There is great debate within the financial and technology circles. Are we in an AI bubble?
The short answer to that is "no".
In this article I will go through what I believe to be some relevant points regarding the advancement of this technology.
Compute Versus The Internet
A common practice is to try and compare the progression of AI (or SI) with the Internet. While this might make sense since it was an innovation that changed the world, the comparison doesn't apply.
The Internet is nothing more than networking. Basically, it is the tying together of computers to allow the sharing of information. Of course, there are many layers to it but that is a simplified version of it. Thus, the rollout (and adoption) of the Internet was based upon a series of factors.
When it comes to SI, we are dealing with compute. This means the adoption curve is going to resemble the usage rates of the 1960s-1990s for early stage comparisons. Naturally, since the turn of the century compute has exploded, especially after the introduction of the smartphone.
Cost Curve and Adoption
What happens when costs drop? Typically, demand increases. This is a crucial factor in the discussion.
We had debate whether the data centers being built will get utilized. Leading the charge is Elon Musk, via SpaceX. That company is looking to secure as much compute as possible. Unlike other entities. this is not for the sole purpose of its models. Instead, SpaceX is building a service similar to AWS. They are in the rental game.
There are two factors at work.
First, we have the technology adoption curve. This is a generalization of what happens when new technologies are introduced. The early stages are technologically inclined individuals. Here we see flawed products or services. Over time, things improve which brings in more users. As ease of use is implemented, the average person starts to participate.
Another factor is the cost curve. When more supply hits the market, costs come down. Of course, if demand is increasing at a faster rate, this will be delayed. That is what we are seeing now. The cost for compute is increasing due to the fact that supply is outpaced by demand.
Will It Continue?
The multi-trillion dollar question is whether the demand will continue. Here is where the debate enters.
My answer stems from the fact we are dealing with a general purpose technology. This is not a single product or service. Instead, it is akin to electricity. When we view things through this prism, we can see how, over time, demand keeps increasing.
Companies all over the world are developing applications tied to this technology. The idea of "companies making money from SI" is an improper way to look at it. Going back to the electricity idea, who makes money there? Of course, the producers, i.e. utilities, earn profits. However, they are the lowest end of the stack. Some of the highest margin companies in the world make their money from electricity. This is what powers the digital world.
In other words, Google's entire profit structure is based upon electricity usage.
The same will be with SI. Companies that are providing the applications that are used by hundreds of millions (or billions) will be the winners.
Each week more utility is rolled out. Simply put. there are a lot more services available that have SI integrated as compared to a year ago. We can expect the same to be true in another 12 months.
Financial World
In closing, we have to discuss the idea of a bubble. For the most part, this is a financial concept, referring to markets. Are SI related companies overvalued?
That is not the question I seek to answer. Markets are markets. They can do some strange things. Is SpaceX worth $2 trillion? Does $1.7T for Anthropic make sense? Will OpenAI even survive (my guess is no)? All of these are legitimate questions for investors and market participants.
The answer I am providing is the idea that the technology is ahead of itself. We are looking at things such as hype, which certainly feeds into markets. My conclusion is the speculation around this technology is legitimate. Utility will keep growing over time. It will be years before the supply of SI compute outpaces demand.
Again, we are dealing with a general purpose technology. Like electricity, it will eventually shift into the background. People will use it without consideration. After all, when one turns on Netflix, does he or she consider that electricity is being used? Of course, we are aware of that but few think about it at the time.