[AI](https://inleo.io/@leoglossary/leoglossary-artificial-intelligence) is going to generate fortunes like we never saw before.


That said, everything that is related to AI is not going to make [money](https://inleo.io/@leoglossary/leoglossary-money).  This is the lesson from the Dotcom era.  Many thought that simply because a [company](https://inleo.io/@leoglossary/leoglossary-company-business) was "on the Internet" or "developing an app" that it was going to be worth a fortune.


In reality, most of what was out there was garbage.


For that reason, we will take a look at the AI stack to see where the money is going to be made.    In my mind, it shapes up like this.


<center> https://ideogram.ai/api/images/direct/EPXAUDZQS4-ih5V6mkRE4A.png
image made using Ideogram </center>




## Hardware


Here is the money printing machine.


It is the [world](https://inleo.io/@leoglossary/leoglossary-world) of [NVIDIA](https://inleo.io/@leoglossary/leoglossary-nvidia-company) and whomever else is making AI chips.  For now, GPUs dominate which makes NVIDIA the clear cut winner.  Others like [Tesla](https://inleo.io/@leoglossary/leoglossary-tesla) and [Google](https://inleo.io/@leoglossary/leoglossary-google-company) design their own chips.  There also could be newer [technology](https://inleo.io/@leoglossary/leoglossary-technology) relating to this that causes a massive breakthrough.


Whatever it is, this is the engine that drives everything.  Compute, aka processing [power](https://inleo.io/@leoglossary/leoglossary-power), is going to be required and those that make this hardware are going to do well.


## Infrastructure


This might not be as lucrative as chip making yet it is easier to get into.  Whereas the first will have a few major players, this is more open.


What are some examples?


We have agent orchestration tools.  This also would include [software](https://inleo.io/@leoglossary/leoglossary-software) designed to evaluate what is going on.  Then there is deployment.


Basically, we are looking at all that is inserted in the hardware to help produce the outcomes.


## LLMs


The next up is the Large Language Models themselves.  This is the level where things can get interesting.


Personally, I foresee a [time](https://inleo.io/@leoglossary/leoglossary-time) when these are commoditized.  At some point, the ability to separate from everything else is going to be difficult.  That are basically all training on the same [data](https://inleo.io/@leoglossary/leoglossary-data), which then will lead to similar synthetic data, resulting on similar systems.


At present, we see the arms race because we are early in the game.  However, as more play around with the technology, especially with [open source](https://inleo.io/@leoglossary/leoglossary-open-source) systems out there, and it is easy to see how this could flip to a [commodity](https://inleo.io/@leoglossary/leoglossary-commodity).


## Applications


Like the first two [layers](https://inleo.io/@leoglossary/leoglossary-layer-insurance), we can see enormous value here.


What I find fascinating is those AI companies that already have working products could see that value enhanced by creating the LLMs.  Nevertheless, it is the integration where the money will be made.


For Meta, this is integrating Llama into all its products.  Google is doing something similar.  We are likely going to see some of these [building](https://inleo.io/@leoglossary/leoglossary-building-structure) robots, like Tesla and [OpenAi](https://inleo.io/@leoglossary/leoglossary-openai), which will utilize the LLMs.


This is what most will focus upon, including those in [Web 3.0](https://inleo.io/@leoglossary/leoglossary-web-3-0).


## Web 3.0 Applications


The key, to me, is to build as much of this stack as we can.


Obviously, the hardware layer is going to be difficult.  Perhaps would could end up seeing distributed computing coming into play here.  However, can this be used to efficiently [train](https://inleo.io/@leoglossary/leoglossary-train-railroad) these models.  We are looking at a ton of compute to even begin to rival what Big Tech is doing.


That said, the rest of the structure can be addressed to varying degrees.


The central premise is going to be eventually offering services.  AI is like anything else, people will not use it if there is no need.


So far, with Web 3.0, that is the case.  What services are being offered?  That is the billion dollar question, one which there really isn't an answer.


This could quickly change, especially as AI penetrates, well, everything we do in the [digital](https://inleo.io/@leoglossary/leoglossary-digital) world.  This is the very early stages and [blockchain](https://inleo.io/@leoglossary/blockchain) could be a great ally in this endeavor.


For this reason, Web 3.0 applications can certainly [leverage](https://inleo.io/@leoglossary/leoglossary-leverage) the technology.  An example of this is AI personal assistants.  Naturally, this is not going to be all on public [databases](https://inleo.io/@leoglossary/leoglossary-database).  However, these could be offered within Web 3.0 applications that incorporate different aspects including [cryptocurrency](https://inleo.io/@leoglossary/cryptocurrency) and decentralized data.


Each layer of the stack offers some potential except, ironically, the LLMs.  If things like Grok and Llama become truly open source, as well as rivaling what else is out there, then we could see something completely different in a few [years](https://inleo.io/@leoglossary/leoglossary-year).


The race might be to become the standard AI operating system and let everyone else build value for the model, which is then integrated into that company's products.


Interesting times to say the least.


____


<center> [What Is Hive](https://inleo.io/@leoglossary/leoglossary-what-is-hive) </center>


Posted Using [InLeo Alpha](https://inleo.io/@taskmaster4450le/the-ai-stack-and-where-the-money-resides)