The disconnect between the companies looking to fast-track the development of AI data centers and the communities/politicians fighting back against them is massive.
The likes of Microsoft, Meta and Amazon Web Services (AWS) stand out as some of the major tech institutions that are funneling money into the continued expansion of AI data centers on a national scale. But with those fully built data centers comes a negative impact on our environment, with precious fossil fuels being used to power them and a huge strain on our power grid resulting from their computing demand.
It's certainly encouraging to see a collective of politicians and the people living in the towns where these AI data centers are pushing back against them. And with a coalition of states placing moratoriums on AI data centers, that opposition is only getting stronger.
To get more insight into the growing negative sentiment towards AI data centers, what could happen if the US builds them faster instead of responsibly and more, we spoke to Balaji Tammabattula, an AI infrastructure engineering expert and COO of BaRupOn.
Delving into the development and rising backlash against AI infrastructure
Tom’s Guide: Why do you think the growing backlash against data centers was inevitable? And where do you think the AI industry got it all wrong?
Balaji Tammabattula: I think the backlash was inevitable because the industry focused heavily on how quickly we could build compute and not enough on what was required to support it.
When a large data center comes into a community and suddenly needs hundreds of megawatts of electricity, water and infrastructure, people naturally ask, "What does this mean for us?"
Those are legitimate questions.
At LAMP in Texas, we've taken a different approach. We're developing dedicated behind-the-meter power generation so that the campus is designed around producing the power it needs rather than depending on existing community grid capacity.
We're applying the same thinking to water and public infrastructure.
Responsible AI infrastructure shouldn't arrive in a community and ask, "What resources can we take?" The better question is, "What infrastructure can we build, and what can we leave behind?"
That's where I think the industry needs to go.
Tom’s Guide: What do you see as the biggest misconception the data-center industry has about the local communities it wants to build in?
Tammabattula: That communities are simply resistant to development. They're not. People want to understand the tradeoff.
If you're bringing a massive project into their community, they want to know whether their electricity rates will be affected, whether you're competing for their water, what happens to emergency services and roads, and whether their children will actually have opportunities created by the investment.
At LAMP, we're trying to address those questions in the physical design of the campus.
We're developing our own behind-the-meter power generation. We're planning large-scale rainwater harvesting and have identified the potential to capture more than 500 million gallons of rainwater based on the site's characteristics. We're also planning a fire station and support for local emergency-response capabilities.
And we want the relationship to extend beyond infrastructure. We plan to support STEM education and local skills development, so young people in the surrounding area can potentially build careers around the industries coming to the campus.
To me, that's a much healthier relationship between a major infrastructure project and its community.
Tom’s Guide: With so many communities imposing moratoriums or permitting pauses on new data centers, should the industry see those moves as an obstacle or as a signal that the way they build AI infrastructure needs to change?
Tammabattula: It's a signal. If communities across America are raising similar concerns about power, water and infrastructure, we shouldn't dismiss that as opposition to technology. We should change the development model.
Imagine if a major AI campus came into a community with its own power-generation strategy, its own water-management strategy, investment in emergency services, workforce development and a long-term commitment to the community.
That's a very different conversation from simply asking a utility for hundreds of megawatts and asking the community to absorb the consequences.
Moratoriums should force our industry to build better projects, not simply find easier places to build them.
Tom’s Guide: When you talk about "Responsible Compute," what does that actually look like in practice?
Tammabattula: For me, Responsible Compute is very practical. If you're going to consume enormous amounts of power, participate in creating that power. If you're going to use water, develop a responsible water strategy. If your project changes the demands placed on local infrastructure and emergency services, invest in those services.
And if you're bringing billions of dollars of technology into a community, create a pathway for the people who live there to participate in that economic opportunity.
That's the philosophy behind what we're developing at LAMP.
Our model includes dedicated behind-the-meter generation designed to minimize reliance on community grid capacity. We're planning significant rainwater harvesting, with the site's rainfall potentially providing more than 500 million gallons of recoverable water. We're planning investment in fire and emergency-response capabilities. And we want to fund STEM and skills-development programs that can connect local students and workers with careers at the innovation campus.
Responsible Compute isn't a certification or a slogan for us. It's the idea that the infrastructure supporting AI should strengthen the place where it is built rather than simply consume its resources.
Tom’s Guide: Do you think data-center developers can realistically earn community trust before construction begins rather than trying to win people over after a project is already announced?
Tammabattula: Yes, but trust doesn't come from a press release. It comes from giving people specific answers.
Where is your electricity coming from? Where is your water coming from? What are you doing about emergency response? What happens to local infrastructure? What opportunities are you creating for people who already live there?
We've learned through our own experience at LAMP that developers need to answer those questions much earlier. We've also learned that listening matters. Sometimes community concerns identify things that should have been part of the development plan in the first place.
Community engagement shouldn't be something you do after designing the project. It should help shape the project.
Tom’s Guide: As AI dramatically increases the electricity demand, what do policymakers and data-center developers need to change to prevent AI growth from putting additional strain on the national grid and local communities?
Tammabattula: We have to change the assumption that every new gigawatt of compute can simply become another gigawatt of demand on the existing grid. At some point, the math doesn't work.
Large AI infrastructure developers increasingly need to become part of the energy solution. That's one reason we're pursuing behind-the-meter generation at LAMP. Instead of starting with
"How much power can the grid give us?" we started thinking about "How do we build the power infrastructure required for what we're creating?"
I think you're going to see much more convergence between energy development and data-center development because AI is making the two inseparable.
Policymakers should encourage that. If developers bring significant new generation, transmission or other infrastructure with their projects, there should be regulatory pathways that recognize and accelerate responsible development.
Tom’s Guide: If America wants to remain the global leader in AI, what happens if we continue to focus on building data centers as quickly as possible instead of building them responsibly?
Tammabattula: Eventually, speed becomes the thing that slows you down.
If every major project creates another fight over electricity, water and community resources, we're going to see more moratoriums, permitting delays and political opposition. America absolutely needs to build AI infrastructure quickly. I don't think speed and responsibility are opposites. In fact, I think responsibility ultimately allows us to move faster.
Build the power with the compute. Develop the water strategy with the campus. Invest in the community infrastructure your project affects. Develop local talent so the economic opportunity doesn't simply pass through the community. That's the model we're trying to demonstrate at LAMP.
If America can prove that the next generation of AI infrastructure can generate its own power, responsibly manage resources and create lasting economic opportunity for the communities where it operates, that becomes a competitive advantage for the United States.
Responsible Compute isn't about slowing AI down. It's about building the infrastructure that allows America to keep scaling AI without eventually running into a wall.
Follow Tom's Guide on Google News and add us as a preferred source to get our up-to-date news, analysis, and reviews in your feeds. Subscribe to Tom's Guide on YouTube and follow us on TikTok.
More from Tom’s Guide
- 'Raising the standard': Inside Talkspace's bold new AI mental health support tool
- 'Not a chatbot on the sidelines': Ridley AI wants to automate your home sale and eliminate 6% commission fees
- 'Used right, AI is a patient tutor. Used wrong, it's a shortcut around the learning': we spoke to the creator behind an AI tool that's changing the way students learn
