How AI's Data Appetite Is Fueling the Data Center Boom

The scale of AI's hunger for compute is reshaping the data center industry into one of the most capital-intensive buildouts in history - and two recent developments illustrate exactly why.

Elon Musk's AI company, SpaceXAI, is laying the groundwork for at least one new large-scale data center in Texas, according to The Information, in a move that would expand its AI computing capacity beyond its existing Memphis hub as it seeks to become a major cloud provider. The company has recently signed agreements to lease large portions of its compute capacity to customers including Anthropic and Google, turning what was once purely an internal AI-training asset into a new revenue stream. The Texas expansion could match or exceed the scale of SpaceXAI's existing facilities, which house around a gigawatt of compute capacity and hundreds of thousands of Nvidia GPUs. 

That single detail  a gigawatt of power for one data center campus captures the core dynamic driving this entire sector. AI workloads, especially large model training and high-volume inference, consume electricity and compute at a scale traditional enterprise IT never approached. Meeting that demand means data centers are no longer just buildings full of servers; they're increasingly becoming power-plant-adjacent infrastructure projects, which is exactly where the economics start to get complicated.

Oracle's experience shows the other side of that coin. The company, which committed to a staggering $165 billion long-term investment plan, is now grappling with the reality that pouring concrete and pulling power lines at unprecedented scale comes with unprecedented problems, particularly at its AI megacampus projects in Wisconsin and near El Paso, New Mexico, where cost overruns are ballooning well beyond initial projections. Oracle's capital expenditures jumped from $21.2 billion in fiscal 2025 to $55.7 billion in fiscal 2026, with fiscal 2027 spending projected between $90 billion and $95 billion. Power constraints sit at the top of the risk list, since securing dedicated grid capacity in regions like central Texas and the Midwest is neither fast nor cheap, and environmental approval obstacles recently forced a change in the power supply plan for one project, causing costs to surge by several billion dollars. 

Together, these two stories reveal the same underlying force from opposite directions: AI's insatiable demand for data is pulling hyperscalers and AI labs alike into a race to build gigawatt-scale infrastructure, while the physical and regulatory limits of the real world land, power grids, permitting, financing  are pushing back just as hard. The data center business isn't just growing because AI adoption is rising; it's growing because AI has changed what a data center fundamentally needs to be.