Anthropic Is the Most Interesting IPO Because the People Who Buy It Won't Be in Charge

Everyone is talking about Anthropic's valuation. It went from $4.1 billion in early 2023 to $965 billion in May 2026. That's the number everyone repeats. But the valuation is a consequence, not a cause. It doesn't tell you what's happening.

What's happening is something that has almost no precedent in the history of software: a company went from near-zero revenue to a $47 billion annualized run-rate in 24 months and it's about to post its first quarterly operating profit. By year-end, its run-rate will out-earn every public software company on earth except Microsoft. The growth curve is so steep that historical comparisons don't work. There's no historical comparison.

The more interesting question is what happens when you build this kind of economic engine inside a governance structure designed to say no to the people who will own it.

Anthropic is a public benefit corporation. It filed its S-1 with the SEC on June 1, 2026, giving the company the option to go public after SEC review. The expected valuation is near a trillion dollars. Reports put it in the $2 to $3 trillion range if the market cooperates. That would make it the largest IPO in history.

Here's the thing about Anthropic's governance that doesn't get repeated as often as the valuation: a body called the Long-Term Benefit Trust sits above the shareholders. It has five trustees with the legal power to prioritize AI safety over financial returns. It can elect a majority of the corporate board. When the company says its "mission will not change" after going public, that's not corporate PR. It's a structural feature. The shareholders are going to own a stake in a company they don't control.

That's not nothing. In every other massive IPO — Tesla, Shopify, SpaceX — the public markets eventually bend the company toward what they can measure. Quarterly earnings, margin expansion, market share. Anthropic is designed so that the mission can override all three. The Long-Term Benefit Trust can tell a trillion-dollar company to slow down because moving faster would be unsafe. The shareholders get to hold the stock but not the steering wheel.

Now put that structure next to the financials and the contradiction becomes more interesting.

Anthropic grew from roughly $1 billion in annualized revenue at the end of 2024 to over $30 billion by April 2026. Its Q1 2026 revenue was about $4.8 billion. Q2 projected to $10.9 billion. The cost to serve one dollar of Claude usage dropped from $0.71 in Q1 to $0.56 in Q2. That's not just growth. That's the unit economics of a business learning how to operate at scale — faster than any software company in modern history has learned.

The company is projected to post its first quarterly operating profit in the second half of 2026, around $559 million in Q2. That sounds small next to the revenue numbers, but it matters because it proves the curve isn't just top-line momentum funded by investor patience. The margins are actually working.

Compare this to OpenAI, which filed its own S-1 around the same time and is valued at roughly $852 billion, $113 billion less. OpenAI generates about half of Anthropic's revenue, burns cash at a much higher rate, and doesn't expect to approach profitability before 2029 or 2030. OpenAI's model is consumer-first: ChatGPT has over 900 million weekly active users. Anthropic's model is enterprise-first: about 80 percent of its revenue comes from enterprise clients, with 300,000 business customers and long-term contracts that renew at 90 percent or higher.

That mix difference matters more than most people realize. Enterprise revenue is stickier and more margin-friendly. Consumer free-users are a distribution story. Enterprise contracts are a cash-flow story. And cash flow is what eventually determines whether a trillion-dollar valuation is justified or just deferred arithmetic.

Anthropic's peak training cost is estimated at roughly $30 billion. OpenAI's is estimated at $120 billion. That factor-of-four difference doesn't just mean Anthropic is cheaper. It means the company was designed around capital efficiency rather than raw scale. You can see it in the product line too — Claude has Sonnet and Haiku models for everyday enterprise work, not just the frontier Opus tier. The capital-efficient architecture is a product strategy, not just an accounting preference.

This is the kind of thing that compounds. Double the revenue with the same training architecture and your margins don't just improve — they bend in your direction in a way that's superlinear. That's why the cost-to-serve ratio dropped so fast in a single quarter. The infrastructure amortizes across more usage without needing proportionally more spend.

But then you come back to the governance question.

Because the Long-Term Benefit Trust exists, public-market investors are going to buy Anthropic stock with a constraint built in. If the trustees decide that safety concerns require pulling back on model releases, or limiting deployment in certain sectors, or investing more in research than in growth — the shareholders don't get to vote that down. This isn't a dual-class share structure where the founder keeps extra votes and eventually steps aside. It's a permanent institutional brake.

Some people will call this visionary. Some will call it a governance red flag. Both interpretations might be right, depending on what kind of risk you think AI actually represents.

I suspect the market will price Anthropic in two ways at once. The financial trajectory will command a premium. The governance structure will demand a discount. The question is which force is stronger. And the answer depends on something harder to estimate: whether public-market investors will treat the Long-Term Benefit Trust as insurance or as a hostage situation.

There's one more thing. Anthropic is structurally dependent on two cloud providers — Amazon and Google — that are also its biggest investors. Amazon alone has committed $13 billion. Google holds roughly a 14 percent stake, capped at 15 percent. The company pays SpaceX $1.25 billion per month for compute through 2029. Its total committed cloud infrastructure spend is estimated at $80 billion through 2029. This isn't a diversified revenue story underneath. It's a relationship-heavy business that happens to be growing at the speed of light.

That concentration risk gets worse, not better, when you layer the benefit trust on top. If a shareholder meeting wants to diversify cloud dependencies to reduce supplier risk, and the trustees say the current arrangement supports the safety mission better, the shareholders can't override it. The incentive alignment between capital providers and governance holders isn't clean.

So here's what the Anthropic IPO is actually asking the market to evaluate. Not whether the revenue growth is impressive. No one doubts that. Not whether the valuation is high. No one disputes that either. The real question is whether you can build a company that generates the kind of returns software investors are used to — enterprise margins, compounding unit economics, 14-fold year-over-year revenue growth — inside a governance wrapper that legally reserves the right to reject those returns when they conflict with a mission the shareholders can't change.

You can build it. Anthropic has. Whether the market will pay for it at a trillion dollars — or treat the governance structure as a permanent cap on what the stock is actually worth — is the test.

Here's how I'd think about it before the shares trade: if you value Anthropic as a pure cash-flow generator and the Long-Term Benefit Trust looks like insurance against reckless execution, the premium makes sense. If you value it as a governance experiment where financial discipline is permanently subordinate to an unelected trust, the discount is the rational response. Both are internally consistent. Only one will be right. The market will tell us which one when the roadshow starts.