Alberto Lamadrid on Data Centers

Image by iStock/Gri-spb

In this episode of Lehigh University’s College of Business ilLUminate podcast, host Stephanie Veto talks with Alberto Lamadrid about the rise in data center production. Alberto is a professor in the Economics Department, part of the Integrated Networks for Electricity Research Cluster and the Institute for Cyber Physical Infrastructure and Energy. 

Listen to the podcast here and subscribe and download Lehigh Business on Apple Podcasts or wherever you get your podcasts.

Below is an edited excerpt from the conversation. Read the complete podcast transcript [PDF]. 

Veto: You study energy systems and electricity economics. What's driving the growth of data centers right now? 

Lamadrid: One of the main drivers is the need that we have for artificial intelligence. This is a need that is happening on both the supply side and the demand side. So, on the supply side, there are several companies that have been growing in terms of capitalization, and they are trying to come up with new and innovative models. They're going to be using increasing amounts of power. I assume that listeners have probably used ChatGPT or have used Claude or any of these kinds of artificial intelligence systems. Part of it is, how do you train that system? And training that system is very energy intensive. 

On the other hand, we have the demand side. All of us are using artificial intelligence. I am a faculty member in the economics department. Our students are learning how to use artificial intelligence. Some of them are going to be using it for their homework, for doing research. Some of them are going to be using it for personal questions. A lot of people are going to be using it for planning their next vacation or whatever. You're going to be having increases on the supply side and on the demand side. And all that goes to increase the amount of electricity that you're going to be using because all these systems are going to be running on that infrastructure. 

Veto: What makes AI and cloud computing different in terms of energy demand? Why do we need these huge facilities? 

Lamadrid: First, these are completely new models. These AI data centers are not your grandmother's data centers. We've been using data centers for a very long time. Before, you had a data center that were basically a couple of servers that provide local services. So, they're going to be, for example, for a company if they need to process their local payroll or whatever it is. The AI data centers, on the other hand, measure football fields. They're going to be having racks and racks of servers that are going to be used to process. 

There are two main things that artificial intelligence requires. One is the training of the model that is going to be ingesting massive amounts of data to come up with a model that is going to be useful. And the other part is called inference. There’s a lot of electricity used for training, and you need large data centers to do it. Some of them are going to be doing the process in parallel. Some of them are going to be waiting for some of the processes to end. That is different from the previous way that we were doing processing. 

The previous revolution that was like this is when we were expanding internet services. However, a Google search is much smaller than a Gemini. In that sense, this is a fundamental change in terms of the amount of electricity and the facilities required to perform them. 

Veto: What are some of the environmental concerns about these big data centers? 

Lamadrid: Because it's computing processing power, you're going to be needing electricity. You need to power those data centers. There's going to be direct and indirect use of water. You need to have water use because you need to cool down the computers that are going to be doing calculations. Also, you are going to be using all that electricity to generate the plants that are going to be producing that electricity. They also need to be cooled down. 

We produce electricity in Pennsylvania. So, we are part of a market. There is not a central authority that owns the plants and then decides to dispatch something–it is a restructured system. We use a lot of natural gas in the plants here, and this is a legacy of Pennsylvania. So, there's going to be pollution if you're going to be producing electricity using natural gas. However, it's much more benign in terms of the number of emissions than other forms of producing electricity like coal. But it still has emissions. So, there's going to be those kinds of concerns in terms of what is going to be the emissions being produced by the production of electricity. 

There is emissions or pollution that is coming from the noise, too. One thing that happens with a lot of these data centers is that they have an uninterrupted power supply next to the data center. That uninterrupted power supply typically is diesel-powered. So, you never use it, or you use it very rarely, because here we have a very reliable electricity system. So, you're going to be mostly buying your electricity from the network, the grid. But every month or every couple of weeks, you're going to be testing that this backup equipment is working. You’re going to be running those diesel generators. So, there's going to be the emissions that are going to be produced during that period. And there's going to be the noise that they're going to be producing. So, there's going to be some noise pollution. That can be an environmental concern as well. 

There's another way in which you can think about this. So, I've been talking about pollution that happens in the air because of the production of electricity. They're also going to be using a lot of land because these are very large data centers. You might have a forest and you're going to be removing the trees. In certain cases, there's not going to be a lot of absorption of emissions, local pollutants or things like that. So, there's going to be that consequence of having impermeable surfaces or things that cannot be cooled down. You might have a little bit of urban heat island effect in this case because of the data center.