More than a year after former President Trump’s announcement of the ambitious $500 billion “Stargate” AI data center initiative at the White House, the project has reached an impasse. OpenAI, initially planning to construct its own data centers, has shifted its strategy, opting instead to partner with existing cloud service providers to address urgent computing power demands.
Informed sources indicate that the joint venture involving OpenAI, Oracle (ORCL.US), and SoftBank Group (ADR) (SFTBY.US) has yet to finalize staffing or initiate the development of any data centers. Mere weeks after the project was announced, the three parties encountered significant disagreements over leadership, roles, and operational framework, resulting in a stall of the “Stargate” initiative.
Facing a critical shortage of computing power, OpenAI had originally envisioned constructing extensive data center campuses to lessen its dependency on cloud providers. However, this plan faltered when potential lenders declined to extend billions in financing to a company operating on an untested business model that incurs annual losses in the tens of billions. Only after grappling with these financial realities did OpenAI seek to advance the “Stargate” project with its partners.
This development has put considerable strain on OpenAI’s financial standing. Last year, the company’s gross profit margin fell short of projections, as it was forced to procure costly computing resources on short notice. OpenAI has since adjusted its forecast for computing power expenditures from $450 billion to $665 billion by 2030. Compounding matters, competitors like Google and Anthropic have solidified their positions in the market over the past year.
Disagreements among the three companies have significantly hindered progress. When Trump unveiled the “Stargate” project in January 2025, the companies pledged to expedite its development, with an initial investment of nearly $100 billion earmarked for constructing 10 gigawatts of computing capacity. However, according to insiders, the initiative faced challenges soon after its launch, lacking effective leadership and coordination. OpenAI, Oracle, and SoftBank struggled to agree on responsibilities and the overall collaboration framework.
Proposals surfaced to create an independent entity to manage the construction and leasing of facilities back to OpenAI, potentially transferring some employees to this new organization. Other discussions considered whether “Stargate” could function as a vehicle for financing chips and infrastructure. However, none of these concepts materialized.
OpenAI’s initial goal of securing a 10-gigawatt capacity commitment from SoftBank and Oracle over three years did not come to fruition. Instead, as the “Stargate” negotiations progressed, OpenAI entered into computing power agreements with alternative suppliers like Amazon Web Services and Google Cloud to fulfill its computational requirements.
After the “Stargate” announcement, OpenAI explored the possibility of self-constructing data centers. Its teams scoured the country for suitable sites capable of supporting campuses with capacities between 800 megawatts and 1.2 gigawatts, focusing on locations with the potential to provide ample power by 2026 and 2027. However, as costs were assessed, executives recognized the necessity of securing substantial financing, leading to the conclusion that leasing arrangements with stronger credit tenants like Oracle could significantly reduce financial burdens.
Conversations with Oracle resumed in the spring, culminating in a new agreement for a major data center project that would see a joint venture develop 4.5 gigawatts of capacity. This collaborative framework enabled the two companies to distribute risks across multiple campuses rather than pursuing individual projects one by one. Shortly thereafter, OpenAI suspended its independent data center plans, transferring initial projects to Oracle for development.
By the end of July 2024, Oracle and OpenAI announced a deal to develop 4.5 gigawatts of data center capacity across various U.S. locations. This rare arrangement allowed both parties to share economic risks, meaning they would both bear the costs associated with delays or overruns and also benefit from any savings if the project costs were less than anticipated.
Post-announcement, OpenAI employees sought to revive their independent data center ambitions, with particular focus on a crucial 1-gigawatt campus located between Milam County, Austin, and Waco, Texas. However, negotiations with Oracle led to further delays, resulting in the loss of a general contractor on the project. Consequently, OpenAI turned to SoftBank for collaboration, announcing a joint development plan with SoftBank Energy, a subsidiary of SoftBank Group.
Despite aspirations for the Texas campus to become its first self-built facility, clashes emerged over control issues, with OpenAI asserting its desire for ownership while SoftBank aimed to develop and retain control of the project. Negotiations spanned months, including multiple trips to Japan to reconcile differences with SoftBank’s CEO, Masayoshi Son. Ultimately, a compromise was struck: OpenAI would manage the facility design while SoftBank Energy would handle development and ownership, marking a significant shift in OpenAI’s immediate infrastructure goals.
Construction began in October, with the agreement allowing OpenAI design control without requiring direct funding or ownership of the project. This arrangement covers decisions related to key technical aspects such as cluster architecture, cooling systems, and power infrastructure.
OpenAI Chief Financial Officer Sarah Friar underscored this strategic pivot during her remarks at the World Economic Forum in Davos, emphasizing that leveraging cloud partners keeps the company’s balance sheet light. Although OpenAI maintains its intention to develop its own data centers, this has receded as a near-term priority. Instead, the company is now focusing on building a series of cloud and infrastructure agreements under the “Stargate” initiative, allowing it to retain special permissions regarding data center design.
The evolution of the “Stargate” project alongside OpenAI’s recognition of its current limitations coincided with significant changes in its infrastructure organization. In November, OpenAI appointed Sachin Katti, formerly of Intel, to lead its infrastructure team, reporting directly to President and co-founder Greg Brockman. As the industry landscape continues to shift, OpenAI’s adaptive strategies may play a critical role in its future trajectory in the competitive AI landscape.
See also
Getty Images Wins Limited Trademark Infringement Case Against Stability AI
Germany”s National Team Prepares for World Cup Qualifiers with Disco Atmosphere
95% of AI Projects Fail in Companies According to MIT
AI in Food & Beverages Market to Surge from $11.08B to $263.80B by 2032
Satya Nadella Supports OpenAI’s $100B Revenue Goal, Highlights AI Funding Needs
