Comment by deepwoods
1 day ago
About a quarter of Anthropic's revenue for the nine months ended 9/30/26 came from just five customers. That's not an absurd concentration, but I would be very interested to know who they are and how mobile their business is. From a financial perspective, the lack of moat seems like the whole ballgame, particularly because they're getting ~80% of their revenue from consumption-based spending.
Moat? An under examined aspect of AI is not just the lack of moats, but the need for some word that connotes the opposite of a moat. I have a side project that I used to explore local AI. I've got a configuration screen that lets you download any model from Huggingface that's compatible with the local AI engine. There are dozens! I've only scratched the surface of trying them and comparing them for the features that use AI. It's a glut, and it's likely to be that way throughout the AI market.
Under examined? No. This is a very common talking point.
A glut of models doesn't mean there is no moat. There are several dimensions on which to compete, and it might turn out that being able to produce models that optimize on different dimensions is a difficult and valuable thing to do. It also might not, but the "there is obviously no moat" only comes from people who don't use these things systematically.
If one or more are AWS, GCP, etc. that would make a lot of sense. Connecting to Claude via API via AWS Bedrock is really nice.
Back when I was at Amazon they used Anthropic models almost exclusively. Not sure how much that's changed in the last year.
I wonder where the China labs supposedly distilling anthropic models rank among the top customers
true, but it's also super easy to switch models via Bedrock (my harness switches subagents between A\, OAI, Kimi, Grok, etc... GLM 5.3 is locked behind some dumb request system which I personally believe is because Amazon is heavily invested in A\) and you're not locked in with the Claude Enterprise features (i swear to god if someone sends another claude artifact instead of publishing to git...)
At a guess - corpos like Big 4 consulting agencies and some extreme fintech ones, also corpos like Oracle and the like. And at another guess, outside of a legal agreement with a fixed time period, probably nothing hold them from quickly switching to a better model provider, if any comes ahead.
I'd say more likely the FAANG type companies, suspect big 4 are much smaller.
But the FAANG customers are the opposite of sticky, given that they all have their own models.
And the cost of switching is near zero. It's not like trying to get SAP out of your org.
Think of that dependency tree -- those 5 customers have how many customers on their stack?