There are many components to the data center conversation: environmental concerns, job sustainability, transparency, and community impact. Until we have a tangible understanding of what it means to “win” AI, HKS’s Rachel Mural argues that it will be difficult to create a strategy to address the questions and concerns surrounding AI.
Rachel Mural: The data center pushback isn't surprising. When we think about data centers, we can effectively think of them as just warehouses that are, you know, housing these servers and these compute centers. These chips take a lot of power to operate, especially the very cutting edge models. With the power comes a lot of heat that they give off, and because of that, it takes power both to run these chips themselves, but then it takes power to also make sure that the chips stay cool enough to be able to run these processes.
A lot of these data centers use a water based cooling system where they can run the water just right up against the servers to keep those chips cool. The challenge that we've come across in our work, especially when we started, was there was very limited transparency in a lot of these researchers in general. It's still very hard to get the data. And I think in some ways, some of these moratoriums are also attempting to address that by putting in your new zoning and citing regulations that you have to disclose your energy, your water, everything else.
It also gives the communities a way to really understand the impact beyond just what they can kind of guess. The questions around employment, for example, are really hard to answer right now. We can see that there are jobs that come in with data centers, but I think there are still a lot of questions around whether those are just short term construction jobs, whether those are really growing the economy long term. The kind of data center being built really matters.
But then also things like how many other data centers might be in that immediate area are also impacting what the job site looks like. It's hard to say really whether the impact of the energy, environment, water, everything else versus that job creation is really balanced. One thing that I think is sort of missing from this conversation is what does it actually mean to win the AI race? That's a really challenging thing to put a metric on.
You know, I think when we think about technology races in the past, I think the natural analogy is the race to the moon, which had a very tangible endpoint. If you were the one that got your country safely to land on the moon and you had an astronaut step foot on the moon, you won the space race. With AI, we don't really have something that's that tangible or easy to really understand as a win. Until we really understand what does it mean to win AI, building any sort of national strategy around that, and dealing with the repercussions of the data center impacts and everything else is just gonna be that much harder.