedrock, AWS’s managed AI platform — reaching Anthropic’s models through AWS’s existing enterprise relationships without requiring Anthropic’s direct sales team to close those accounts. The relationship is symbiotic: Amazon has committed up to $25 billion in total investment in Anthropic, securing preferred compute capacity; AWS’s AI credibility depends in part on Claude’s continued performance leadership.

Consumer Growth and 2-3 Year Projections

Claude.ai paid subscriptions have quadrupled since the beginning of 2026, establishing a growing consumer base that serves as both a feedback loop for model improvement and a brand asset that drives enterprise adoption. Based on current trajectory, analysts project Anthropic could serve more than one million enterprise API customers by 2028, with revenue per account rising substantially as agentic use cases multiply.

PART FIVE: THE FUTURE OF THE BUSINESS — AND WHAT IT MEANS FOR HUMANITY

The Agentic Pivot: Beyond the Chatbot

The next phase of Anthropic’s development is not about better chatbots. It is about agents: AI systems that can autonomously plan, execute multi-step tasks, browse the web, write and run code, manage files, and complete extended workflows with minimal human intervention. The difference between a conversational AI assistant and an autonomous agent is roughly the difference between a consultant who gives advice and an employee who does the work. The addressable market implications are transformational.

“Machines of Loving Grace” — Dario’s Vision for Humanity

In October 2024, Dario Amodei published a fifty-page essay titled “Machines of Loving Grace: How AI Could Transform the World.” His central argument is that most people are dramatically underestimating the positive potential of advanced AI — not because the risks are not real, but because the upside is so large and so unfamiliar that human intuition fails to process it accurately. He suggests that within five to ten years of the emergence of genuinely powerful AI, the world could look radically different — if the technology is developed carefully and deployed responsibly.

The Medical Revolution: Compressing a Century of Progress

Biology is, in Amodei’s framing, the domain where advanced AI will have its most profound impact. He describes a future in which AI systems act as autonomous research collaborators — generating hypotheses, designing experiments, interpreting results, and driving scientific discovery at a pace that dwarfs anything human researchers can achieve working alone. The specific prediction: AI could compress fifty to one hundred years of biological and medical research progress into five to ten years. Cures for cancers that have resisted treatment for decades. Effective therapies for Alzheimer’s. Healthy human lifespans extended meaningfully beyond current limits.

AI as Autonomous Scientist and Economic Transformer

Beyond medicine, Amodei envisions AI systems operating as autonomous researchers across every field of human knowledge simultaneously — physics, chemistry, materials science, climate science — accelerating progress in every discipline at once. Anthropic’s investor materials describe AI potentially contributing $10 to $20 trillion in annual economic value globally by 2030, with Anthropic aiming to become the infrastructure layer for that economy in the way AWS became the infrastructure layer for cloud computing.

The Geopolitical Dimension

Amodei has been explicit: the race to artificial general intelligence is geopolitically decisive. A world in which AGI is first achieved by a company committed to democratic values and safety-first development looks categorically different from a world in which it is achieved by a state actor operating under authoritarian constraints. The IPO, in this framing, is partly a war chest — capital to fund the continued development of frontier models that Anthropic believes must be led by safety-conscious actors.

The Shift to Outcome-Based Pricing

One of the most underappreciated elements of Anthropic’s long-term strategy is the potential migration from compute-based API pricing to outcome-based pricing: charging for what the AI actually accomplishes rather than the compute consumed. A legal research task completed. A software feature built. A customer support issue resolved. This model would structurally transform Anthropic’s revenue economics — replacing a ceiling bounded by compute capacity with one bounded by the value created, a ceiling orders of magnitude higher.

PART SIX: LEGAL & REGULATORY RISKS — WHAT THE PROSPECTUS WILL HAVE TO SAY

The PBC Governance Structure: Unprecedented and Untested

Anthropic is organized as a Public Benefit Corporation with a Long-Term Benefit Trust designed to ensure the company cannot be fully captured by profit motives at the expense of its safety mission. For public market investors accustomed to shareholder primacy, these structures create genuine uncertainty. What rights do common shareholders actually have? Under what circumstances can the Trust override shareholder preferences? These questions have never been answered for a company of this size, and the S-1 will be the first opportunity to see the answers in legally binding language.

