What is the future?

In this article, I will answer this question by focusing upon the infrastructure that we are seeing built out.

AI is a paradigm shift. It will affect everything in society. For this reason, I think it is important to understand how things are being altered.

This is an evolving process. Things are moving rapidly, with the world advancing at a pace it never saw before. Consider this point, GPT3 from Open AI came out on November 30th, 2022. We are not even to the 3 year mark.

Now ponder how much things changed in the digital world since then. Extrapolate that out into the "real" world to see the impact that occurred.

Of course, as they say, you ain't seen nothing yet.

Source

GPU-as-a-Service: The Future

GPU-as-a-Service is the future. This is something that should be logical to conclude.

CPUs are no longer adequate to handle out massive computing needs. Naturally, at the local level, there is little issue. However, we we work our way up the scale, we see that anything that is driving AI has to be on GPUs (or Google's TPU).

There is no other way to handle deep learning or advanced inference. This is because massive parallel processing is crucial.

It is exactly what GPUs provide. Because of that Jensen Huang is now akin to a rock star. Nvidia is the most valued company in the world. Earnings are expected to keep increasing. In 2026, we will see Blackwell 3 rolling out, a chip that should fare better than the introduction of the B200.

Ultimately, this is the basis for the future. Everything will be built atop of this.

Digital Highway System

We are going to see everything tied to the digital world. Over the next decade or two, the real world will become digitized.

For now, we can focus upon something such as robots. As those roll out, they will be connected to the digital realm. AI systems are being integrated into them.

What do they run on? GPUs of course.

Here is where the service starts.

The other day, Elon Musk and Jenson Huang announced an agreement where xAI will become a customer of Nvidia's project with Humain. Essentially, xAI is renting compute from this company, incorporating more GPUs into its framework.

Why would Musk do this?

Simply put, he needs the compute. XAI, along with a host of other companies, are buying chips as quickly as possible. The problem is that one cannot simply put 100K GPUs in a warehouse and turn them on. Energy remains a problem.

Under this agreement, Humain is building the data center in Saudi Arabia, handling the energy set up. Once this is live, it simply rents the compute out, similar to how cloud companies operated for most of the last decade.

Think of this as the national highway system. Roads are built which fosters other businesses. In some places, the roads are "rented" in the form of tolls.

GPUs are the roads of tomorrow. They are the basis of all that is taking place. Companies all over the world are building out data centers, with many others looking at the energy problem.

Companies such as xAI will still build their own. However, since enough is not available, they will rent space from those who have it.