AI's Energy Crisis: More Power, Less Silicon

AI's energy demands are outpacing grid capacity, with Goldman Sachs forecasting US datacenter power needs to reach 66 GW by 2027. Envision's Mission Gobi aims to build 5 GW of green AI computing in deserts by 2030, while SpaceX explores orbital datacenters powered by solar energy. The AI race is shifting from silicon to sustainable energy as the key bottleneck.

AI's Energy Crisis: More Power, Less Silicon AI's biggest hurdle isn't silicon anymore, it's power. As datacenter energy demands skyrocket, new strategies like Envision's Mission Gobi aim to harness renewable resources to fuel AI's future. The AI race is no longer defined by who has the best semiconductors or the most GPUs. It's all about energy now. As we've poured billions into datacenters, the key bottleneck for AI isn't compute /glossary/compute capability, but the power to run it all. This shift was the focus of Envision CEO Lei Zhang at VivaTech in Paris this June. AI is undergoing an energy revolution, much like the industrial revolution of old, but this time, it's electricity turning the wheels of innovation. Power Hungry AI Goldman Sachs forecasts US datacenter power demand to hit 31 GW by 2025, soaring to 66 GW by 2027. A staggering rise, considering only about 72% of planned facilities make it on time due to electricity shortages. The International Energy Agency reports datacenters used 1.5% of the world's electricity in 2024, likely doubling by 2030 as AI's thirst for power triples. While AI models evolve at breakneck speed, our power grids remain stuck in the past, incapable of keeping up with the tech-driven demands. As AI server power density shot up elevenfold from 2020 to 2025, and expected to quadruple again by 2027, the question isn't just about keeping the lights on. Who's footing the bill for all this power, and what's the societal cost? Communities worldwide are wondering whether AI should hog the same electrical resources vital for homes, hospitals, and industries. Mission Gobi: AI Chasing the Sun Enter Mission Gobi, Envision's ambitious plan to build 5 GW of green AI computing capacity in deserts by 2030. The premise flips the usual script: let AI follow energy, not the other way around. Deserts, abundant in solar and wind potential and scarce in competing demands, offer an enticing solution. Rather than sucking power from existing infrastructure, Mission Gobi seeks to craft new renewable systems solely for AI. Utility, not hype. That's the point. In Chifeng, Inner Mongolia, Envision already runs a fully renewable 2 GW system, integrating wind, solar, and computing in real time. It's a glimpse into what energy-native AI infrastructure could look like. Retrofitting old city grids for gigawatt AI loads is a complex task. Why not start fresh where power is cheapest and most plentiful? The Future of AI: Beyond Silicon Even Elon Musk's SpaceX is eyeing energy as AI's main hurdle, envisioning orbital datacenters powered by endless solar energy in space. Both Envision and SpaceX recognize a shared truth: the future of AI isn't just in more powerful chips. It's in securing the energy to fuel them. If the industrial age was about coal, and the electrical age about power grids, the AI age hinges on abundant, sustainable energy. The question isn't solely about where the next great AI model will be built, but where the power to sustain it will come from. The next chapter of AI won't just be written in silicon, but in solar panels and wind turbines. On-device AI isn't coming. It's here, and it's hungry for power. Get AI news in your inbox Daily digest of what matters in AI.