AI Growth Drives $200 Billion Data Center Costs and 50 Gigawatts Power Need by 2030, xAI’s Colossus Consumes 300 MW

Epoch AI just dropped a map of the world’s 500+ AI supercomputers. • Performance doubling every 9 months. • Hardware cost & power use doubling every year. • xAI’s Colossus already gulps 300 MW — equal to 250 k homes — and that’s only 2025. If the trendline holds, the 2030 https://t.co/OBVFyBLnaB https://t.co/9w9oDblz8W

AI’S MEGA BRAINPLANTS COULD COST $200B AND SUCK POWER LIKE A SMALL NATION A new study says the future’s smartest machines might need nuclear-reactor-level juice to stay online. By 2030, the top AI data center could pack 2 million chips, cost $200 billion, and draw 9 https://t.co/PPYcxkjX6d https://t.co/919UrJcR0N

Data centers play a critical role in our lives. But AI has turbocharged the need for bigger data centers, putting a growing demand on the existing power grid. So what happens when the digital backbone of our lives needs more power than the world can offer? https://t.co/zZwR60QAsl https://t.co/iJDTqJ4dZH

Anthropic co-founder Jack Clark said 50 gigawatts of new power capacity will be needed by 2027 to support AI. That is the equivalent of about 50 new nuclear plants. I’ve done some calculations to see what this means for Nvidia.

The energy required to run data centers will total Japan’s entire energy grid by 2030, but I wouldn’t panic. AI can slash emissions by discovering ways to optimize systems, giving rise to more renewable and efficient energy production.