Anthropic Now Spends More on AI Than on Salaries
In short
- Anthropic’s amended IPO filing shows that its internal Claude use, valued at retail prices, passed its monthly employee pay (excluding stock) back in April.
- By September that internal usage was up more than 250x since January, and over 90% of new code was written by Claude.
- Money angle: tokens are becoming a payroll line. A solo founder can copy this for a fraction of a salary.
Anthropic now “spends” more on its own AI than on its people. That is the quiet number inside the amended S-1 filed on Monday. If you run a small business, it tells you where the cost of work is heading.
What happened
On November 2 Anthropic filed an amended Form S-1 with the SEC. It added third-quarter numbers. Q3 revenue was about $17 billion. Growth slowed to roughly 47% over Q2, after 144% in Q2 and 132% in Q1.
Business Insider spotted the detail that matters for builders. Anthropic values the Claude it uses internally at the prices it charges customers. In April that value was already higher than its monthly employee compensation, not counting stock awards. By September the internal usage had grown more than 250 times since January.
One caveat. This is not cash leaving the bank. It is internal usage priced at retail. Still, the filing also says more than 90% of the code added in September was written by Claude. That part is real work done.
How to use it
Do what they do, at your size. Count tokens as a cost of labour, next to salary.
Do the math: a freelance developer or support helper costs about €2,500 a month. Claude Ultra costs $200, roughly €185. If it takes over even a third of that work, you save around €800 a month, or about €9,600 a year. These are my assumptions, so plug in your own numbers.
- List the repeated tasks that eat your week: replies, product descriptions, small scripts, reports.
- Give each task to an agent for two weeks. Track hours saved.
- Only then decide if you hire, or pay for a bigger plan.
What it means
Revenue per employee is going to explode at companies that do this well. Anthropic grew revenue far faster than headcount. Teams that treat AI as a tool on the side will lose to teams that budget for it like staff.
The slowdown to 47% is the other signal. Even the leader cannot grow 130%+ every quarter, and it carries about $895 billion in compute and infrastructure commitments. Prices may stay low while that race runs. Lock in what works now.
Ivan’s Take
This is the model I want for my own businesses: few people, many agents, measured by results. I would start by moving one repeated job to AI this week and measure it. Do not copy the hype, copy the accounting.
Sources: CNBC · Business Insider
If it doesn’t save you time or money, you won’t see it here
- Facts, not hypeEvery price and number is checked against the official source
- Always the real priceYou see what it costs before you click
- I say when it’s weakEven when it’s OpenAI or Google
