Help me understand where, from your perspective, the generation for new data centers is going to come from. I know there are gas turbine shortages, but also that solar and wind are dealing with headwinds in the United States given cuts to the Inflation Reduction Act.
There are a lot of stories out there about certain technologies coming out to the forefront to solve the problem, whether it’s gas generation or something else. But the scale and the scope of this stuff … I don’t think there is a silver bullet where it’s all going to come from one place.
The Energy Department put out a request for information looking for ways to get to 3 gigawatts quickly, but I don’t think there is any way to do that quickly in the United States. It’s going to take work from generation developers, batteries, thermal generation, emerging storage technologies, and transmission. Reality is, whether it is supply chain issues or technology readiness or the grid’s readiness to accept that load generation profile, none of it is ready. We need investment and innovation on all fronts.
How do conflicts over siting play into solving the data center power problem? Like, how much of the generation that we need for data center development is being held back by those fights?
I do have an intuitive sense that the local siting and permitting concerns around data centers are expanding in scope from the normal noise and water considerations to include impacts to energy affordability and reliability, as well as the selection of certain generation technologies. We’ve seen diesel generation, for example, come into the spotlight. It’s had to do with data center permitting in certain jurisdictions, in places like Maryland and Minnesota. Folks are realizing that a data center comes with a big power plant – their diesel generation. When other power sources fall short, they’ll rely on their diesel more frequently, so folks are raising red flags there. Then, with respect to gas turbines or large cycle units, there’s concerns about viewsheds, noise and cooling requirements, on top of water usage.
How many data center projects are getting their generation on-site versus through the grid today?
Very few are using on-site generation today. There’s a lot of talk about it and interest, but in order to serve our traditional cloud services data center or AI-type loads, they’re looking for really high availability rates. That’s really costly and really difficult to do if you’re off the grid and being serviced by on-site generation.
In the context of policy discussions, co-location has primarily meant baseload resources on sites that are serving the data centers 24/7 – the big stories behind Three Mile Island and the Susquehanna nuclear plant. But to be fair, most data centers operational today have on-site generation. That’s their diesel backup, what backstops the grid reliability.
I think where you’re seeing innovation is modular gas storage technologies and battery storage technologies that try to come in and take the space of the diesel generation that is the standard today, increasing the capability of data centers in terms of on-site power relative to status quo. Renewable power for data centers at scale – talking about hundreds of megawatts at a time – I think land is constraining.
If a data center is looking to scale up and play a balancing act of competing capacity versus land for energy production, the competing capacity is extremely valuable. They’re going to prioritize that first and pack as much as they can into whatever land they have to develop. Data centers trying to procure zero-carbon energy are primarily focused on getting that energy over wires. Grid connection, transmission service for large-scale renewables that can match the scale of natural gas, there’s still very strong demand to stay connected to the grid for reliability and sustainability.
Have you seen the state of conflict around renewable energy development impact data center development?
Not necessarily. There is an opportunity for data center development to coincide with renewable project development from a siting perspective, if they’re going to be co-located or near to each other in remote areas. For some of these multi-gigawatt data centers, the reason they’re out in the middle of nowhere is a combination of favorable permitting and siting conditions for thousands of acres of data center building, substations and transmission –
Sorry, but even for projects not siting generation, if megawatts – if not gigawatts – are held up from coming to the grid over local conflicts, do you think that’s going to impact data center development at all? The affordability conversions? The environmental ones?
Oh yeah, I think so. In the big picture, the concern is if you can integrate large loads reliably and affordably. Governors, state lawmakers are thinking about this, and it’s bubbling up to the federal level. You need a broad set of resources on the grid to provide that adequacy. To the extent you hold up any grid resources, renewable or otherwise, you’re going to be staring down some serious challenges in serving the load. Virginia’s a good example, where local groups have held up large-scale renewable projects in the state, and Dominion’s trying to build a gas peaker plant that’s being debated, too. But in the meantime, it is Data Center Alley, and there are gigawatts of data centers that continue to want to get in and get online as quickly as possible. But the resources to serve that load are not coming online in time.
The push toward co-location probably does favor thermal generation and battery storage technologies over straight renewable energy resources. But a battery can’t cover 24/7 use cases for a data center, and neither will our unit. We’re positioned to be a bridge resource for 24/7 use for a few years until they can get more power to the market, and then we can be a flexible backup resource – not a replacement for the large-scale and transmission-connected baseload power resources, like solar and wind. Texas has benefited from huge deployments of solar and wind. That has trickled down to lower electricity costs. Those resources can’t do it alone, and there’s thermal to balance the system, but you need it all to meet the load growth.
The model for the evening of June 27. Darker colors indicate higher PM2.5 levels, the particles associated with wildfire smoke. The numbered circles refer to the number of regional wildfires.
The model for the morning of June 28.
The model for the evening of June 28.
