CoreWeave Inc. has announced plans to invest up to US$6 billion in a data center in Lancaster, Pennsylvania.

The facility will start with an initial capacity of 100 megawatts, with the potential to expand to 300 megawatts, according to CEO Michael Intrator.

The project is expected to create 600 construction jobs and 175 permanent positions.

It will feature Nvidia’s latest chips and aims to support large-scale AI models in the Mid-Atlantic region.

The first phase of 100 megawatts is projected to be operational by 2026.

CoreWeave, which recently agreed to acquire Core Scientific Inc. for US$9 billion, is expanding its data center network due to rising demand from cloud providers and AI developers.

🔗 Source: Bloomberg

🧠 Food for thought

1️⃣ Data centers are experiencing unprecedented power scaling

CoreWeave’s plan to build a 300-megawatt facility illustrates a dramatic shift in data center scale that’s transforming the industry.

Just three years ago, as CoreWeave’s CEO notes, 10-megawatt facilities were standard, but today’s AI workloads are driving conversations toward gigawatt-scale operations, facilities that consume electricity equivalent to powering 750,000 homes.

This trend is documented industry-wide, with Deloitte research showing AI data centers require 10 times more energy per square foot than traditional facilities, potentially increasing from 5 to 50 megawatts for similar-sized spaces 1.

Goldman Sachs forecasts a 50% increase in global power demand from data centers by 2027 and a 165% increase by 2030, directly attributed to AI computing requirements 2.

The largest US data centers currently draw less than 500 megawatts, but future projects like CoreWeave’s are pushing capacity boundaries, with some planned facilities requiring up to 2,000 megawatts 1.

2️⃣ Power infrastructure has become the critical bottleneck for AI advancement

CoreWeave’s strategic Lancaster location represents a crucial element of AI infrastructure planning: securing reliable power access.