Artificial Intelligence drives massive electricity demand and corporate token cost controls

The International Energy Agency (IEA) warns that the rapid expansion of artificial intelligence will push data‑center electricity use to about 3 % of global consumption by 2030 – roughly the same amount of power Japan uses in a year. The agency estimates nearly $2 trillion will be invested in digital infrastructure through 2030, with about $400 billion earmarked for new energy generation, grid upgrades and storage. The United States is expected to account for almost half of the increase in electricity demand linked to AI, surpassing the consumption of traditionally heavy industries such as aluminum, steel and cement.

At the same time, Microsoft’s executive vice‑president Jay Parikh sent an internal memo urging engineers to curb “token‑maxxing,” the practice of using excessive AI tokens that can generate costly bills. New internal guidelines set token‑budget targets for divisions starting July 2026 and promote the use of the more economical GPT‑5.6 model for internal projects, aiming to keep monthly token expenses – which can run from hundreds to thousands of dollars per employee – under tighter control.