Saad Ullah
⬤ Data center construction is hitting overdrive as Amazon and Nvidia push forward with massive infrastructure projects to keep up with exploding AI demands. More than 7 gigawatts of fresh capacity is set to go live in 2025, with another 10 gigawatts starting construction. The current wave includes huge multi-gigawatt facilities from AMZN, NVDA, Meta, OpenAI, and other major players.
⬤ The biggest projects are genuinely massive. OpenAI's building a 1.2-gigawatt Stargate data center in Texas. Meta's Louisiana complex, called Hyperion, will hit 5 gigawatts and cover space comparable to a big chunk of Manhattan. In Abu Dhabi, G42 and OpenAI are working on a 5-gigawatt campus spanning 10 square miles. Amazon's Indiana site will eventually house over 30 data centers at 2.2 gigawatts, while xAI is putting up a 1-gigawatt AI training facility in Tennessee. Nvidia and Microsoft are backing a 1.2-gigawatt operation in Portugal called Start Campus.
⬤ These numbers reflect how hungry AI is for power. US utility power going to data centers should hit 61.8 gigawatts by year-end 2025, up 11.3 gigawatts from last year. Worldwide, hyperscale data centers doubled to 1,136 in just five years through 2024. With AI models getting bigger and needing more computing muscle, companies like AMZN and NVDA are competing hard for land, energy deals, and long-term power contracts.
⬤ This buildout represents a real shift across tech, energy, and infrastructure. Multi-gigawatt facilities have become critical to the AI ecosystem, driving demand for power generation, grid improvements, semiconductors, and reshaping regional economics. As these giant AI data centers keep multiplying, success for AMZN and NVDA will depend on who can deploy the most compute power fastest.
Saad Ullah
