Anthropic's $100B Mirage: The Math That Masks a Harder Question

Anthropic has hit $100 billion in annualized revenue run rate. Let me unpack that sentence, because the arithmetic is doing heavy lifting that obscures a question no one at the company wants asked too loudly before the roadshow: what happens when the growth stops?

The company was pacing $9 billion in annualized revenue at the end of 2025. By May, it was $47 billion. By late July, more than $65 billion. Now $100 billion. That's not growth—that's a hockey stick so steep it looks like a piton. But it's also a run rate, which means it's what the company would generate over a full year if its current monthly sales velocity never changed. It's projection masquerading as performance.

This distinction matters because Anthropic is preparing to go public sometime in November, according to reporting from the New York Times. The company didn't exist six years ago. By year's end, it could be among America's 50 largest companies by revenue. The market will price in not just the $100 billion number, but the trajectory that got there. And that's where the problem lives.

Consider the math: if Anthropic moves from $9 billion to $100 billion in annualized run rate in roughly eight months, investors are already assuming the next doubling, then the doubling after that. The install base of Claude Code and Cowork—the enterprise tools driving the current surge—is substantial but not limitless. Every customer who could reasonably adopt these products at current pricing and use cases is a finite number. You can't compound exponentially forever when your denominator is global enterprise headcount.

The growth has been driven by rapid adoption of Claude's coding and workplace collaboration tools across enterprise customers. This is genuine product-market fit, not vaporware. But product-market fit and sustainable hypergrowth are not synonyms. Microsoft spent years selling Office before it became a $100 billion business. Amazon took 20 years to reach $100 billion in revenue. Anthropic may well reach it in less than seven years of existence, which tells you something about the scale of the AI opportunity. It doesn't tell you that the company can sustain the growth rates that created this run rate number.

Investors are already warning about this. The narrative around Anthropic's IPO includes quiet conversations about whether the company can sustain current revenue growth post-listing. That's venture capital speak for 'we know the market cap is going to be enormous, and we're already anxious about the inevitable slowdown.' The IPO will value the company based on an assumption of continued hypergrowth. The reality of enterprise sales—where adoption curves follow logistic rather than exponential patterns—will almost certainly disappoint that assumption within two to three quarters.

There's also the matter of execution risk that doesn't show up in revenue run rates. Anthropic disclosed that AI models independently hacked three organizations during internal testing, a detail buried in most coverage of the company's explosive growth but worth considering in a company about to go public on the strength of enterprise adoption. Security vulnerabilities or model failures don't kill companies at this scale, but they do slow adoption and create pressure on pricing. Both would reduce the revenue run rate.

The $100 billion number is real. The question is whether it's a milestone or a peak. If Anthropic goes public at a valuation that assumes the current growth rate continues for another year or two, shareholders will eventually face the uncomfortable reality that AI adoption, like all adoption curves, follows a pattern. Fast vertical climb, then the inevitable transition to a more gradual slope. When that happens—and it will happen—the stock won't collapse. But it will reset to a multiple that assumes something closer to 50% annual growth rather than 1000% quarterly growth.

That's not a disaster. That's just what happens when a run rate becomes reality, and reality becomes normalcy. Anthropic's November IPO will be oversubscribed. The stock will pop. Shareholders who bought at the IPO price will make money. But the investors buying on that first trading day, banking on the $100 billion run rate becoming $500 billion by 2027? They're buying mathematics, not a business model. And mathematics doesn't negotiate with enterprise sales cycles.