In May 2026, the artificial intelligence startup that began as a breakaway faction of disgruntled OpenAI safety researchers became the most valuable private AI company on Earth. The path from $4 billion to $965 billion in three years runs through a personal feud, a public refusal to hold hands, and a fundamentally different theory of how to build AI.

By Kenny Le

On May 28, 2026, Anthropic disclosed a Series H funding round that raised $65 billion at a post-money valuation of $965 billion (Pequeño, 2026). The round was led by Altimeter Capital with participation from Greenoaks and Sequoia Capital. With that single financing event, Anthropic surpassed OpenAI to become the highest-valued AI startup in the world, eclipsing OpenAI’s $852 billion post-money valuation set in its $122 billion round on March 31, 2026 (Capoot, 2026).

The headline figure obscures a more complicated story. By revenue, OpenAI is still nearly twice the size of Anthropic. By origin, the two companies share a building, a research lineage, and seven founders who used to sit on the same side of the table. By temperament, they have spent five years drifting toward opposite poles of nearly every public question facing the AI industry, from the Pentagon to the Super Bowl to, most famously, whether their CEOs will hold each other’s hands on stage.

For investors, the rise of Anthropic is the most consequential reshaping of the AI industry since ChatGPT launched in November 2022. For everyone else, it is a case study in how a company built by people who left their previous employer over safety concerns ended up running ahead of that employer in the very market they helped create.

The Numbers Behind the Crossing

The chart below tracks the post-money valuation of both companies from 2023 to 2026, and compares their annualized revenue (ARR) as of February 2026, the most recent comparable figures publicly disclosed.

Two patterns are worth pulling out.

First, the valuation crossover happened in a single round. As recently as February 2026, Anthropic closed a Series G at $380 billion (Anthropic, 2026), still less than half of OpenAI’s $852 billion mark a month later. In the 14 weeks between those two events, Anthropic’s valuation roughly tripled while OpenAI’s was effectively flat in the private market.

Second, revenue tells the opposite story. As of February 2026, Anthropic disclosed run-rate revenue of $14 billion, with the figure growing more than 10x annually for three consecutive years (Anthropic, 2026). OpenAI, as of the end of February 2026, was at $25 billion in annualized revenue, a 17% rise from $21.4 billion at year-end 2025 (Reuters, 2026). OpenAI is also generating roughly $2 billion in revenue per month, according to its March 2026 funding announcement, with more than 900 million weekly active users and over 50 million paying ChatGPT subscribers (OpenAI, 2026).

The market is paying Anthropic a premium not for its current revenue, but for its growth trajectory and a particular bet about where enterprise AI spend is going. At $965 billion on $14 billion ARR, Anthropic trades at roughly 69x ARR. At $852 billion on $25 billion ARR, OpenAI trades at roughly 34x. Investors are quite literally pricing Anthropic at double the revenue multiple of its closest competitor.

A Breakaway in Slow Motion

The origin story of Anthropic is the origin story of a personal split.

Dario Amodei was not a founding member of OpenAI. He joined roughly a year after the lab was created in 2015, taking on the title “Team Lead for AI Safety” (Griffiths, 2026). By September 2018, OpenAI had named him research director. By December 2019, he was Vice President of Research. He played a central role in the development of GPT-2 and GPT-3, telling The New York Times after the GPT-3 release that the model had “this emergent quality” that surprised even its creators (Griffiths, 2026).

Tensions surfaced in late 2020. According to Keach Hagey’s biography of Sam Altman, The Optimist, Amodei successfully lobbied to keep co-founder Greg Brockman off the team that built GPT-3, a power play that “ruffled feathers” within OpenAI (Griffiths, 2026). The disputes escalated around what Amodei saw as inadequate caution in releasing increasingly capable models. He later told friends he “felt psychologically abused by Altman.” Altman told colleagues the friction “was making him hate his job” (Griffiths, 2026). On December 29, 2020, OpenAI announced that Amodei and a “handful” of colleagues were leaving.

