Anthropic Goes Public: The Defectors Strike Back
Anthropic has confidentially filed to go public, which means the AI nerds now have a fresh reason to attack each other with the passion normally reserved for college football message boards and HOA meetings. This is a big deal. Anthropic is the company founded by OpenAI defectors, and now it may beat OpenAI to the public markets. The people who walked out of the most famous AI company in the world are suddenly standing at the IPO window first, checking their pockets for a trillion-dollar valuation and pretending this is all perfectly normal.
The company announced that it confidentially submitted a draft S-1 to the SEC. In normal English, that means Anthropic has started the process to go public, but we do not yet get the full financials, share count, or official IPO price. The company said the offering depends on SEC review, market conditions, and other factors. Dario said, this move gives them the options to go public after the SEC completes its review, which is CEO for, “Get in, loser, we’re going public.” Anthropic’s most recent private round valued the company at $965 billion post-money, which is close enough to a trillion dollars that the zeroes probably need their own parking permits.
The origin story is almost too good. Dario and Daniela Amodei, along with other former OpenAI employees, founded Anthropic in 2021 after leaving OpenAI. The simple version is that they wanted to build powerful AI with a stronger safety culture. The more cinematic version is that a group of researchers walked out of the Death Star, built their own lab, named the assistant Claude, and now might IPO before the empire they left. If someone pitched this to Netflix, the notes would say, “Too obvious. Needs more subtlety.”
Since then, the lore has been ridiculous in the best possible Silicon Valley way. Anthropic was designated a supply-chain risk by the Pentagon, then fought the government in court, with a federal judge temporarily blocking part of the blacklisting after saying the Pentagon appeared to have retaliated against Anthropic over its AI-safety views. At the same time, Axios reported that the NSA was still using Anthropic’s most advanced model despite the Pentagon dispute. So, depending on which hallway in Washington you were standing in, Anthropic was either too dangerous to trust or too useful to stop using. That is not a normal corporate development. That is a John le Carré novel with a data center budget.
Then there is the Vatican chapter. Anthropic cofounder Chris Olah appeared at the Vatican around Pope Leo XIV’s AI encyclical, which warned about the moral, economic, and human consequences of artificial intelligence. There is something wonderfully strange about an AI company built in San Francisco ending up in a conversation with the Pope about human dignity. Wall Street has seen founder-led companies, dual-class shares, roll-ups, SPACs, meme stocks, and crypto miners. It has seen fewer companies that can plausibly say, “We had a meeting in Rome about whether our product threatens civilization.”
And because this story apparently needed one more plot twist, Anthropic also signed a major compute agreement with SpaceX. Elon Musk had previously mocked Anthropic as “Misanthropic” and accused it of hating Western civilization. Then SpaceX agreed to lease Anthropic access to massive AI data-center capacity. Musk later clarified that the arrangement begins as a short-term lease with termination rights, but the larger point remains: in AI, yesterday’s civilization-ending enemy can become today’s infrastructure vendor. There are awkward Thanksgiving dinners, and then there is paying Elon Musk’s company for compute after he called you evil.
The numbers are the reason everyone is paying attention. Anthropic says its revenue run rate crossed $47 billion earlier this year, and it raised $65 billion in its latest round at that $965 billion valuation. Earlier in 2026, Anthropic said it had a $14 billion revenue run rate, more than 500 customers spending over $1 million annually, and eight of the Fortune 10 as Claude customers. Claude Code alone had reached more than $2.5 billion of run-rate revenue. That is not a cute chatbot business. That is enterprise software growth after someone dropped it into a particle accelerator.
There are also reports that Anthropic is approaching its first operating profit. Reuters Breakingviews reported that Anthropic generated $4.8 billion in Q1 revenue, projected $10.9 billion in Q2 revenue, and expected an adjusted operating profit of $559 million for the second quarter. The reported compute cost per dollar of revenue dropped from 71 cents to 56 cents. That part matters because the bear case on AI has always been that these models are fantastically expensive to run and get more expensive as demand grows. Anthropic is trying to show the opposite: that scale can create efficiency instead of just a larger electric bill with a Stanford hoodie.
