If I'm understanding the study correctly, I'm pretty sure that ChatGPT use is still growing, but the rate of growth appears to have slowed. I'm not quite sure whether this is a big deal -- download growth also appeared to fall last December to January -- but it is a useful piece of context for sure.
Yes, if people were shifting to (say) Claude that wouldn’t be a big problem for the AI industry. But I suspect that lots have gone back using Google with Gemini, which is great for Google but doesn’t bring in any extra revenue.
It is great to see such carefully prepared work, rather than hype people vs skeptics making narrow arguments.
All of this still hinges on AI being able to improve and deliver as fast as many companies hope. I think some disappointment and some losers are quite likely, especially among weaker players and the ones who overextend themselves. Even OpenAI may not be immune, given just how much they raise the stakes.
Great visuals. It makes all of these complex trends easy to see.
The first chart is confusing me, though. Is it saying that Meta is spending $400 billion in capex this year? Or is that the cumulative capex of all of the companies?
The chart is showing the cumulative capex. So Amazon + Microsoft + Google + Oracle + Meta spent at a $400 billion annualized rate in Q2 (97 billion for the quarter). But Meta is only the bit between the Meta line and the Oracle line.
This is a very annoying quibble on my part, but the header "money burning machine" under #13 is a bit ruthless. If you extended the chart one additional year to the right, into 2031, the title would have to be changed to "money printing machine" for the chart to make sense.
I certainly don't disagree with your primary point which is that OAI's operating losses to breakeven are large and (for now) unfunded.
Thanks for explaining this so well, it feels like this AI boom has more energy then my most intense Pilates class! Do you see this capex trend continuing indefinitely?
Good catch! Chart 1 is annualized while chart 12 is quarter by quarter amounts. The chart would look the same, just with different labels, if on an annual basis.
Excellnt compilation! Chart #12 really stands out for Oracle investors - the comparison showing Oracle with negative free cash flow while accumulating debt through aggressive buybacks is concerning. When you combine that with the massive capex requirements for AI infrastrucutre, Oracle's financial position looks much more precarious than the other tech giants. The fact that Amazon can have negative FCF sometimes but still be in "excellent financial health" while Oracle is singled out for debt accumlation tells the whole story. Great work visualizing the AI boom's winners and potential losers.
Thank you for the excellent overview. I wasn't aware that the electricity prices are a major advantage that the US has over the EU in data center construction. It's mind-blowing that commercial electricity prices in the EU averaged €0.19 per kWh, around double the comparable US rate.
Every AI article ends up concentrating on about 6 stocks. The same ones. There isn't much to learn. Can someone do a write up on other investment ideas please.
ChatGPT use is reported to have peaked https://futurism.com/artificial-intelligence/chatgpt-peaked-data
If I'm understanding the study correctly, I'm pretty sure that ChatGPT use is still growing, but the rate of growth appears to have slowed. I'm not quite sure whether this is a big deal -- download growth also appeared to fall last December to January -- but it is a useful piece of context for sure.
(https://techcrunch.com/2025/10/17/chatgpts-mobile-app-is-seeing-slowing-download-growth-and-daily-use-analysis-shows/#:~:text=To%20be%20clear%2C%20this%20is%20a%20look%20at%20download%20growth%2C%20not%20total%20downloads%2E)
It would be useful to have aggregate data, as ChatGPT isn't the only game in town.
Yes, if people were shifting to (say) Claude that wouldn’t be a big problem for the AI industry. But I suspect that lots have gone back using Google with Gemini, which is great for Google but doesn’t bring in any extra revenue.
It is great to see such carefully prepared work, rather than hype people vs skeptics making narrow arguments.
All of this still hinges on AI being able to improve and deliver as fast as many companies hope. I think some disappointment and some losers are quite likely, especially among weaker players and the ones who overextend themselves. Even OpenAI may not be immune, given just how much they raise the stakes.
Great visuals. It makes all of these complex trends easy to see.
The first chart is confusing me, though. Is it saying that Meta is spending $400 billion in capex this year? Or is that the cumulative capex of all of the companies?
The chart is showing the cumulative capex. So Amazon + Microsoft + Google + Oracle + Meta spent at a $400 billion annualized rate in Q2 (97 billion for the quarter). But Meta is only the bit between the Meta line and the Oracle line.
Chart for #11 has incorrect sequencing of months. Another case of using AI without adequate supervision?
Haha good catch. That was entirely human error. We will fix it.
This is a very annoying quibble on my part, but the header "money burning machine" under #13 is a bit ruthless. If you extended the chart one additional year to the right, into 2031, the title would have to be changed to "money printing machine" for the chart to make sense.
I certainly don't disagree with your primary point which is that OAI's operating losses to breakeven are large and (for now) unfunded.
Thanks for explaining this so well, it feels like this AI boom has more energy then my most intense Pilates class! Do you see this capex trend continuing indefinitely?
Chart 12? Capex only 100 bil? That an avg amongst hypers?
Good catch! Chart 1 is annualized while chart 12 is quarter by quarter amounts. The chart would look the same, just with different labels, if on an annual basis.
I've now updated the chart to be annualized like the other charts
Quarterly! Ahh yes ty
Excellnt compilation! Chart #12 really stands out for Oracle investors - the comparison showing Oracle with negative free cash flow while accumulating debt through aggressive buybacks is concerning. When you combine that with the massive capex requirements for AI infrastrucutre, Oracle's financial position looks much more precarious than the other tech giants. The fact that Amazon can have negative FCF sometimes but still be in "excellent financial health" while Oracle is singled out for debt accumlation tells the whole story. Great work visualizing the AI boom's winners and potential losers.
Excellent charts and explanations. Thanks for sharing this.
Thank you for the excellent overview. I wasn't aware that the electricity prices are a major advantage that the US has over the EU in data center construction. It's mind-blowing that commercial electricity prices in the EU averaged €0.19 per kWh, around double the comparable US rate.
Thanks for sharing
I wonder what's the main reason they don't build datacenters in the north
Every AI article ends up concentrating on about 6 stocks. The same ones. There isn't much to learn. Can someone do a write up on other investment ideas please.