In early 2021, Amodei founded Anthropic with seven other former OpenAI employees, including his sister Daniela Amodei, who became president. According to AI Business, only one of Anthropic’s initial employees had not previously worked at OpenAI (Griffiths, 2026). The company name was chosen to emphasize a human-centered mission.

Amodei has been measured in public about why he left. In a 2024 interview with Lex Fridman, asked directly about his departure, he said simply: “It is incredibly unproductive to try and argue with someone else’s vision” (Griffiths, 2026). It is one of the more diplomatic exits in modern Silicon Valley history, and also one of the most consequential. By 2026, the company Amodei built had become the second-largest AI lab in the world by enterprise market share and the largest by private valuation.

Claude and the Restraint That Almost Cost Everything

Anthropic’s first major model, Claude, finished training in the summer of 2022. The company chose not to release it.

The decision was rooted in Anthropic’s founding thesis that powerful AI should be deployed carefully, with safeguards. Amodei later told Time Magazine, “I suspect it was the right thing to do,” before adding the qualifier that has come to define Anthropic’s public posture: “But it’s not totally clear-cut” (Griffiths, 2026).

Months later, on November 30, 2022, OpenAI released ChatGPT. The launch turned OpenAI into a household name, attracted unprecedented investor interest, and reshaped the public’s understanding of what AI could do. Anthropic, having held back, watched its closest competitor capture both the consumer mindshare and the funding gravity of the entire AI sector.

That early restraint became part of Anthropic’s brand and arguably part of its business case. By the time Claude reached enterprise customers in scale, the safety-first positioning had matured into a real commercial differentiator. Large companies, regulated industries, and AI-cautious boards found in Anthropic a vendor whose conservatism on deployment was a feature, not a bug. Industry analysis from early 2026 noted that more than 1,000 companies were spending more than $1 million each on Anthropic’s services, a level of enterprise concentration that fueled the company’s vertical revenue curve (Roetzer, 2026).

Claude Code, the agentic coding tool Anthropic launched publicly in May 2025, alone reached more than $2.5 billion in annualized revenue by February 2026, having more than doubled since the beginning of that year (Anthropic, 2026; SaaStr, 2026). The product the company once delayed releasing has become one of the fastest-growing software-as-a-service offerings on record.

The Hands That Did Not Touch

On February 19, 2026, at the India AI Impact Summit in New Delhi, Prime Minister Narendra Modi orchestrated a group photo with leaders from across the global AI industry. The arrangement was deliberate. Modi raised his hands together with Sam Altman and Alphabet CEO Sundar Pichai in front of an applauding audience, encouraging the other CEOs on stage to do the same. The visual was meant to signal unity.

Altman, positioned next to Amodei, kept his arm extended toward Modi but did not reach for Amodei’s hand. Amodei, in turn, grasped the hand of the person on his other side. The two men, standing shoulder to shoulder, raised their fists rather than clasp each other’s hands (CNBC, 2026; Anadolu Agency, 2026). The photograph went viral within hours.

The moment crystallized something that had been building for years. The same week as the India summit, Anthropic ran a major Super Bowl advertising campaign that pointedly mocked AI chatbots that push products at users, a thinly veiled critique of OpenAI’s announced plans to test advertising in ChatGPT. Altman responded on X with a mixture of compliments and accusations: “The good part of the Anthropic ads: they are funny, and I laughed,” he wrote. “But I wonder why Anthropic would go for something so clearly dishonest.” He added that “Anthropic serves an expensive product to rich people,” and that OpenAI, by contrast, was committed to bringing AI to “billions of people who can’t pay for subscriptions” (Griffiths, 2026).

The rivalry has occasionally turned operational. Days after the Super Bowl exchange, both companies released major updates to their coding-focused models within minutes of each other. Earlier, in November 2023, when OpenAI’s board briefly ousted Altman, it reportedly approached Amodei about replacing him and even merging the two companies. Amodei declined both offers (Wikipedia, n.d.).

