While concerns about AI driving an energy crisis have gained traction, the reality may be less dire than some fear—but still poses serious risks to the U.S. power grid and climate goals. Tech companies are indeed consuming more electricity to train and run AI models, prompting utilities to plan for massive expansions in gas-fired power plants and transmission infrastructure. However, much of this projected demand could be based on speculation rather than solid, long-term need. The surge in data center energy use is real. Traditional data center racks use about 6 to 8 kilowatts—similar to the power of three homes. But AI workloads require far more powerful hardware, with high-density racks consuming 80 to 100 kilowatts—equivalent to the power of 80 to 100 homes. That’s like powering a small town with a single rack. This spike in demand has disrupted decades of flat electricity consumption, driven by energy efficiency improvements. Now, utilities are rushing to build new gas infrastructure. Between January 2023 and January 2025, proposed gas capacity jumped 70%, largely fueled by data center growth. If all these projects go forward, the U.S. gas fleet could grow by nearly a third—undermining the Biden administration’s goal of a carbon-free grid by 2035. Yet, the scale of future demand remains uncertain. Many developers are making speculative requests for power, sometimes before securing funding or customers. Some even approach multiple utilities to get competing quotes, leading to inflated and potentially double-counted demand forecasts. According to Vistra Energy CEO Jim Burke, proposed projects could be overstated by three to five times what will actually materialize. This creates a dangerous cycle: building too much infrastructure risks stranded assets—expensive facilities that sit idle, burdening consumers with higher bills. Meanwhile, over-reliance on gas undermines climate progress, locking in fossil fuel use just as the world needs to decarbonize. Experts from As You Sow and the Sierra Club warn that without better transparency and planning, the AI boom could derail clean energy transitions. They recommend that utilities require developers to disclose how many utilities they’re approaching, enforce long-term contracts, increase nonrefundable deposits, and charge cancellation fees. Tech companies also have a responsibility. While giants like Amazon, Meta, and Google have long invested in renewable energy, they must now prioritize sustainability alongside AI ambition. By committing to renewable-powered data centers and backing new solar and wind projects, they can help offset fossil fuel dependence—even in the face of weakened federal incentives under the Trump administration. The path forward isn’t about rejecting AI’s potential, but managing its energy impact with foresight, accountability, and a commitment to clean energy. Without deliberate action, the promise of AI could come at a steep environmental and economic cost.