Musk's strategic bet on power is not a spur-of-the-moment decision; it was clearly stated at the Davos Forum in February 2026 that "the fundamental constraint on AI deployment is power," followed by a complete chain of actions including acquisition, deployment, and expansion within three months.

On the capital path, xAI adopts a "vertical integration" model: from chip procurement (bulk orders of NVIDIA H100/H200) to power acquisition (APR Energy's 1GW mobile power plants) to data center construction (Memphis + Texas dual bases), forming a closed loop of "computing power - power - site," similar to Tesla's logic of building super factories in the 2010s.

Comparing to Google's acquisition of renewable energy developers in 2018 and Amazon's commitment to 100% renewable energy in 2020, the competition for power among tech giants has upgraded from "green PR" to "strategic resource control." xAI is currently in the "early expansion" stage, prioritizing power supply over cost optimization.

Essentially, this represents a restructuring of the industrial chain: AI training is shifting from "relying on the grid" to "building its own power plants," due to the long approval cycles (3-5 years) and high expansion costs of traditional power grids, while mobile gas turbines can be deployed in 30-90 days to meet the immediate demands of exponential growth in AI computing power.

ABAB News · Cognitive Laws

The endgame of AI is not algorithms, but the cost and speed of every kilowatt-hour.
Computing power is superficial; electricity is fundamental. A GPU without power is scrap metal.
From competing for chips to competing for power plants, the next battle for tech giants is energy.