The Pentagon Split: Two Theories of AI Citizenship

The most consequential divergence between the two companies in 2026 was not financial. It was geopolitical.

In late February 2026, Anthropic refused to sign a usage agreement with the U.S. Department of Defense, citing what Amodei called “red lines” against deploying its models for “the mass domestic surveillance of Americans and fully autonomous weapons” (Griffiths, 2026). Hours after Anthropic walked away, OpenAI struck a deal with the Defense Department.

Altman acknowledged during an X-based AMA on February 28 that the deal was “definitely rushed,” and offered a frank assessment of how it might be remembered: “If we are right and this does lead to a de-escalation between the DoW and the industry, we will look like geniuses, and a company that took on a lot of pain to do things to help the industry. If not, we will continue to be characterized as rushed and uncareful” (Griffiths, 2026).

The fallout escalated. According to Business Insider’s reporting, Defense Secretary Pete Hegseth indicated he would formally designate Anthropic as a supply chain risk, a move that could prevent any company holding Pentagon contracts from working with Anthropic at all. Altman, perhaps unexpectedly, defended his rival’s right to set red lines: “I think it is an extremely scary precedent and I wish they handled it a different way. I don’t think Anthropic handled it well either, but as the more powerful party, I hold the government more responsible” (Griffiths, 2026).

The episode laid bare two theories of AI corporate citizenship. OpenAI’s theory is that engagement with state power, even when imperfect, produces better outcomes than refusal. Anthropic’s theory is that drawing firm lines and accepting the commercial cost is itself a safety mechanism. Both companies argue their approach is the responsible one. Both have evidence on their side.

What the Valuation Premium Actually Means

Anthropic’s $965 billion price tag is striking. So is its 69x ARR multiple, double that of OpenAI. Three forces appear to be at work in the valuation.

The first is the growth curve itself. Anthropic’s ARR has expanded more than tenfold annually for three consecutive years, from roughly $1 billion in December 2024 to $4 billion by mid-2025, $9 billion by year-end 2025, and $14 billion by February 2026 (SaaStr, 2026). That is the steepest enterprise SaaS revenue curve in recorded history.

The second is enterprise concentration. While OpenAI’s revenue is heavily weighted toward consumer subscriptions, with ChatGPT Plus and Pro making up roughly 85% of ARR (FutureSearch, 2025), Anthropic’s revenue base is built on large enterprise contracts. Enterprise revenue carries higher gross margins, lower churn, and more predictable expansion. For investors valuing future cash flows, a dollar of enterprise ARR is meaningfully more valuable than a dollar of consumer subscription ARR.

The third is leadership stability. Anthropic still has all seven of its founding executives. OpenAI has been through Altman’s brief 2023 ouster, the departure of Ilya Sutskever, Andrej Karpathy, and John Schulman (who left for Anthropic in 2024), and a continuous churn of senior researchers. Amodei has used this contrast in public, telling a Bloomberg audience in 2024: “We have 7 cofounders. Three and a half years later, we’re all still at the company” (Griffiths, 2026).

What Comes Next

Anthropic’s earnings trajectory will be tested in the coming quarters. Reporting from The Wall Street Journal in May 2026 indicated that Anthropic’s earnings were projected to more than double in the second quarter, putting the company on track for its first operating profit (Hyatt & Schechner, 2026). Profitability, if achieved, would be unusual for an AI lab at this stage of investment. OpenAI, despite its larger revenue base, continues to burn cash and has not yet reported a profitable quarter (Capoot, 2026).

The competitive trajectory is genuinely open. OpenAI maintains a commanding lead in consumer AI and is closer to a public offering. Anthropic leads in enterprise share, revenue multiple, and, for the moment, valuation. Both companies are scaling their data center footprints at a pace that has reshaped the global semiconductor and power supply chains.

The Bottom Line

References