The FTC and Antitrust Scrutiny

The Federal Trade Commission launched a formal inquiry in January 2024 into the investments that Google, Amazon, and Microsoft had made in frontier AI laboratories, specifically naming Anthropic. The concern is structural: if the same companies that serve as Anthropic’s cloud infrastructure providers also hold significant equity stakes, market competition may be compromised in ways not immediately visible. A former senior FTC official told Reuters in 2026 that the agency is “almost certain to revisit” its scrutiny of these arrangements.

The Amazon-Google Dependency: Concentration Risk

Amazon has committed up to $25 billion in total investment and secured preferred cloud capacity. Google has invested $40 billion and secured agreements for 3.5 gigawatts of TPU compute capacity beginning in 2027. Both are simultaneously Anthropic’s infrastructure providers, its investors, and through their own AI products, its competitors. This tripartite relationship is unlike anything previously disclosed in a major technology IPO, and the risk section of the S-1 will need to address it with unusual candor.

National Security: The Autonomous Weapons Dispute

In a development that has received less attention than it deserves, the U.S. government formally designated Anthropic as a supply chain risk following the company's refusal to allow its AI systems to be used for fully autonomous weapons systems and mass surveillance of U.S. citizens. Anthropic's Acceptable Use Policy explicitly prohibits these applications, a position consistent with its safety mission and the norms of responsible AI development as understood by most of the research community. U.S. defense officials, however, have argued that these restrictions compromise the military's ability to operate effectively in domains where AI will be decisive. A preliminary injunction is currently blocking enforcement of the designation, but the underlying case is live, and its resolution could have material implications for Anthropic's ability to serve government customers.

EU AI Act and Global Regulatory Compliance

The European Union's AI Act classifies Anthropic as a provider of general-purpose AI with systemic risk, triggering mandatory transparency requirements, safety testing obligations, incident reporting requirements, and compliance audits. Non-compliance carries fines of up to three percent of global annual revenue — which at Anthropic's current scale would represent more than $1 billion in potential exposure. Beyond the EU, Anthropic faces a patchwork of emerging AI regulations in the United Kingdom, Canada, China, India, and a growing number of U.S. states — most significantly California, where the company is headquartered and where legislators have passed some of the most aggressive AI safety legislation in the world. The compliance burden of operating at global scale under fragmented regulatory regimes will be a recurring theme in Anthropic's S-1 risk factors, and it represents a genuine operational challenge that grows more complex with every new market the company enters.

U.S. Export Controls: Revenue Constraints

The Commerce Department's AI chip export controls — designed to prevent advanced AI compute from reaching adversaries — have the practical effect of restricting where Anthropic can deploy its technology and which customers it can serve. Large potential markets, including China and parts of the Middle East and Southeast Asia, are subject to restrictions that limit Anthropic's commercial reach. As those controls evolve in response to ongoing geopolitical developments, the company's addressable market may expand or contract in ways that are difficult to forecast and impossible to control.

Model Safety Incident Risk: The Public Company Problem

Perhaps the most unusual risk factor that will appear in Anthropic’s S-1 is also the most important: the risk that one of its AI models causes real-world harm in a high-profile way. Private companies can manage safety incidents — which are an inevitable reality for any organization deploying complex AI systems at scale — through quiet remediation, internal investigation, and selective disclosure. Public companies cannot. A single widely-reported incident in which Claude provided dangerous advice, assisted in a criminal act, or generated harmful content at scale could trigger both regulatory action and immediate, severe damage to Anthropic’s market capitalization. The pressure to ship faster, expand more broadly, and reduce the friction that safety measures introduce — pressure that public markets reliably generate — is precisely the pressure that Anthropic’s founders believe led them to leave OpenAI in the first place.

Intellectual Property and Training Data Litigation

Like every frontier AI laboratory, Anthropic faces ongoing legal exposure related to the use of copyrighted material in training data. A series of lawsuits by authors, publishers, and media organizations allege that training on copyrighted material without license constitutes infringement. The outcomes remain uncertain, but potential adverse results include mandatory licensing fees, damages awards, or court-ordered restrictions on training practices.

