AI Data Centers

AI data centers cluster in power-rich, fiber-connected US regions like Virginia, Texas , California, Arizona, Georgia, and emerging spots in Ohio, Pennsylvania, and the Midwest. This page lets users pinpoint exact locations of public companies' AI data centers using satellite imagery, ideal for tracking construction progress.

The US hosts over 5,000 total data centers (with thousands more planned or under construction), dominating globally. While not all are purely AI-focused, the AI boom is driving rapid hyperscale/HPC growth, and many companies are converting existing sites to support these workloads. Currently we are tracking across the United States and Canada.

Hyperscalers are massive cloud service providers like Amazon Web Services (AWS), Microsoft Azure, Google Cloud, Meta, and Oracle that operate enormous, highly scalable data centers to deliver vast computing, storage, and AI/HPC resources on demand, driving much of the AI data center expansion with their huge power and infrastructure needs. Many hyperscalers are now striking major long-term deals with former Bitcoin miners turned AI data center operators such as IREN (Microsoft), CIFR (AWS and Fluidstack), APLD (CoreWeave), WULF (Fluidstack and Core42), HUT (Anthropic and Fluidstack), and CORZ (CoreWeave). These deals typically provide power, cooling, and colocation for GPU/HPC workloads in exchange for secured multi-year revenue (often billions in contracted value), prepayments, financial backstops to de-risk financing, and phased buildouts.

A large AI data center, typically one with a power capacity of 100 megawatts (MW) or more, can require the same amount of power as that of hundreds of thousands of homes. Such a facility can consume 300,000 to 5 million gallons of water per day primarily for evaporative cooling, with US data centers collectively using billions of gallons annually and projections for hyperscale/AI facilities reaching 16-33 billion gallons yearly by the late 2020s. Innovations like closed-loop, air, or immersion cooling can reduce this significantly.

The following is how we categorize different capacities:

• Energized capacity refers to infrastructure that is fully built, powered, and actively delivering compute workloads. This includes facilities currently running AI training, inference, or high-performance computing services. Some companies describe this as "operating" capacity in their disclosures.
• Under development capacity represents projects that are in active construction or build-out. These sites typically have announced timelines, capital commitments, and expected energization dates, but are not yet live.
• Pipeline capacity includes power agreements, land control, or secured sites that a company has contracted or announced but has not yet begun developing. These projects may move forward depending on financing, customer contracts, or regulatory approvals.
• Total portfolio capacity is the combined sum of energized, under development, and pipeline capacity. It reflects the full scale of a company’s AI and HPC data center footprint, including both active infrastructure and planned expansion.

We update capacity figures whenever companies release new operational disclosures. In most cases, updates occur following quarterly earnings, SEC filings such as 10-Q, 10-K, or 8-K reports, investor presentations, or major hyperscaler and enterprise contract announcements.

Our approach is event-driven rather than calendar-based. When a company revises energized megawatts, signs a new AI hosting agreement, adjusts development timelines, or expands its power portfolio, we update the data accordingly.

All information is mainly sourced directly from primary company disclosures. This includes official press releases, investor presentations, earnings transcripts, and regulatory filings.

Several publicly traded companies operate or are building AI data centers, many of which are former Bitcoin miners that have pivoted their power and cooling infrastructure toward high-performance computing and AI workloads. Notable examples include:

• Core Scientific (CORZ), one of the largest AI/HPC data center operators with major CoreWeave hosting contracts
• IREN (IREN), which secured a significant Microsoft Azure hosting deal for AI workloads
• Applied Digital (APLD), which is building next-generation AI data centers with CoreWeave as an anchor tenant
• Cipher Mining (CIFR), which is expanding into HPC with AWS and Fluidstack partnerships
• WULF (WULF), which is pivoting infrastructure toward AI with Fluidstack and Core42 contracts
• Hut 8 (HUT), which is diversifying into AI infrastructure with Anthropic and Fluidstack hosting

Use the portfolio comparison chart above to see each company's energized, under development, and pipeline capacity in megawatts.

We make every effort to maintain accurate and current data on AI and HPC data center information. However, companies often use different terminology when describing energized, operating, contracted, secured, or potential megawatts. Reporting standards are not uniform across the industry, and development timelines or capacity targets may change.

While we continuously monitor updates, discrepancies or timing gaps may occur due to delayed filings or revised corporate disclosures.

All information on this platform is provided for informational purposes only. It should not be considered investment advice, financial advice, or a substitute for reviewing original company filings. Users should conduct independent due diligence and consult professional advisors before making capital allocation decisions.

For related operational metrics, see our hash rate data and BTC production pages. Compare hyperscaler deal economics across companies, model lease terms with the colocation deal simulator, or upgrade to a Professional plan for satellite imagery of every tracked facility.