This week’s conversation is with Harmol Samra, CEO of data center developer Host Digital. I asked Samra for a conversation because his team expressed a willingness to offer candid views on the industry’s problems with social license to operate — namely, why is the industry botching this so badly? From his point of view, many could learn a lot from industrial revolutions of the past, and they should be careful not to let mega-project expansions get in the way of broader acceptance of the data center sector.
The following conversation was lightly edited for clarity.
Can you start with your experience right now trying to get permission to build and operate data centers in the U.S.?
What we do is we focus on data centers between 20 megawatts and 100 megawatts — we call them “right-sized” — that are currently energized and delivering in the next 10 to 14 months. So really focusing on the front end of that curve. To your question of social license, we’re focused on building the right-sized data centers that communities are used to seeing and have been seeing for the last 20 years. Very similar to any other industrial warehouse you would see in your community in a business park or things like that. We’re not building multi-gigawatt campuses in a community that has 20,000 people and has an overarching, overbearing impact.
On the social license piece of it, we understand that this is the base layer of infrastructure we’re building. We’re providing the base layer as long as we agree that we all want to store our data for cloud usage, we do need to train our models, we all want to interact with AI and have some imprint — that’s what we support.
On the community benefits piece, we don’t ask for any tax breaks or incentives anywhere we go. In fact, we do the opposite — in most cases we set up community benefits programs. Our data center might become operational in a year, and that’ll pay real estate taxes, but what about in between now and then? In instances, it’s been a new AI training program in high school. In other instances, it has been [people saying], “AI’s great, but using it is expensive, so what if we could provide tokens so students can learn?”
Can you give us an example of a time you were challenged in developing data center infrastructure? Help us get a better understanding of how you’ve faced this backlash yourself.
Our first site in northeast Oklahoma, in Tulsa, enacted a moratorium on data centers after we purchased a site and were going through our leasing and financing process for it. It was a bitcoin mining site we were converting to a data center, in this case. We went to engage with the community, with the city planning office, and with the city council members. Unanimously, by 9-0, we got an exemption [from their moratorium] for us to be able to go build their data center.
When we went and made our case, what we are doing in our strategy was retrofitting this site. Not really adding new load to the system. They said, “The moratorium was not put in place to capture developments like you.” It was put in place for the mega campuses. It was about walking them through our strategy with a site that has been drawing power for the past few years.
Do you feel retrofitting sites changes the perspectives of local officials?
Yes it does, because there are a couple of things you can address up front. We’re not asking for new load; 55 megawatts has been drawn to the site for the past few years. The building we’re converting has been there for years. It calms people’s fears about a new data center coming. It’s just a building that has its computers.
It becomes easier and more bite-sized to digest. Now, obviously there are no gigawatt-scale things you can retrofit.
What do you think other energy sectors can teach the data center sector from their experiences dealing with backlash?
Overarching, as an industry, we have to be net positive to a community.
Other energy sectors and players have worked with their actual impacts to the climate and to the community, and that’s what we have to learn. If you build a data center, what is the social benefit? Through jobs? Through revenue? Through infrastructure they need? It’s the point we need to hammer home.
What do you think of some of the strategies raised thus far to improve community benefits? I’ve heard some new ideas like data center dividends, and I wonder what you make of all the proposals floating out there.
One idea is tax offsets.
Let’s say you’re going to build a data center in a town with 10,000 people. They’re all paying on average real estate taxes of $3,000. That’s a total impact of $3 million that people are paying out of their pocket. Now, if that can be supplemented by having a new data center there, offsetting that entire dollar amount, that is a direct benefit to people living there. It’s a net benefit to local residents.
There’s a reason some of the impact plans we are seeing today are direct and dollar-denominated, right? We can offset taxes, pay for our own infrastructure upgrades, add new generation to the grid — wherever we build a data center, we can build a new renewable asset and bring energy back to the grid. Frankly, some of these ideas have resonated with communities.
Why do you think the data center unrest and backlash is happening, then?
I think the biggest reason we’re seeing that is that data centers are a physical embodiment of what artificial intelligence is. It’s the base asset for AI to work. There’s a lot of unease about how we perceive AI will impact our society and our lives and livelihoods, and its harder for communities to tell an Anthropic how to change their models. But they do think if they stop the buildout of a data center, they can stop the expansion of AI until we all figure out where it goes.
Some of the factual opposition in the beginning was high electricity usage that might impact people’s bills. That’s been addressed. Excess water? That’s been addressed. Noise impacts? You can put cooling equipment outside and wrap it in a way with sound barriers where it doesn’t impact.
The factual impacts, Jael, I think have been addressed. It’s more a corollary function of how people feel about AI — that with the data centers, people see a physical asset and they wonder how to stop that.