The caution is that these are still private-company numbers. We do not yet have the audited S-1. We do not know exactly how revenue is recognized, how compute commitments are treated, what stock-based compensation looks like, how customer concentration breaks down, or whether a first profitable quarter is the beginning of a durable margin story or a beautiful one-quarter postcard from the land of adjusted accounting. Private-company financials can be like dating profiles. Usually true-ish. Occasionally filtered. Always better once you meet them in daylight.
The valuation trajectory is almost comical. Anthropic was valued at about $4 billion in 2023 and is now valued privately at $965 billion. That is not normal growth. That is financial mitosis. They are inventing new numbers for this company. The closest comparison may be the pandemic-era growth stocks, except this one has better enterprise traction and far larger infrastructure needs. Zoom grew fast because everyone suddenly needed video calls. Anthropic is growing because companies are beginning to shove AI into coding, analysis, workflows, support, finance, and anything else involving a keyboard and a human who would rather go home before 9 p.m.
The most unusual part of the Anthropic IPO may not be the revenue, the valuation, or even the compute costs. It may be the governance. Anthropic is a public benefit corporation with something called the Long-Term Benefit Trust. The Trust is an independent body of trustees with backgrounds in AI safety, national security, public policy, and related fields. Its purpose is to keep Anthropic aligned with the long-term benefit of humanity while also balancing shareholder interests. The Trust now appoints a majority of the board. This is not a standard “maximize shareholder value and please enjoy our risk factors” structure. This is closer to putting a council of safety guardians inside a trillion-dollar company and asking Wall Street to please read the footnotes before sprinting toward the buffet.
A recent Harvard Law paper called this the “Ben & Jerry’s Risk.” That is not a joke, although it should be served with sprinkles. Ben & Jerry’s had a mission-protection structure, then Unilever acquired it, and that structure eventually led to conflict between the company’s independent mission guardians and its corporate owner. The Harvard authors argue that OpenAI and Anthropic both raise a similar governance question: what happens when investors want one thing and mission guardians want another? In Anthropic’s case, a supermajority of stockholders can terminate the Trust and remove directors appointed by it. So the safety guardrail has a kill switch. Does anyone else see the irony here, or do I need to ask Claude to diagram it?
That governance tension may become the most important part of the story. Investors love mission statements until those missions interferes with margins. They love safety until safety delays a product launch. They love long-term thinking until the stock is down 18% and some analyst with perfect hair asks why the company sacrificed revenue for humanity. The hard question is not whether Anthropic has good intentions. The hard question is who gets to define “good” when the stock price disagrees with the ethics committee.
That brings us to the harder question: is Anthropic the start of a new category, or is the market once again trying to put a trillion-dollar wrapper around a very expensive dream?
My answer is fairly simple. I believe this is the start of the next industrial revolution. In this case, the hype may actually be underselling the reality. That does not mean every AI company will win, or that every valuation will make sense, or that investors should sprint blindly into anything with “AI” taped to the side of it like a garage-sale price tag. But the change itself looks real. Software is beginning to do work. Compute is becoming a form of industrial capacity. Intelligence is becoming a product.
For that reason, my clients and I will be investors in not only Anthropic, but OpenAI and SpaceX as well, where appropriate and when access makes sense. These are not small side stories in technology. They are part of the infrastructure being built underneath the next version of the economy. Anthropic is building frontier AI. OpenAI is pushing consumer and enterprise adoption at massive scale. SpaceX is building the communications, launch, and satellite backbone that may help connect and power much of what comes next. Different companies, different risks, same broad theme: the physical and digital foundations of the next economy are being laid in real time.
The valuation is where the debate begins. A company can be excellent and still be priced for perfection. Anthropic may become one of the most important companies in the world. It may also discover that public investors are less forgiving than private investors when growth slows, margins tighten, or the power bill starts eating like a teenage offensive lineman. Both things can be true at the same time, which is what makes markets so much fun and occasionally why they require antacids.
When the public S-1 finally arrives, the headline valuation will get all the attention, but the real story will be underneath it: gross margins after compute, revenue quality, customer concentration, retention, capex commitments, related-party infrastructure deals, and how much power the Long-Term Benefit Trust really has once public shareholders enter the room. Anthropic is asking investors to price intelligence, safety, compute, governance, and belief all at once.
Good luck putting that into a spreadsheet.
As always, if you would like to talk about how this may affect your portfolio, or how we are thinking about AI, infrastructure, software, and the companies powering this next wave, please give us a call.
Your capital, our experience, a bespoke creation.