The Secret Engine Powering the Greatest Growth Story We’ve Ever Seen

By multiple measures, Anthropic is the fastest growing company in the history of business. Faster than Google before its IPO, faster than Facebook, faster than Zoom at the peak of the pandemic lockdown when the whole world suddenly needed video calls overnight, even faster than fellow frontier AI lab OpenAI, the company behind ChatGPT.

So the obvious question is: how? How does a company almost nobody outside of tech had heard of three years ago suddenly become a machine running at $50 billion a year?

Most folks assume the answer is the AI chatbot, where you ask a question and Claude answers you. That’s part of it. But it’s a small part. The real engine is something most of Anthropic’s own users don’t really see or understand.

First, though, why this matters for you as an investor…

Because on June 1^st, the company confidentially filed the paperwork to go public, setting it up for an historic blockbuster IPO later this year and giving ordinary investors their first opportunity to buy a piece of one of this decade’s defining AI companies.

It also matters because of the company Anthropic now keeps. Its late-May $65 billion funding round valued Anthropic at $965 billion post-money, which pushed its valuation past archrival OpenAI.

Just a couple years ago, OpenAI was ahead by a country mile. Now the two are neck and neck on the road to trillion-dollar IPOs. (OpenAI confidentially filed its IPO paperwork on June 8.)

Now, the how behind Anthropic’s historic growth…

Three things are driving it. Two are well understood. The third—the big one doing the heavy lifting—is hiding in plain sight and remains surprisingly under-appreciated.

The first driver sounds boring but provides the strategic foundation for the rest: a focus on selling to businesses with real budgets rather than the crowd. Anthropic didn’t chase the viral consumer wave that attracts a lot of AI attention. From early on the company built Claude with enterprise needs in mind: strong performance on long documents and complex code, compliance, data controls, and integrations with the big cloud platforms businesses already use.

Anthropic focused its attention there because that’s where the money and the real problems are. A teenager messing around with a chatbot might pay $20 a month, or nothing at all. But a bank that wants Claude reading and analyzing thousands of contracts, or a software company that wants it working across its entire engineering team, will sign a contract worth millions of dollars—and then spend even more next year.

A deeper reason this worked: big organizations are terrified of AI that’s unpredictable, leaks data, or goes off the rails. Anthropic made safety and reliability its whole calling card. It developed a “build AI you can trust” identity before it was fashionable. To a cautious Fortune 500 executive signing off on adopting a risky new tech, that reputation meant everything.

Today Anthropic has hundreds of thousands of business customers, including most of the world’s largest companies, and more than 1,000 that each spend north of $1 million a year on its tech.

That’s a deep, sticky, high-value customer base—not a fickle crowd of casual users.

The second driver is the one that made Anthropic famous among the talking heads and folks who build things: Claude Code.

Claude Code is basically an AI tool that writes and analyzes software to make engineers more productive and allows non-engineers to actually build useful applications. It’s like a very capable junior-to-mid level software engineer you control with simple commands in plain English (or another language) who never sleeps and can hold your whole project in its head.

Traditionally, writing software is one of the most expensive things a modern company does. It’s also one of the highest-leverage business activities, meaning a small improvement doesn’t stay small. It ripples out and makes a lot of other things better at the same time.

So an AI tool that speeds up software development delivers meaningful and visible returns in short order.

Claude Code went from a late-February 2025 launch to $1 billion in annual run-rate revenue before the end of last year, then to $2.5 billion just a few months later. Customers include Netflix, Spotify, Salesforce, and the big consulting firms.

But Claude Code is doing something sneakier than just making money as new customers come on board. Every time it reads a codebase or writes a feature, it quietly consumes a huge amount of AI horsepower under the hood. And that leads us to the third driver—the big hidden engine doing most of the work.

I’m talking about Anthropic’s API (application programming interface). This is the one almost nobody sees, but it’s the real rocket fuel behind the story.

The API is a way for customers’ software to plug directly into Anthropic’s AI and use its brainpower, without a human ever visiting a chatbot website.

If you think of Claude’s intelligence as electricity, the API is the wall socket. Any company’s software systems and tools—a customer-support operation, a hospital’s records system, a travel app, a law firm’s document tool—can plug into that socket and draw as much intelligence as it needs. And it gets billed for exactly what it uses.

It’s like a utility bill for thinking and it’s metered in tokens—basically little chunks of text or code that work like AI’s “units” of intelligence. The more thinking a company asks Claude to do, the more tokens it burns, and the more it pays Anthropic.

The API engine is powerful for several reasons.

Let’s start with reach. It lets thousands of companies add Claude’s AI capabilities directly into tools they already have or are building. Anthropic doesn’t have to invent and maintain every end-user application itself. Other companies do that work and bring their own distribution.

In other words, Anthropic gets paid to be the intelligence inside thousands of other companies while letting them do all the work of building and selling.

Then there’s the way usage grows. When a customer plugs into Anthropic’s API, it rarely stays put. One team tries it, gets a big productivity and/or efficiency boost, then spreads the word, and usage climbs—month after month.

So enterprise customers tend to spend more over time. And Anthropic grows not just by landing new customers, but by existing customers constantly increasing their spending too.

Coding and other complex, multi-step work burns through tokens at a rapid rate. So as the world’s businesses point Claude at their hardest and most valuable problems through the API, the meter doesn’t tick, it spins. Heavy enterprise use turns into serious revenue, fast.

And when customers build their workflows around Claude—connecting it to their internal data and systems and tuning it to their specific needs—switching to something else becomes very painful and expensive. The longer they use it, the more it’s worth to them, and the harder it is to leave.

About 75 cents of every dollar comes from this pay-as-you-go API, and about 80% of total revenue comes from businesses. The consumer chat side of the company is still meaningful, but much smaller by comparison.

