OpenAI Stargate Phase 1 Construction of 200 Megawatts and 980,000 Square Feet

The flagship site for the OpenAI Stargate data center project is currently under construction in Abilene, Texas. The initial phase includes two buildings totaling 980,000 square feet and over 200 MW of power capacity. This phase is slated to be energized in the first half of 2025, setting the stage for subsequent GPU installations.

big. beautiful. buildings. pic.twitter.com/8lAvUufXFm

— Sam Altman (@sama) January 23, 2025

Construction of the second phase has already begun, adding six additional buildings to the campus.

Crusoe announced the second phase of construction, which includes six additional buildings, expanding the facility to eight buildings encompassing approximately 4 million square feet and a total power capacity of 1.2 GW.

Each of the data center buildings is planned to run a single network fabric with up to 50,000 of NVIDIA’s GB200 NVL72s AI GPUs. This expansion is anticipated to be completed by mid-2026, marking a significant milestone in scaling AI infrastructure. ​OpenAI’s Stargate project aims to invest $500 billion in data centers requiring over 5 GW of power, while Meta has proposed a $200 billion data center initiative with similar power needs.

Crusoe Chief Executive Officer Chase Lochmiller said there will be eight data center buildings each designed to hold as many as 50,000 Nvidia GB200 semiconductors.

🔎Until recently, @OpenAI was tightly aligned with @Microsoft. But a series of strategic shifts has changed the dynamic:

The $11.9B, 5-year OpenAI-CoreWeave deal signals a major shift in inference infrastructure.

Oracle + @Crusoe‘s 1.2GW Stargate site in Abilene is now the… pic.twitter.com/gDMv5okvaW

— SemiAnalysis (@SemiAnalysis_) May 8, 2025

Microsoft’s CapEx growth is decoupling from OpenAI’s. The biggest winner? Oracle, with massive forward RPO growth and potential share gain in hyperscale AI.

Microsoft is freezing 1.5GW of near term self-build datacenter projects – projects that were previously scheduled to go online in 2025 and 2026. This is a deliberate move by Microsoft to slow the expansion of its self-built capacity. These cancellations involve non-binding Letters of Intent (LOIs), not firm contracts, correcting earlier misconceptions about the scale of lease terminations. Despite the freeze on new leases, Microsoft retains approximately 5GW of pre-leased capacity under binding contracts, which is set to begin operations between 2025 and 2028. The leasing slowdown primarily impacts capacity planned for 2027 and beyond, rather than 2025 and 2026 directly.

GPU Installations: Summer 2025 and Mid-2026

By the end of summer 2025, the Abilene data center is scheduled to have 16,000 Nvidia GB200 GPUs installed.

The current plan is to deploy 64,000 Nvidia GB200 GPUs by the end of 2026 at the Abilene site. However, the data center complex is being designed with a capacity for up to 400,000 Nvidia GPUs. This means that while the infrastructure will support 400,000 GPUs in the future, the confirmed installation schedule only accounts for 64,000 by the end of 2026. There’s no specific commitment yet to reaching 400,000 GPUs by mid-2026, but the capacity is there for potential future expansion.

The 400,000 figure refers to the total design capacity of the Abilene complex rather than a firm installation target for mid-2026. If OpenAI accelerates its plans or adds more sites, this number could become feasible, but as of now, it’s 16,000 by summer 2025 and 64,000 by the end of 2026 for Abilene.

xAI Memphis Colossus Data Center

The xAI Memphis data center building itself is 1 million square feet (approximately 92,905 square meters). This is the size of the facility located at 5400 Tulane Road in the Whitehaven area of Memphis, which includes an existing warehouse repurposed for xAI’s operations. xAI has purchased 100 acres (about 404,686 square meters) of land for the data center. This includes three parcels at the Tulane Road site, providing space not only for the current 1 million square foot building but also for potential future development.

xAI has leased additional space beyond what they own. They have secured 522 acres (approximately 2.11 million square meters) in the Pidgeon Industrial Park, located near their supercomputer site in southwest Memphis. This 21-year lease, approved by the Economic Development Growth Engine (EDGE), includes options for extensions and supports xAI’s growing operations.

There are firm plans for building expansion at the xAI Memphis data center. The company is scaling its Colossus supercomputer, which currently operates with about 300,000 GPUs which should expand to about 500,000 GPUs in the next two months and possibly to 1 million GPUs around the end of 2025 or Q1 in 2026.

xAI has already installed mobile natural gas generators for about 540 Megawatts to go along with 150 MW of grid power. This is enough power for 500,000 GPUs.

Plans for Other US Locations and Globally
The Stargate project isn’t confined to Abilene—OpenAI has big plans for expansion both within the U.S. and internationally:

OpenAI aims to build five to ten campuses across the U.S., each capable of supporting 1 GW or more of power capacity. The company is exploring options in 16 states, indicating a broad domestic rollout. However, Abilene is the only site currently under construction, with others still in the planning phase.

OpenAI is eyeing locations in the UK, Germany, and France for future Stargate data centers. Through the OpenAI for Countries initiative, the company also plans to collaborate with national governments to develop AI infrastructure worldwide. While these international plans are ambitious, specific details—like GPU counts or construction timelines—haven’t been finalized yet.

Funding Details
The project is part of a planned $500 billion joint venture between OpenAI, Oracle, SoftBank, and Abu Dhabi’s MGX. This massive investment will fund the construction of multiple large-scale data centers across the U.S. and potentially globally over the next four years.

OpenAI recently raised $40 billion in a funding round at a $300 billion valuation. A portion of this capital is earmarked for the Stargate project, adding further financial muscle to the initiative.

Crusoe is indeed leading the construction of the Abilene data center. They’re responsible for designing, building, and operating the campus, which is being developed on land owned by Lancium. Crusoe’s expertise in high-performance computing infrastructure makes them a critical partner.

Oracle: Oracle is heavily involved, overseeing the sourcing and operation of the supercomputer at the Abilene site. They’re also a core member of the $500 billion joint venture.

SoftBank and MGX: These partners provide strategic and financial support as part of the joint venture, though they’re not directly involved in the physical construction.

Crusoe is the boots-on-the-ground construction lead for Abilene, while Oracle ensures the technical integration of the supercomputer systems. The collaboration between these companies is driving the project forward.

Here’s a quick rundown of the latest news on the OpenAI Stargate data center:
Construction: The Abilene, Texas site is underway, with the first phase (200 MW+) energizing in early 2025 and a second phase already in progress.

GPU Installations: 16,000 Nvidia GB200 GPUs by summer 2025 and 64,000 by the end of 2026, with capacity for up to 400,000 in the future (not confirmed for mid-2026).

Global Plans: Expansion to 5–10 U.S. campuses and potential sites in the UK, Germany, and France, plus global partnerships via OpenAI for Countries.

Funding: A $500 billion joint venture (OpenAI, Oracle, SoftBank, MGX) plus $40 billion from OpenAI’s recent raise.