Whole Truths and Half Stories

Faryal Akbar, Physiology and Neurobiology Undergrad ’27

I’ve spent most of COP31 going from one AI and data center talk to the next. My takeaway is the biggest problem in this debate isn’t the technology. It’s that we’re all in separate rooms, agreeing with people who already agree with us.

In one room, AI is the hero. Panelists described AI balancing the grid, bringing more renewables online with battery storage, and helping companies cut fuel use and emissions. The research partly backs them up. The IEA estimates AI grid tools could unlock up to 175 gigawatts of transmission capacity, more than data centers’ added needs through 2030 (IEA, 2025). Turkey launched the Antalya AI Pledge here, calling for data centers to run on clean power and efficient water systems. One speaker even argued that AI’s energy use isn’t the real problem, and that food corporations and fast fashion do far more damage.

In the other room, opposite. Frontline communities from Louisiana described facilities built on wetlands and in low-income neighborhoods that get the pollution but not the jobs, while elected officials sign NDAs instead of answering questions. A Harvard study released this week found that 403 U.S. HDCs ran on 54% fossil power and emitted 25.7 million metric tons of CO2 in a year (Guidi, 2026). In Memphis, xAI’s proposed gas turbines could cause $30 to 44 million in health damages a year (SELC, 2026).

My hot take is both rooms are telling the truth, and neither is telling the whole story.

The pro AI crowd talks about the grid of 2035 while skipping the communities living next to gas turbines today. Even their own projections show the tension, in the IEA’s high-growth scenario, data center electricity use reaches about 1,637 terawatt hours by 2035 (IEA, 2026). And China, held up as the AI leader, still got 58% of its electricity from fossil fuels in 2025 (Ember, 2026).

Data centers and the global south panel at COP31But the anti-AI crowd leans on misconceptions and fearmongering as well. The viral “bottle of water per prompt” claim covered a whole conversation and mostly counted water at power plants, real estimates range from 0.26 to 45 milliliters per prompt (CASW, 2025). Where a data center is built matters far more than any per prompt number. Globally, data centers are projected to stay under 1.5% of energy-related emissions through 2035 (IEA, 2025) which is real and growing, but not the main driver of climate change. They add to electric bills where demand is concentrated, but they aren’t the main reason prices are rising (Cybernews, 2026).

Even the Harvard study couldn’t separate AI from ordinary cloud computing. Shaming people for typing prompts misses the point too, since one person’s usage barely registers next to decisions about siting and power contracts.

The real question isn’t whether AI is good or bad. It’s who decides where these facilities go, what powers them, and who pays. Keep data centers away from coal plants and pair them with renewables and storage. Require companies to disclose their energy and water use. And bring frontline communities in before the permits are signed, not after the turbines are running.

Environmentalists are good at rallying people who already agree with us, but that won’t decide where the next wave of data centers gets built. This debate doesn’t need more people cheering in their own room. It needs people willing to walk into the other one.