Apoorv, great analysis! I’m wondering if we can translate your blog into Chinese and post it in AI community. We will highlight your name and keep the original link on the top of the translation. Thank you!
I can't figure out how to restack images, in the app at least, but I've been saying since early last year that the lion's share of revenue will be in the apps people develop that make AI as easy to use as every other kind of software now works, without having to learn prompt engineering.
We're only just barely starting to see that, as UIs slowly start to improve and non-geeks like us start using them, and that pyramid on the left will get far, far more top heavy.
I am not sure if this is correct comparison because of timing and maturity. If you last cycle - late 90s during internet build out - Cisco revenue higher compares to app layers (google, meta, etc).. as infra maturity is achieve then value moved to app layer. same will happen in AI - we are yet to see 'good' app and edge devices. currently its still AI infra build out phase, within couple of years (later in cycle) - value creations will be in app layer.
Great analysis, and interesting visualizations, i liked the triangles you used when splitting the value between semi, infra and software.
It will be interesting too see what will happen from 2025 and forward, as the heavy lifting made between 2023-2025, but cloud services will begin to offer more diverse solutions of AI hardware, and not just Nvidia's chips. Not that it needs to worry about Nvidia's future, but I don't think this moat will last forever.
I also wrote about a similar thing, more of a high level overview of the semi-conductors industry, with a section about AI chips, but you gave much more details about this specific vertical. well done!
Couldn’t find article you referenced on value shifting to tech enablers. Can you please explain what you mean by that and article by Vivek you are referring to?
Most of it resonated immediately and analysis was interesting throughout, thank you!
is best way to analyze the value capture of mobile by looking at the stock price? Not clear to me that apple was only second tier winner here. Also the categorizing of Amazon as "mobile internet" companies only seems to be off and skew results there. Maybe another cut would be to look at how the profit pools have shifted overtime between categories.
I'm not convinced the app layer will make a good investment, even as it accrues more value over time. GenAI is making software development easier, which will lead to more fragmentation at the app layer, and more competition in every category.
Apoorv, great analysis! I’m wondering if we can translate your blog into Chinese and post it in AI community. We will highlight your name and keep the original link on the top of the translation. Thank you!
Sure thing go ahead!
I can't figure out how to restack images, in the app at least, but I've been saying since early last year that the lion's share of revenue will be in the apps people develop that make AI as easy to use as every other kind of software now works, without having to learn prompt engineering.
We're only just barely starting to see that, as UIs slowly start to improve and non-geeks like us start using them, and that pyramid on the left will get far, far more top heavy.
I am not sure if this is correct comparison because of timing and maturity. If you last cycle - late 90s during internet build out - Cisco revenue higher compares to app layers (google, meta, etc).. as infra maturity is achieve then value moved to app layer. same will happen in AI - we are yet to see 'good' app and edge devices. currently its still AI infra build out phase, within couple of years (later in cycle) - value creations will be in app layer.
Give it time. The wheel was immensely valuable thousands of years ago, it is mobility now. The world is still shaping itself around the AI railroads
Great analysis, and interesting visualizations, i liked the triangles you used when splitting the value between semi, infra and software.
It will be interesting too see what will happen from 2025 and forward, as the heavy lifting made between 2023-2025, but cloud services will begin to offer more diverse solutions of AI hardware, and not just Nvidia's chips. Not that it needs to worry about Nvidia's future, but I don't think this moat will last forever.
I also wrote about a similar thing, more of a high level overview of the semi-conductors industry, with a section about AI chips, but you gave much more details about this specific vertical. well done!
https://reassembler.substack.com/p/inside-the-chip-wars-power-profit-fight-over-the-new-oil
Amazing analysis with nuanced focus on core drivers that would create the momentum for curve inversion!
Couldn’t find article you referenced on value shifting to tech enablers. Can you please explain what you mean by that and article by Vivek you are referring to?
Most of it resonated immediately and analysis was interesting throughout, thank you!
is best way to analyze the value capture of mobile by looking at the stock price? Not clear to me that apple was only second tier winner here. Also the categorizing of Amazon as "mobile internet" companies only seems to be off and skew results there. Maybe another cut would be to look at how the profit pools have shifted overtime between categories.
I'm not convinced the app layer will make a good investment, even as it accrues more value over time. GenAI is making software development easier, which will lead to more fragmentation at the app layer, and more competition in every category.