Post

Next newsletter on wheresyoured.at I believe, outside of ChatGPT, that the combined generative AI industry is smaller than Pokémon Go at its peak, with apps like Gemini and Copilot having less than 20 million monthly active users. These unprofitable, unsustainable apps are also deeply unpopular.

114 reposts660 likes

I think you may be neglecting the fact that these apps are not the real product. They are a demo of the real product, which is not a chatbot, but the models themselves which are being used behind the scenes at many companies to replace or augment human labor.

Revenue is what I’m thinking about. I would have said the AI industry includes all the products using AI your analysis doesn’t account for like Salesforce’s.

That's because sales force doesn't disclose it, Microsoft says $13bn in annual revenue from Ai which is pathetically small and includes literally every single Microsoft product that makes any money from AI including the cash they get from the oAI revshare. Teeny tiny industry!

Also curious about this. It’s confusing because Microsoft bought GitHub right? But I’m pretty sure GitHub Copilot is different and not included in this analysis.

It sucks to me that they seem ok with this naming collision. They are clearly, imo, favoring the general purpose one just because it has a higher market cap. GitHub Copilot was first and I know people actually use and like it. I use Claude more now I guess but still think (GitHub) Copilot is good

Yeah this doesn’t pass the sniff test for me if its meant to include GH Copilot. SO many software companies are buying IDE plugin licenses for all their engineers, often whether the engineer actually asks for it or not.

Well cool you might want to include that in your piece then instead of saying “the app and the site” if you have those numbers. Its BANANAS how hard its being pushed in the software industry and it’d be great to have a reference to how much is talk and how much is real.

Ed, not saying your wrong about sustainability, but I think you should also be counting API clients here which I think (based on developer interaction) would be pretty high

It’s more of a vibe… but I’m sitting in my small share office and there are three tiny start ups paying hundreds of dollars a month to Google, OpenAI and all the AI developer forums are extremely busy compared to the usual

Funny thing about gen AI is, that almost anything I would actually use one for, can be done locally on my machine, in a way that doesn't make any revenue to anyone. Big proprietary models are still better with some tasks like coding, but with some performance gains, that gap is closing fast.

was about to say this. copilot is also on like every modern windows install and pins itself to the taskbar. 11 million is crazy low on comparison

It's especially galling to be considered a user of something like Gemini, when Google doesn't allow me to remove it from many of my Workspace services. If anything I am a Monthly Conscripted User

Especially because it keeps moving around on my Android. In Messages, it's now floating right beside the floating "new messages" button, and the same colour. Accidental touches make one a "user," I guess.

I'm sure just another few billion dollars on investment will send those AI usage numbers skyrocketing! But if not, here's some ideas to help companies: - Microsoft can build office apps directly into copilot instead of the reverse. - meta ai being able to add legs to all of your messages.

- Grok promotes your posts to more users based on the number of prompts or swastikas you used. - Gemini gives your phone an "iron dome" that protects you from cyber criminals and terrorists alike. (unless Israel wants to do the terrorism) Feel free to take my ideas guys! You were going to anyhow.

You know there are thousands of other AI tools/startups (some of them very popular), right? Anthropic for example earns much more from the API services that they sell to those tools than from their Claude service.

Anyway, that's not very important for the big picture here. They are investing billions because they think they can reach AGI soon (and change the whole economy), not because of their current chat app numbers. In fact, they are investing billions in companies that only do research for now.

In order to have profitability for AI, you need a couple things. 1) You need cheap API access for data training 2) You need low energy costs 3) Independence from Nvidia (Google uses TPUs) 4) Scale 5) Customers 6) Focus on efficiency Google is doing that in spades.

They have their own chips that they've designed specifically for AI. They have been doing this AI thing for a long time now despite the recent hype, and they've been able to implement it profitably for some tasks. Also Gemini is becoming much much more efficient than competitors including deepseek.

no way, Google already failed miserably by putting AI search results above regular search results, even when the AI search is frequently wrong and we cannot opt out of it.

Their AI search results are not great at all, but they've made Gemini 2.0 very cheap to use, even more than deepseek, and that's not them trying to take a loss on every prompt. They've found a way to make Gemini more efficient while being on par with many of the other models.

I'm no fan of LLMs and generally despise them. But...what is the point of this? Without the thing that eclipses Pokemon Go generative AI isn't even as popular as Pokemon Go? ChatGPT is by far the most well known and talked LLMs and no one talks about LLMs without talking about them...so of course?

ah if you knew what a balm these words are to my soul in a time of many torments knowing that these fucking pinheads are going to get their nuts slammed in the car door of consensus reality is soothing

yeah it’s just a feeling that the models are more useful b2b than b2c, but i don’t have a reference