National Security, Export Controls, and Model Safety Risk

The U.S. government formally designated Anthropic a supply chain risk following the company’s refusal to allow its AI systems for fully autonomous weapons and mass surveillance. A preliminary injunction is blocking enforcement, but the case is live. U.S. Commerce Department export controls additionally restrict where Anthropic can deploy its technology, limiting commercial reach in markets including China and parts of the Middle East. And as a public company, a single high-profile safety failure — a Claude model causing real-world harm — could trigger both regulatory action and severe market cap damage in ways that private companies can more easily manage.

EU AI Act, Global Regulatory Patchwork, and the Unanswerable Disclosure

The EU AI Act classifies Anthropic as a provider of general-purpose AI with systemic risk, triggering mandatory safety testing, transparency requirements, and incident reporting, with fines up to three percent of global annual revenue for non-compliance. Beyond the EU, a growing patchwork of AI regulation across the U.S., UK, Canada, and India creates compliance complexity that grows with every new market entered. And at the center of the S-1 will be a disclosure without precedent in securities law: the company’s own acknowledgment, in legally operative language, that the technology it is building may pose an existential risk to humanity. No investment bank has ever had to underwrite that sentence.

PART SEVEN: PEOPLE, COMPENSATION, AND THE TALENT WARS

Headcount and the Research Talent Moat

Anthropic has grown from approximately 800 employees in early 2024 to an estimated 3,000 to 5,000 by mid-2026 — four to six times growth in two years. Its most defensible competitive asset may be the density of world-class AI safety and research talent on its payroll: alumni of DeepMind, Google Brain, Meta AI, OpenAI, and the top AI research programs at MIT, Stanford, Berkeley, and Carnegie Mellon. This talent concentration is difficult to replicate because the people who care most deeply about AI safety are disproportionately drawn to Anthropic specifically — a conviction that functions as a recruitment and retention mechanism that money alone cannot fully substitute.

What Anthropic Pays

Total compensation for software engineers in 2026 ranges from approximately $300,000 to $490,000. Senior research scientists earn between $350,000 and $550,000 in total annual compensation. The founding team holds aggregate equity worth well in excess of $100 billion at current valuations, with Dario and Daniela Amodei holding the largest individual stakes and co-founders estimated at roughly two to four percent each — translating to approximately $19 billion to $38 billion per co-founder at the $965 billion valuation. The IPO will be one of the largest per-capita wealth creation events in Silicon Valley history for a relatively small employee base.

The Post-IPO Retention Problem

Once employees can sell their shares — typically six months after the IPO, when the lockup period expires — many of those who joined for the mission and the upside will have both. Anthropic’s answer has historically been culture: the genuine belief, shared across the organization, that the work is important in a way that transcends its commercial value. Whether that culture survives the transition to a publicly traded company — with quarterly earnings calls, analyst pressure, and shareholder activism — is one of the most important unanswered questions in the company’s future.

The Contractor Workforce

Behind Anthropic’s products is a global workforce of contractors whose labor is essential but less visible. RLHF data labeling, safety red-teaming, and content moderation require thousands of human raters working on sensitive and psychologically demanding material. The compensation and working conditions of this workforce have become an area of growing scrutiny from ESG-focused investors and labor advocates — a reputational risk that will be harder to manage as a public company than as a private one.

PART EIGHT: WHAT YOU MIGHT BE MISSING

The OpenAI Rivalry Is Personal

The competitive dynamic between Anthropic and OpenAI is widely covered as a business story. What is less often acknowledged is how personal, ideological, and historically freighted it is. The Amodei siblings and their co-founders did not leave OpenAI over a compensation dispute. They left because they believed OpenAI was making choices that prioritized commercial velocity over safety rigor — and they bet their professional reputations on proving that a different approach was possible. Sam Altman and Dario Amodei represent genuinely different visions of how transformative AI should be built and governed, and the public markets are about to force a comparison.

The Investor Register: Who Is Really Betting

Beyond Amazon and Google, Anthropic’s capitalization table includes Spark Capital, Salesforce, Nvidia, Microsoft, and the Saudi Arabian government’s Public Investment Fund. The geopolitical implications of a government with a documented record of human rights abuses holding an equity stake in a company building AI at civilizational scale — AI eventually capable of surveillance, autonomous decision-making, and information generation — have been noted by critics and remain largely unaddressed publicly. The S-1 will require full investor disclosure, and that disclosure will trigger debate.