It’s the difference between owning one popular restaurant and supplying the secret sauce thousands of restaurants use in their kitchens.

Anthropic’s API turned Claude from a clever chatbot into something more like essential plumbing for a huge chunk of the business world.

What’s more, the three key growth drivers I’ve been talking about fit together and reinforce each other in a powerful, accelerating flywheel. The enterprise focus brings in customers with big budgets, big problems, and the patience to go deep once they trust you. Claude Code hands those customers an immediate, visible win in one of their most expensive areas. And the API gives them a frictionless socket that allows all that intelligence to pour into their systems and lets Anthropic collect a toll whenever they take a sip.

Better models attract more serious users. These users generate more revenue and more real-world feedback about what to improve. That money and feedback gets poured into even better models and more computing power, which attracts more users. And around it goes.

That’s how Anthropic’s growth has been so extreme. And it’s why the humble API deserves the credit almost nobody gives it.

Why this matters when comparing the two big upcoming AI IPOs

Here’s a question many investors may be wrestling with later this year: if both Anthropic and OpenAI go public within months of each and you can only back one, which one do you want to own?

It may be tempting to go for the famous one. OpenAI built ChatGPT, which showed the whole world what AI could do. The company deserves an enormous amount of credit. But when you look past the brand and into the engine room, the case for Anthropic—right now—appears much stronger.

Let’s start with the money, because that’s what matters most. OpenAI is a marvel and a money pit at the same time. In 2025, according to reportedly leaked figures, it lost around $21 billion on about $13 billion in revenue. So it lost more than $1.60 for every dollar it took in. And it doesn’t expect to stop bleeding cash until the end of this decade.

Anthropic burns cash too—every frontier AI lab does—but it’s on a clear near-term path to profitability. It expects to post its first operating profit this quarter (Q2 2026), years ahead of its rival.

Then there’s the customer mix. While OpenAI’s own (more recent) enterprise and API efforts are growing fast, its business still leans heavily on revenue from consumer subscriptions—hundreds of millions of regular folks, many of whom use the free version of ChatGPT and may never pay a cent.

Anthropic, on the other hand, leans on revenue from business customers paying by the token. And businesses are much more lucrative and much stickier (harder to lose).

Think about how easy it is to cancel a $20 monthly ChatGPT subscription and switch to a competitor. You can do it in under 30 seconds, from your phone, on a whim.

Now think about a company that’s integrated Claude deep into its workflows and trained engineers around it. Switching to a competitor involves a months-long migration that costs a fortune and causes countless hours of lost productivity without AI’s assistance.

A subtler point: regular folks’ AI needs are nowhere near as intense as business customers. Most consumers use AI like a slightly smarter Google search—a quick question here, a rewritten email there. The models are already good enough for that. So when OpenAI launches a brilliant new model, a big chunk of its customer base just shrugs, because they didn’t need it and don’t know how to take advantage of it.

Anthropic’s customer base is the opposite. It’s comprised of software engineers and companies pointing AI at their hardest, most valuable problems. And they’ll gladly pay more every time the model gets better, because a smarter model does more real work for them.

As AI takes on a larger and larger share of the actual economy, investing in the model that entire corporations run on looks like the safer bet than owning the one the world uses to write wedding and birthday toasts.

I’m not saying OpenAI is a bad company; it’s not. It has the most recognized brand in the field, a staggering consumer reach, its own less substantial but still fast-growing enterprise and API business, and its latest models have kept pace with the cutting edge.

But the question wasn’t “is OpenAI good?” It was “which one do you want to own?” And as things stand right now, the answer for me is clearly Anthropic.

One more thing I want to hit on real quick: risk.

Heavy handed regulation in general is the biggest risk to both companies. But Anthropic is currently suffering from acute pain inflicted by the US government.

The feud between the feds and Anthropic dates back to at least February when the administration ordered federal agencies to stop using Anthropic’s models after the company refused to sign a Pentagon contract it thought lacked sufficient guardrails against their use for things like autonomous weapons and spying on American citizens.

Things seemed to be improving this spring, then fell apart again more recently. In early June, Anthropic launched its most powerful models yet. Within three days the government issued an order that resulted in Anthropic needing to pull the new models off the market entirely to comply.

As I write, weeks later, they’re still offline and members of Congress are now demanding the government explain itself.

There’s another, subtler risk worth knowing, and it ties right back to the hidden API growth engine.

Much of Anthropic’s API revenue flows through the big cloud platforms of Amazon, Microsoft, and Google, and there’s an ongoing argument about how to count it. Critics, including OpenAI, claim Anthropic books the full customer payment as its own revenue while paying the cloud companies out the back, which puffs up the headline number. When the paperwork to go public—the confidentially filed S-1—gets revealed to the world before the IPO and auditors weigh in, it’s possible the official revenue figure gets substantially trimmed. That wouldn’t change Anthropic’s trajectory, but it would damage the story for a news cycle and impact how much the company could raise at IPO.

While it’s always important to weigh risks like these, what’s fascinating is how little Anthropic’s commercial engine flinched in the face of a blacklisting by the US government, a subsequent lawsuit that’s still grinding through the courts, and the new flagship models getting yanked. While the feds were slamming doors, businesses around the world were doubling down and integrating Claude deeper into their operations to keep revenue growth right on roaring. That tells you something important about the durability of this flywheel.

The chatbot and the government fight get the headlines. But the quiet API socket humming away inside thousands of companies keeps powering the fastest revenue ramp the world has ever seen.

The Secret Engine Powering the Greatest Growth Story We’ve Ever Seen

Morning Habits That Drive Higher Productivity for Financial Advisors

The Tech Sell-Off Is Hiding a Stronger Bull Market