This probably isn’t a very accurate article. But I still find it plausible anyway. I feel like most AI subscriptions have only now started showing value to me. And even then it’s because I’ve been willing to work and experiment with it to get it to do things on my behalf.
The vast majority of users are likely comfortable with just asking it quick questions, and probably don’t need agents running in the background all the time.
A somewhat misleading article or at least headline as this is about households which will not account for people paying for AI subscriptions at work unless they are using their own personal card.
Imagine creating something people still like to pay for after over 20 years. Incredible! Disclaimer: I also have still a subscription.
Regarding LLM: I guess if you count in enterprise/company subscription, LLMs would win. Interesting would be how many really want them or just have to use them…
According to this [1] data which uses actual subscriptions instead of households, it seems the number is around 8-10million LLM subscribers in USA. This was March so I expect the figures to be around 40% higher now.
Another proxy: OpenAI said [2] they have 50M subscribers and in another place they said 40% of mobile revenue comes from USA. That itself accounts for 10M paid subscribers within USA which is 5 times the estimation from this post using ChatGPT alone, that too from February.
This probably isn’t a very accurate article. But I still find it plausible anyway. I feel like most AI subscriptions have only now started showing value to me. And even then it’s because I’ve been willing to work and experiment with it to get it to do things on my behalf.
The vast majority of users are likely comfortable with just asking it quick questions, and probably don’t need agents running in the background all the time.
I guess that WoW is more fun to play than using LLMs. Guess what will always have more costumers than LLMs?
A somewhat misleading article or at least headline as this is about households which will not account for people paying for AI subscriptions at work unless they are using their own personal card.
True, but personal use tracks direct consumer adoption, whereas sustained enterprise use is still very much in the air. (See "tokenmaxxing".)
1) This could not have been farthest from truth
2) WoW will never be billed via a corporate card.
Imagine creating something people still like to pay for after over 20 years. Incredible! Disclaimer: I also have still a subscription.
Regarding LLM: I guess if you count in enterprise/company subscription, LLMs would win. Interesting would be how many really want them or just have to use them…
This is false or misleading.
According to this [1] data which uses actual subscriptions instead of households, it seems the number is around 8-10million LLM subscribers in USA. This was March so I expect the figures to be around 40% higher now.
Another proxy: OpenAI said [2] they have 50M subscribers and in another place they said 40% of mobile revenue comes from USA. That itself accounts for 10M paid subscribers within USA which is 5 times the estimation from this post using ChatGPT alone, that too from February.
[1] https://epoch.ai/data/polling/2026-03
[2] https://openai.com/index/scaling-ai-for-everyone