Here's a closer look at the largest AI construction site in the world

Crusoe Energy Systems is developing an 826-acre data center in Abilene, Texas, leased to Oracle to support OpenAI's AI compute demands, potentially housing 400,000 Nvidia GPUs and drawing 1.2 gigawatts of power; UBS analysts estimate the initial phase could generate over $2 billion in annual revenue for Oracle, with Nvidia potentially seeing up to $20 billion in GPU revenue from the deployment, though challenges remain regarding power distribution, component procurement, and labor shortages. While frequently referenced as "Stargate 1", analysts believe that Stargate branding is currently aspirational, as funding to date has come from a separate joint venture.

Analysis

A significant artificial intelligence infrastructure project, potentially the world's largest AI construction site, is underway in West Texas, developed by Crusoe Energy Systems and leased to Oracle to support OpenAI's compute requirements. This 826-acre Lancium Clean Campus is planned to house 400,000 Nvidia GPUs across eight buildings, consuming 1.2 gigawatts of power at full capacity. UBS Global Research, after a March site visit with over 2,200 workers present, reports the first two buildings (200 megawatts, ~100,000 GPUs) are expected to be powered up in the coming months, with a phased ramp-up into 2026 and the facility aiming to go live in late 2025; the full GPU cluster is anticipated to be operational within 12 months thereafter. UBS estimates this initial phase could generate over $2 billion in annual revenue for Oracle, with Nvidia potentially realizing up to $20 billion in GPU revenue from the entire Abilene deployment, and total project hardware and networking demand reaching $25 billion. Oracle may begin recognizing revenue from this project as early as its fiscal second quarter ending November 2025. While referred to as "Stargate 1," a nod to a $500 billion U.S. AI initiative involving OpenAI, Oracle, and SoftBank, UBS notes current funding is from a separate joint venture (Crusoe, Blue Owl Capital, JPMorgan) and views the Stargate branding as aspirational pending resolved capital and structural commitments. The project faces challenges including delays from faulty transformers, labor shortages necessitating out-of-region electricians, and long lead times for high-voltage transformers. Crusoe Energy has pivoted from bitcoin mining to become a vertically integrated builder of GPU data centers. The site’s remote location is deemed more suitable for model training due to latency, but the infrastructure's minimal customization allows for flexibility. Contractors remain confident in completing the first phase within months and scaling up through 2027.

Key Decisions for Investors

• Investors should monitor Oracle's execution on this large-scale data center project, as its successful and timely completion could significantly boost revenue, with over $2 billion annually projected from the initial phase, and affirm its capability in supporting major AI workloads.
• The deployment of 400,000 GPUs, potentially generating up to $20 billion in revenue for Nvidia, reinforces the sustained, high-volume demand for its AI chips; investors should consider this an indicator of continued strength in Nvidia's data center segment.
• Given the project's scale and reported challenges in power distribution, component procurement, and labor, investors should track its progress as a bellwether for the broader AI infrastructure build-out, which could influence sentiment and valuations across the AI supply chain.
• While the "Stargate 1" branding is currently aspirational, any concrete developments or funding allocations from the larger $500 billion Stargate initiative towards this or similar projects could present further upside for involved parties like Oracle and should be closely watched.