the sheer scale¶

I think people struggle to grasp the sheer scale of our energy usage.

In 2022, global primary energy consumption was ~180,000 TWh. Data centers consumed 460 TWh, or ~0.25%.

With AI, that number is projected to hit 1,000 TWh by 2026. That'll be 0.5% of the projected 192,000 TWh. https://t.co/9ktUVnIHvu That figure, by the way, projects AI contributing 20% of the growth of data energy usage, with traditional data centers accounting for most of the rest. It is more energy-intensive, but in terms of scale, it still has a lot of ground to cover before it overtakes anything. If these worst case trends hold steady, AI will eventually outpace traditional data centers. However, total energy consumption will continue to grow -- slower than the rate for data, but again, data would have to make up quite a bit of ground to even get close. Agriculture, transportation, manufacturing, construction, and mining remain the driving forces of global energy usage. Residential and commercial electricity is growing rapidly, but until AI can harvest resources from the ground, it will never come close to primary production. 2.5% growth of an astronomical number is still much, much more than 10% or even 20% growth of a tiny number. The world will be slow-roasted long before AI is even in the top 10 heating elements.