The Long-Term Benefit Trust: The Most Important Document You’ve Never Read

The Long-Term Benefit Trust holds influence over board composition and certain strategic decisions in ways designed to outlast any particular CEO, investor, or market cycle. Its exact powers, legal structure, trustee identities, and conditions for modification have never been fully disclosed publicly. Anthropic’s S-1 will be the first opportunity for the public to read the full terms. It may be the most consequential governance document in the history of artificial intelligence.

The Race Nobody Talks About: Three Trillion-Dollar IPOs at Once

The context for Anthropic’s IPO is not just the AI market. Three companies — SpaceX, OpenAI, and Anthropic — are each preparing to enter the public markets with valuations that could exceed one trillion dollars. This has never happened. The previous record for the most valuable company at IPO was Saudi Aramco at approximately $1.7 trillion in 2019. The sequencing and pricing of these three offerings will be among the most closely watched dynamics in the financial markets through the end of 2026.

The Question at the Center of the Prospectus

Every IPO prospectus ends with the same implicit question: do you trust this company to deploy your capital wisely, to execute on its plan, and to generate returns? Anthropic’s prospectus will carry that question and one additional one, larger and stranger than any that has appeared in a previous S-1: do you trust this team with the future?

Not merely the future of the company, or the future of the AI industry, or even the future of the technology sector. The future, in the largest sense the word can carry. Dario Amodei and his colleagues believe, and have said clearly, that the technology they are building could be the most powerful and the most dangerous that humanity has ever created. They have organized their company, their governance structure, their research agenda, and now their public capital raise around the proposition that they are the right people to build it safely — that the alternative, ceding the frontier to those less committed to safety, is worse than proceeding under their stewardship.

Whether one finds that proposition reassuring or terrifying likely depends on one’s assessment of the team, the mission, the governance, and the technology itself. What is not in doubt is that the question is real, the stakes are genuine, and the answer — whatever markets deliver it — will be one of the most consequential judgments of the era.

Conclusion: The $1 Trillion Test

When Anthropic’s IPO roadshow begins — most likely in the autumn of 2026 — the bankers and executives who take it to institutional investors in New York, Boston, London, and Singapore will deliver a pitch unlike any that has come before. They will describe a company growing faster than any in history, burning cash at a rate that reflects the scale of its ambitions, governed by structures designed to prevent it from being corrupted by the very success it is seeking, building technology that its own leadership describes as potentially civilization-altering, and asking investors to fund the next chapter of that project at a valuation approaching one trillion dollars.

By any conventional financial analysis, Anthropic is an extraordinary business. The revenue trajectory, the customer concentration, the gross margin improvement, the product-market fit demonstrated at extraordinary scale — these are the metrics that institutional investors have spent careers looking for. The path to profitability is credible, if not guaranteed. The competitive moat — in talent, in safety reputation, in the depth of enterprise relationships — is genuine and difficult to replicate.

But Anthropic is not, ultimately, a conventional business. It is an organization that was founded on a conviction — that the most powerful technology in human history must be built carefully, that safety and capability are not in tension but are mutually reinforcing, and that the team that gets this right will have done something more important than building a successful company. The IPO is the mechanism by which that conviction meets the judgment of the world’s capital markets. It is, in the truest sense, the ultimate test.

Dario Amodei has written that he believes the next decade of AI development will be among the most consequential in human history. The investors who buy into Anthropic’s public offering will be placing a bet — not just on a stock, but on that belief. The rest of us will live with the outcome either way.

“Anthropic’s IPO asks investors to decide whether they trust this team with the future — not just of the company, but of the species. No investment bank has ever had to price that into a roadshow.”

For full biographical profiles of each founder, see Part 2 of our Special Report on Artificial Intelligence and the companion piece to this article: “The Seven Founders of General Wisdom.”

This report is part of Short Shot Media's ongoing series on Artificial Intelligence and the IPOs that will define the next decade. The story is only getting bigger from here.