Reading time: 5m 27s
The biggest IPO in human history is almost here. But then you see the headline figure. “Anthropic to IPO at a valuation between $1.3 and $1.4 trillion.”
Your eyebrows raise, eyes squint, and cheeks tighten. That seems like a game that’s already been played. No juice for us plebs to extract, right?
To be fair, your skepticism is warranted. Going to market at a trillion-dollar valuation is pretty hard to justify. But that’s what everybody and their mother is also thinking. So how about we play devil's advocate for a second?
Can Anthropic actually justify its $1.4 trillion valuation? If yes, then how the hell can they do that?
Let’s run through it.
Claude’s getting cheaper, but Anthropic is making more money
In its most recent filings, Anthropic revealed that as revenue and demand continue to grow at a staggering rate, the actual cost to customers is going down.
You would expect that rising demand would mean rising infrastructure and compute costs, and therefore higher costs for the end customer, but with each successive model, Claude has become more efficient.
Between Fable 5 and Fable 5.1, there’s been a 25% reduction in cost to customers, and between Opus 5 and Opus 5.5, there’s been a 40% reduction in cost.
The cheaper it gets for individuals and businesses to use these AI models, the more usage they will get, often outpacing prior usage. This is why revenue grew a reported 1,127% year-over-year in Q3 2026.
The expectation is that models will only continue to get more efficient, and the surrounding infrastructure will also improve, thereby further reducing costs and ultimately making it feasible for Claude to be entrenched in every single business.
In that case, a $1.4 trillion valuation doesn’t seem that outlandish.
900% net revenue retention
On a related note, Anthropic also reported a 900% net revenue retention (NRR).
So what does that mean exactly?
Imagine a group of businesses that, over the first year, spend $1 million collectively on Claude. Then, the next year, that exact same group now collectively spends $9 million. That’s a 900% NRR.
It does not mean that Anthropic has acquired that many new customers; it means that its existing customers have heavily ramped up spending on the products.
You’ve probably seen this play out in real time in a company you’re working at. First, they test Claude in one or two aspects of the business, find it effective, and then gradually introduce it everywhere.
So what we’re seeing is not a gradual but an aggressive takeover of AI in the workplace and life in general, because 900% is an astounding number.
Now things get even crazier with this 900% number if we take into account the next section.
Enterprise AI slowly becomes essential software
Enterprise AI is certainly picking up steam, but the keyword here is slowly.
Anthropic’s IPO Roadshow revealed that:
- More than 30% of Global 2000 companies spend at least $100,000 annually on Claude.
- More than 50% of Fortune 500 companies do the same.
- Average Claude spending at a Fortune 500 customer represents less than 0.1% of its total expenses.
- Even among Anthropic's top 10 Fortune 500 customers, spending represents less than 1% of their total expenses.
So even though that 900% figure is impressive, it’s still barely scratching the surface when it comes to spending by major companies.
It takes time for these major companies to pivot toward new technologies, and once that pivot starts ramping up, the revenue growth is only going to get bigger for Anthropic, in which case, a $1.4 trillion valuation could actually be undervalued, as crazy as that sounds.
Consider what it looks like when Claude is integrated into all systems like customer relationship management systems, compliance workflows, research and development, communications, financial reporting, and so on.
Then trying to take AI out of these systems could be even more disruptive, thereby making models like Claude a crucial part of company infrastructure. So enterprise AI is only going to grow in importance and thereby take Anthropic up with it.

Anthropic projects replacing $40 trillion worth of work
On a related note, in the recent documents, Anthropic identified $40 trillion worth of economic value generated from knowledge work that it believes Claude can realistically replace.
The industries involved include:
- Legal services and professional consulting
- Software engineering and IT
- Accounting, finance and administration
- Marketing, advertising and creative work
- Research and scientific development
- Customer service and business operations
Of course, this is not as simple as it sounds. This does not mean that Anthropic automatically generates $40 trillion in revenue because the way it gets implemented across industries will vary. And of course, Anthropic has a good number of competitors, so it is not going to capture all of this market share.
However, being among the top AI models out there, it’s fair to assume that Anthropic will capture a big chunk of this value, which really helps their bull case.
Great distribution strategy
People always say a great product means nothing if there isn’t an effective way for it to reach the most number of people.
In this sense, Anthropic has a great distribution strategy.
As their S-1 filing showed, Anthropic makes 80% of its revenue from actual Claude usage and only 20% from subscriptions. A major reason for this is their distribution strategy, which leverages cloud service providers.
If Anthropic's aim is to get Claude entrenched in every enterprise, it makes no sense to knock on doors and get people to sign up.
Enterprises already have infrastructure, including cloud service providers like AWS, Microsoft Azure, Google Cloud, Google TPUs and Nvidia GPUs.
Anthropic simply integrates with these existing service providers and offers services directly through them to enterprises, making the integration process seamless for businesses.
Given that these aforementioned companies are also investors in Anthropic, the strength of their distribution channels will only grow. So now more businesses, regardless of size, have an easy way to integrate Claude into their systems, thereby helping Claude grow and Anthropic with it.
Of course, this is not exclusive to Claude, as other frontier models also use this strategy, and you also don’t want to be dependent on other companies for distribution, but nonetheless, it will be a great avenue to grow the business.
High employee retention rate
This is more of a footnote or honorable mention compared to the rest, but nonetheless, it’s a good sign.
Among the AI companies, Anthropic has the highest employee retention rate. The Roadshow reported that it had only a 1% regrettable employee attrition rate.
The reason is that each employee is highly efficient at their job and gets compensated handsomely for it.
Anthropic reported that the company generates around $6.6 million per employee. That’s Tether-level impressive.
So all this time, they have been improving the product, growing demand, and increasing revenue without significantly increasing headcount and actually retaining their best talent.
That’s always a good sign if you’re looking at the bull case for any company. Will this change after they pick up their IPO checks? Maybe. But for the time being, we’ll assume things will keep moving ahead in full steam.
Concluding thoughts
The future of AI is entirely unknown at this point. Nobody knows where the technology will go in the future or how it will integrate into society as a whole.
All we can do is simply present the facts as they come, and based on the data available, Anthropic may not be as bad an investment as it seems.
Of course, this is not financial advice of any kind, so please do your own research, think through all the possible scenarios and then make your decision. Happy investing.











.webp)












.webp)

.webp)

.webp)

.png)



.webp)



.webp)
.webp)

.webp)
.webp)





















.webp)

.webp)


.webp)






.webp)
.webp)





.webp)

.webp)

















.webp)
.webp)

.webp)
.webp)
.webp)


.webp)
.webp)










.webp)


.webp)









.webp)







.webp)




.webp)

























.webp)







.webp)















.webp)

.webp)
.webp)

.webp)














.webp)

.webp)


.webp)








.webp)



